This week’s signals trace a clear trajectory:

from how intelligence is being created,
to how it is being scaled,
to how it is beginning to integrate into everyday workflows —
and ultimately, how it is reshaping the boundaries of creativity, labour, and civilisation itself.

Welcome to MyGeekSpace | Living the Future, where we don’t just track headlines — we map direction.

In this edition:

• AI crosses a cinematic threshold, moving from tool to medium

• Infrastructure scales to unprecedented levels as SpaceX enters the AI arena

• The interface layer evolves rapidly, with AI becoming a true “digital collaborator”

Power up your curiosity and settle into this week’s signals from the expanding frontier.

Support MyGeekSpace

SECTION 1 · THE SIGNAL

The first studio-quality AI-assisted feature film is set to debut at Cannes next month — and its reception may quietly redefine how films are made going forward.

“Bitcoin: Killing Satoshi” blends traditional filmmaking with AI-driven production techniques, signalling a shift not just in efficiency, but in the nature of cinematic creation itself.

What’s happening

• Hybrid production model
  Despite the “AI-generated” label, the film relied on 154 crew members and over 100 actors, combining human creativity with AI-assisted environments and post-production.

• Massive cost compression
  The original concept required filming across 200 locations with a projected $300M budget. AI reduced this to a single soundstage shoot over 20 days, bringing costs down to around $70M.

• AI as world-builder
  Entire environments, locations, and visual continuity were generated or enhanced through AI workflows, enabling scale previously reserved for major studios.

• Creative accessibility expands
  Projects that would never have been financed under traditional models may now become viable.

Why It Matters

This is not just about cheaper filmmaking.

It marks the moment where AI becomes part of the creative substrate, not just a supporting tool.

Cinema, historically constrained by budget, logistics, and physical reality, is beginning to detach from those limits. The barrier between imagination and execution is collapsing.

At the same time, this raises deeper questions:

• Who owns creative output when AI is embedded in the process?

• What happens to traditional film roles and labour structures?

• Does abundance dilute or expand artistic quality?

Closing Thoughts

We’ve seen AI write scripts, generate visuals, and assist editing.

But this feels different.

This is the first time AI is not just helping make films, it is reshaping what filmmaking is.

Cannes will not just be a premiere.

It will be a test of whether audiences are ready for a new creative paradigm.

Source:
Killing Satoshi: $70M Bitcoin Film Will Make Heavy Use of AI
Yahoo Entertainment

SECTION 2 · INFRASTRUCTURE

While creativity evolves at the surface, something deeper is accelerating beneath it: infrastructure.

This week, SpaceX made a bold move into the AI landscape, partnering with Cursor to train elite models on one of the most powerful compute systems ever assembled.

What’s happening

• Cursor × SpaceX partnership
  The AI coding startup gains access to the Colossus supercomputer, reportedly equivalent to around one million H100 GPUs.

• High-stakes strategic play
  SpaceX retains the option to acquire Cursor for $60 billion or instead pay a $10 billion development fee.

• Compute as leverage
  This deal signals that access to compute is becoming the defining competitive advantage in AI.

• Reality check on “Space AI”
  Despite earlier hype around orbital data centres, SpaceX now acknowledges the uncertainty and potential lack of profitability in space-based compute.

Why It Matters

We are entering a phase where:

Intelligence is no longer just designed — it is engineered at scale.

The companies that control compute infrastructure will shape:

• What models are built

• Who gets access

• How intelligence is distributed globally

This is no longer just a software race.

It is an infrastructure war, closer to energy grids or telecommunications than traditional tech competition.

Closing Thoughts

The deeper signal here is simple:

AI is consolidating around power, scale, and capital intensity.

And yet, paradoxically, this same infrastructure is enabling tools that feel increasingly personal and accessible.

The future is being built in massive data centres —
but experienced in a browser window.

Source:
SpaceX x Cursor Partnership Announcement
SpaceX | via X / official communication

Subscribe now

SECTION 3 · INTERFACE

If Section 2 is about power, Section 3 is about proximity.

This week, the interface layer evolved rapidly, with Anthropic and OpenAI both pushing AI deeper into everyday workflows.

What’s happening

• Anthropic enters the workflow layer
  Claude Design and the Word add-in bring AI directly into design, writing, and document editing environments.

• From prompt to product
  Users can now generate prototypes, pitch decks, and marketing assets through conversation — with outputs ready for real-world use.

• “Digital employee” models emerge
  Claude Opus 4.7 is designed for long-running, complex tasks, with improved memory, instruction-following, and multimodal capabilities.

• OpenAI pushes autonomous agents
  Workspace agents can operate continuously in the background, handling workflows, generating outputs, and coordinating across tools.

• Interface becomes invisible
  AI is no longer a destination — it is becoming embedded within existing tools and processes.

Why It Matters

We are moving from:

• AI as a tool you use

To:

• AI as a system you work alongside

The interface is dissolving.

Instead of switching between apps, prompts, and outputs, users are entering a space where:

Thinking, creating, and executing begin to merge.

This has profound implications:

• Work becomes more fluid and less task-based

• Productivity shifts from effort to orchestration

• The definition of “skill” begins to change

Closing Thoughts

The most important shift here is subtle.

AI is no longer asking:

“What do you want me to do?”

It is starting to understand:

“What are you trying to achieve?”

That transition — from instruction to intent- is where the real transformation begins.

Sources:
Claude Design Announcement
Anthropic

Introducing Workspace Agents in ChatGPT
OpenAI

ChatGPT Images 2.0 Overview
OpenAI

SECTION 4 · EMBODIMENT

This week’s “Dose” takes us beyond Earth — and back again.

Artemis II ‘Earthset’ reframes our place in the universe

A new image captured by the Artemis II crew shows Earth slowly disappearing behind the Moon’s horizon — a moment now known as “Earthset”.

What’s happening

• A new planetary perspective
  The image echoes the iconic “Earthrise” photo from 1968, but from a far greater distance.

• Record-breaking human journey
  The Artemis II crew travelled over 250,000 miles from Earth, further than any humans in history.

• A viral moment of reflection
  A simple iPhone video of the event reached millions, resonating far beyond the scientific community.

Why It Matters

Every few decades, an image changes how we see ourselves.

Earthrise did it.
The Blue Marble did it.
Earthset might do it again.

These images are not just scientific artefacts.

They are civilisational mirrors.

They remind us that:

• Borders disappear at scale

• Fragility becomes visible

• Perspective reshapes priorities

Closing Thoughts

In a week dominated by AI, compute, and digital acceleration, this image offers something essential:

Context.

The future is not just about building intelligence.
It is about understanding where we are building it from —
and why.

Source:
Artemis II ‘Earthset’ Coverage
NASA / Associated reporting (NYT, APOD)
Scientific American

SECTION 5 · THE HORIZON

The Moon, sovereignty, and the next geopolitical frontier

Recently, a senior Russian official proposed the creation of “sovereign territories” on the Moon — a statement that directly challenges the foundations of international space law.

What’s happening

• A shift in tone
  Russia signalled its intent to establish a national presence on the lunar surface.

• Legal tension emerges
  The 1967 Outer Space Treaty explicitly forbids national appropriation of celestial bodies.

• A crowded lunar race
  The US, China, and private actors are all accelerating lunar ambitions.

Why It Matters

We are entering a phase where:

Space is no longer just exploration — it is strategy.

The Moon is becoming:

• A testing ground for off-world infrastructure

• A staging point for deeper space missions

• A symbolic extension of geopolitical power

Closing Thoughts

The deeper question is not who gets to the Moon first.

It is:

What kind of civilisation arrives there?

Because the systems we build beyond Earth will reflect the values we carry with us.

And this time, we have the awareness — and the tools — to choose differently.

Source:
Russia Proposes Sovereign Territories on the Moon
Russian Academy of Sciences / TASS

🌀 Final Wrap-Up | Signals from the Expanding Frontier

This week’s trajectory is clear:

• Intelligence is becoming creative infrastructure

• Infrastructure is becoming a strategic power

• Interfaces are becoming invisible collaborators

• Perspective is expanding beyond Earth

• And civilisation itself is preparing to extend outward

💬 Your Thoughts | Join the Living Conversation

Which signal stayed with you most this week?
Where do you feel intelligence is moving next?

✨ Continue the Journey

If this edition resonated, follow the signal here on Substack, explore more at MyGeek.Space, or join me on 𝕏 @frankdasilva for real-time reflections on AI, culture, and future civilisation.

Subscribe now

For deeper insights and early access, you can also join on Patreon.

If you value independent, future-focused writing, consider subscribing to MyGeekSpace (free or paid), or support the work with a one-time contribution below.

Support MyGeekSpace

Futūrum Vīvere.

Each week, we map where intelligence is beginning to live.

At MyGeekSpace | Living the Future, grounded in the Terra 2.0 vision, I track the meaningful signals emerging beneath the daily noise of headlines and hype. This is not a space for fixed predictions, but for orientation: a way of sensing direction while the future is still taking shape.

In this edition, we explore China’s strategic AI pivot after AlphaGo, the intensifying race to make AI wearable through next-generation smart glasses, and Anthropic’s release of Claude Opus 4.7, a major public model upgrade that also reveals a widening gap between what users can access and what remains behind the frontier curtain. We also take a brief detour into creator hardware with GoPro’s new MISSION 1 cinema line, before closing with NASA’s Moon Base blueprint and what it reveals about the shift from exploration to permanent off-world infrastructure.

Power up your curiosity and settle into this week’s signals from the expanding frontier.

→ If this weekly exploration resonates, consider supporting MyGeekSpace on Substack with a free or paid subscription. Your support helps sustain independent, future-focused writing and keep this signal alive. And if these reflections resonate, but you would prefer to support the work without subscribing, you can also make a one-time donation below.

Subscribe now

Support MyGeekSpace

Section 1 | The Great Tech Pivot: How China Redefined the AI Race

The AI race no longer looks like a simple contest over who has the biggest chips or the deepest pockets. Since AlphaGo’s shock victory over Lee Sedol in March 2016, China has treated AI less as a consumer technology trend and more as a national strategic priority, with a formal 2030 leadership goal set out in its 2017 development plan. What is striking now is not merely China’s ambition, but the way it has adapted under pressure: moving from hardware dependence to software efficiency, open models, and data scale.

From AlphaGo to National Strategy

• AlphaGo was more than a symbolic win
  When AlphaGo beat Lee Sedol, it did not just defeat a Go champion. It showed that machine intelligence could produce moves outside established human intuition, with Move 37 becoming the moment that crystallised the shift.

• China interpreted the moment geopolitically
  While many in the West saw AlphaGo as a research breakthrough, Beijing moved quickly to frame AI as a pillar of national power. China’s 2017 New Generation Artificial Intelligence Development Plan set a clear objective: to become a world-leading AI innovation centre by 2030.

• That decision changed the tempo of the race
  From that point onward, AI ceased to be just about labs and start-ups. It became part industrial strategy, part state competition, and part civilisational bet on who defines the next technological order. This is the deeper pivot that still shapes the rivalry now.

The Three-Layer AI Stack

• The hardware layer still matters enormously
  Advanced GPUs, semiconductor tooling, and data centre infrastructure remain the physical foundation of frontier AI. NVIDIA continues to dominate the high-end AI compute stack, while TSMC remains central to advanced chip manufacturing.

• The model layer is where China has become more inventive
  With access to the best chips constrained by export controls, Chinese firms have been pushed to find smarter ways to train and run powerful systems. This is where engineering elegance starts to matter as much as brute-force scale.

• The application and data layer may become decisive
  Models do not grow in a vacuum. They depend on the quality, richness, and volume of real-world data. As AI moves beyond text and into audio, video, and physical-world understanding, data pipelines are becoming strategic assets in their own right.

The Hardware Squeeze Created a Different Kind of Innovation

• Washington tried to slow China at the silicon level
  U.S. export restrictions were designed to limit China’s access to the most advanced AI chips, shifting licence review for top-tier semiconductors into a much tighter geopolitical framework. That has made the hardware race more explicitly political.

• But constraints can create unusual forms of ingenuity
  Rather than waiting to fully match the Western compute stack, Chinese teams have focused on extracting more performance from limited resources. In practice, this has meant treating efficiency not as optimisation at the margins, but as the core design philosophy.

• This is the real pivot
  The narrative is changing from “who owns the best chips?” to “who can do the most with the chips they have?” That is not a complete replacement for hardware power, but it is a significant reframing of the competitive field.

DeepSeek and the Rise of Efficient AI

• DeepSeek became the emblem of this new approach
  Its technical reports describe a Mixture-of-Experts architecture, with only a portion of the model activated per token, rather than the whole system firing at once. That selective activation reduces compute cost while preserving strong performance.

• Memory efficiency is part of the breakthrough
  DeepSeek’s use of Multi-head Latent Attention points to a wider trend: compressing attention and memory requirements so models can handle context more efficiently without the same computational burden.

• Open-source distribution changes the power dynamic
  Whereas some Western labs are building increasingly closed ecosystems, Chinese open-weight releases can spread faster, attract external developer support, and turn a model into a broader platform for iteration. That does not automatically guarantee leadership, but it does widen the field of experimentation.

China’s Data Advantage May Be Even More Important

• The next AI leap is not just textual, but multimodal
  Video, audio, image understanding, and physically coherent generation are becoming central to the next wave of models. That means the most valuable training fuel is no longer only scraped text, but richly structured real-world media.

• China’s platform ecosystem offers scale and continuity
  Companies like ByteDance sit on enormous flows of native video and engagement data through products such as Douyin and related services. This creates a high-volume stream of behaviourally rich multimodal material that is difficult to replicate from fragmented public web data alone.

• This is beginning to show up in product performance
  ByteDance’s Seedance 2.0 is positioned as a unified multimodal audio-video system, and recent reporting suggests Chinese video-generation models are now highly competitive at the frontier. In other words, the data advantage is already becoming visible in the output layer.

Why It Matters

What makes this shift so important is that it challenges a lazy assumption that AI leadership belongs automatically to whoever spends the most on hardware. The United States still holds immense structural advantages in chips, capital, and frontier infrastructure, but China has shown that pressure can accelerate a different path to competitiveness: leaner models, aggressive optimisation, open releases, and vast domestic data reservoirs. If this continues, the next phase of the AI race may be decided less by raw silicon supremacy and more by who can combine efficiency, scale, and multimodal data most intelligently.

Closing Thoughts

The deeper story here is not simply that China is catching up. It is that the terrain itself is changing. AlphaGo may have been the wake-up call, but the real lesson came after: technological leadership is rarely linear, and constraint often breeds a more adaptive form of innovation. That sets up the next question perfectly: if software efficiency and data abundance are now redefining the contest, what happens when AI begins moving from screens into the physical world?

Companion Video:

Section 2 | The Race for AI Vision: Apple’s N50 vs. The Industry

If Section 1 showed how the AI race is being reshaped at the level of models, chips, and data, Section 2 brings that contest directly onto the human body. The next platform battle is increasingly moving from the smartphone in your hand to the interface on your face. Apple, Meta, Google, and Snap are all circling the same prize: the device that can make AI feel ambient, useful, and socially acceptable enough to wear every day. Recent reporting suggests Apple is working on smart glasses designed to interpret the surrounding world and feed that context into Siri, while Meta has already turned smart glasses into a fast-growing commercial category.

From Vision Pro to a Lighter Bet

• Apple appears to be shifting from spectacle to subtlety
  Vision Pro proved that Apple could still build category-defining hardware, but it also revealed the limits of a bulky, premium headset as a mainstream product. The new direction, as reported by Bloomberg, points toward much lighter smart glasses rather than full mixed reality headsets.

• The reported N50 concept is about context, not immersion
  Rather than projecting a full digital layer into your field of view, Apple’s glasses are said to focus on cameras, environmental sensing, and contextual AI. That makes them less like miniature headsets and more like a wearable front-end for an intelligent assistant.

• This is a meaningful strategic pivot
  The ambition now seems to be to make AI disappear into everyday life instead of demanding that users adopt a visibly futuristic device. In other words, the winning hardware may not be the most cinematic, but the most frictionless.

Apple’s Real Advantage May Be the Stack Behind the Glasses

• Apple’s hardware strength is not just industrial design
  What gives Apple leverage is its long-standing ability to integrate chips, software, services, and supply chain control into a single experience. If smart glasses do become mainstream, that level of vertical integration could matter more than isolated product cleverness.

• The assistant layer is likely to matter as much as the frames
  A pair of AI glasses only becomes compelling if the assistant behind it is fast, aware, and actually useful. Apple’s reported strategy relies heavily on a more capable Siri that can interpret surroundings and coordinate actions across the wider Apple ecosystem.

• That could turn glasses into an extension of the iPhone, not a replacement
  At least initially, the most plausible route is a companion device that leans on the iPhone for compute, identity, and ecosystem continuity. This lowers the barrier to adoption while giving Apple millions of existing users a familiar on-ramp.

Meta Is Still the One to Beat

• Meta has the clearest consumer lead at the moment
  Ray-Ban Meta glasses have already sold in the millions, with Meta and EssilorLuxottica saying sales have more than tripled year over year. That gives Meta something most rivals still lack: real behavioural data from people actually wearing AI glasses in the wild.

• This is no longer a prototype market
  The category has moved beyond novelty. Meta’s smart glasses are now established enough for the company and its manufacturing partner to treat them as a scaling consumer line rather than a side experiment.

• Meta’s head start creates pressure on Apple and Google
  Apple may still be able to out-design the category, but Meta has already proven that people will buy AI eyewear when it is embedded in a recognisable brand and offered in a wearable form factor. That gives the entire market more legitimacy.

Google and Snap Are Building the Wider Field

• Google is rebuilding its glasses effort around Android XR and Gemini
  At Google I/O, the company said it is partnering with Samsung, Warby Parker, and Gentle Monster to create stylish glasses powered by Android XR. This suggests Google wants to provide the software and AI layer for a broader wearable ecosystem, rather than rely on a single in-house device.

• Snap remains the experimental wildcard
  Snap has continued developing its Spectacles platform and has publicly confirmed lightweight consumer Specs for 2026, supported by Qualcomm and on-device AI. It still feels more like a frontier player than a mainstream winner, but it remains influential in shaping what face-worn computing might become.

• The race is becoming an ecosystem contest, not just a product race
  What matters now is not only who makes the best-looking glasses, but who can support developers, assistants, services, and everyday use cases around them. That gives companies with mature AI ecosystems an increasingly important edge.

Why It Matters

This matters because the platform war after the smartphone is starting to come into focus. The real question is not simply whether people will wear AI glasses, but whose intelligence layer they will wear. If face-worn devices become the preferred interface for search, memory, guidance, communication, and contextual awareness, then the company controlling that layer may shape the next operating system for daily life. Meta currently leads in volume, Google is laying ecosystem foundations, and Apple appears to be preparing its usual play: arrive later, refine the category, and integrate it more elegantly than everyone else.

Closing Thoughts

The deeper shift here is philosophical as much as technical. AI is no longer being designed only as software we open on a screen. It is being woven into objects that watch, listen, interpret, and quietly mediate our relationship with the world around us. That makes the race for smart glasses about far more than fashionable hardware. It is about who builds the most trusted interface between human perception and machine intelligence. And that leads naturally into the next stage of the story: once AI can see the world through us, what happens when it begins to act within the physical world too?

Section 3 | Claude Opus 4.7 Arrives: Anthropic’s Public Frontier Takes a Step Forward

After chips, wearable interfaces, and the race to embed AI deeper into everyday life, the next frontier in this week’s story is the model layer itself. Anthropic has now officially released Claude Opus 4.7, describing it as its latest top publicly available model and positioning it as a notable improvement over Opus 4.6 in advanced software engineering, vision, and professional creative tasks such as interfaces, slides, and documents. Crucially, Anthropic also makes clear that Opus 4.7 still sits below its more powerful restricted model, Claude Mythos Preview, which remains gated behind the company’s controlled-access frontier track.

Anthropic’s Public Model Has Moved Up, But the True Frontier Remains Gated

• Claude Opus 4.7 is now Anthropic’s new flagship public model
  Anthropic says Opus 4.7 is generally available across Claude products, its API, Amazon Bedrock, Google Cloud Vertex AI, and Microsoft Foundry. The company presents it as a direct upgrade to Opus 4.6, especially for complex, long-running engineering tasks that require greater autonomy and consistency.

• Coding appears to be the headline improvement
  Anthropic says Opus 4.7 shows stronger performance on difficult software engineering work and notes that it outperforms Opus 4.6 across a range of benchmarks. The release also states that comparisons in its charts and tables use the best-reported API versions of GPT-5.4 and Gemini 3.1 Pro, reinforcing Anthropic’s claim that Opus 4.7 now competes at the very top of the public coding tier.

• But Anthropic is openly signalling that public access is no longer the true frontier
  The company explicitly says Opus 4.7 is less broadly capable than Claude Mythos Preview, and that Mythos remains restricted while Anthropic tests cyber safeguards on less capable public models first. That makes this one of the clearest examples yet of a split between what the public can use and what the lab itself considers closer to the frontier.

The Upgrade Is Not Just Smarter, But More Operational

• Anthropic kept pricing flat
  Opus 4.7 is priced the same as Opus 4.6 at $5 per million input tokens and $25 per million output tokens, which lowers friction for existing users considering the switch.

• The real trade-off is token behaviour
  Anthropic says Opus 4.7 uses an updated tokeniser, meaning the same input may map to roughly 1.0 to 1.35 times as many tokens, depending on the content. It also “thinks more” at higher effort levels, especially in agentic settings, which can lead to more output tokens even when reliability improves.

• That means the model may feel more capable, but not always cheaper in practice
  Anthropic argues that the net effect is favourable in its own testing, but it also recommends measuring the change on real traffic. In other words, Opus 4.7 may be a better worker, but teams will still need to understand the cost-performance curve in live use.

Claude Code Is Also Getting More Serious

• Anthropic is expanding effort control for harder work
  Opus 4.7 introduces a new xhigh effort level between high and max, giving developers finer control over the trade-off between reasoning depth and latency. Anthropic has also raised Claude Code’s default effort level to xhigh across all plans.

• A new /ultrareview mode pushes Claude further into engineering workflows
  Anthropic says the new slash command creates a dedicated review session that reads through changes and flags bugs and design issues a careful human reviewer might catch. That is a meaningful signal that Claude is being positioned not merely as a generator, but increasingly as a collaborator in code quality and review.

• The release also broadens the model’s visual and creative role
  Anthropic says Opus 4.7 has substantially better vision, can process higher-resolution images, and is more tasteful and creative when producing interfaces, slides, and docs. So while coding leads the headlines, the model is also being shaped for a broader class of professional output.

A Dual-Track Anthropic Is Now Fully Visible

• Anthropic is running a public track and a gated frontier track in parallel
  The public side is moving at speed through Opus releases and product rollouts. The frontier side, represented by Mythos Preview, remains restricted while Anthropic experiments with cyber safeguards, alignment, and broader deployment conditions.

• This is both a product strategy and a governance signal
  Anthropic is effectively saying that some capabilities are ready for mainstream use, while others still require restricted handling. That creates a new dynamic in which the publicly visible AI race may no longer reflect where the actual frontier sits inside the labs.

• That split may become more common across the industry
  As models grow more capable in coding, cyber, and autonomous workflows, companies may increasingly separate their commercial releases from their most advanced internal systems. Anthropic is now making that distinction unusually explicit.

Why It Matters

This matters because it changes how we should interpret progress in AI. Public launches used to feel like the frontier itself. Now they may be better understood as managed releases from a frontier that has already moved ahead behind closed doors. Claude Opus 4.7 is clearly a meaningful public upgrade, especially for agentic coding, code review, and professional output. But the deeper signal is Anthropic’s growing willingness to acknowledge that its most powerful model is not the one most people can use. That introduces a new kind of asymmetry into the AI race: the gap between public capability and internal frontier capability.

Closing Thoughts

Throughout this edition, we have watched intelligence spread across the stack: from model efficiency in China to wearable interfaces to creator tools, and now to the public-versus-private frontier within one of the leading AI labs. Claude Opus 4.7 is more than a routine model upgrade. It is a glimpse of a new phase in AI, where the most interesting question is no longer only which model performs best, but which layer of capability is being shown to the public, and which layer is being held back. That distinction may define the next chapter of the AI race just as much as the benchmarks themselves.

A Dose of Creator Tech for Your Weekend | Because the Future of Creativity Is Also Hardware

If this edition has explored the race for intelligence across chips, models, and next-generation interfaces, this final signal turns our attention to the tools creators will actually hold in their hands. GoPro’s newly announced Mission 1 Series suggests that the future of filmmaking may be smaller, tougher, and far more capable than many expected. It is a fascinating move from a company best known for action cameras, now stepping more confidently into the world of compact cinema.

• GoPro is aiming beyond the action camera category
  The new MISSION 1 Series is not being framed as a simple HERO update, but as a serious push into professional compact cinema. That gives the launch a different weight and positions GoPro in a more ambitious part of the imaging market.

• The flagship specs are surprisingly strong
  MISSION 1 PRO features a 50MP 1-inch sensor, up to 8K60 capture, 4K240, 1080p960 burst slow motion, and open gate recording, all powered by the new GP3 processor. For a camera this small and rugged, that is a notable statement of intent.

• The lineup covers different levels of creator ambition
  GoPro has introduced three models: MISSION 1 PRO, MISSION 1, and MISSION 1 PRO ILS. The ILS version adds Micro Four Thirds lens compatibility, giving more serious shooters added flexibility without abandoning the compact form factor.

• The wider ecosystem matters too
  Alongside the cameras, GoPro is launching a supporting accessory range including a wireless mic system, new grips, a Media Mod, and a higher-capacity battery. This reinforces the idea that MISSION 1 is being built as a creator platform, not just a one-off device.

• NAB will reveal the commercial picture
  GoPro has said pricing and more details will be unveiled at NAB in Las Vegas from 19 to 22 April 2026. That makes this week’s announcement feel like a strong opening move rather than the full story.

Why It Matters

This matters because creative hardware is entering a new phase. The line between consumer-friendly portability and professional-grade capability is shrinking fast, and that opens up more possibilities for independent filmmakers, hybrid creators, and agile production teams. If GoPro can deliver on image quality, thermals, battery life, and usability, MISSION 1 could become a meaningful tool in the wider democratisation of cinematic production.

Closing Thoughts

GoPro made its name by helping people capture movement, risk, and adventure in places traditional cameras could not easily go. With the MISSION 1 Series, it seems ready to take that same spirit into a more cinematic frontier. In a weekend digest shaped by AI, platforms, and new interfaces, this feels like a fitting reminder that the tools of expression are evolving just as quickly as the intelligence behind them.

One More Thing… | NASA’s Moon Base Blueprint: From Return to Settlement

NASA has now moved the Moon conversation beyond flags, footprints, and symbolic missions. With the release of its Moon Base User’s Guide: Architecture Resources, the agency has outlined a phased plan for establishing a continuous human presence in the lunar south pole region. This is not simply about going back. It is about learning how to stay. The document frames the Moon Base as a long-term effort tied to scientific, technological, and economic development, with the explicit end goal of continuous human presence on the lunar surface.

• NASA is thinking in phases, not stunts
  The guide sets out a three-phase implementation pathway: early technology demonstrations and experiments, then initial crewed Moon Base missions and infrastructure build-out, and finally continuous crew presence. NASA’s phasing also scales surface payload delivery from roughly 4,000 kg in Phase 1 to roughly 150,000 kg in Phase 3.

• The Moon Base is really an infrastructure stack
  The document makes clear that sustained lunar presence depends on far more than rockets. NASA highlights shared systems for power, logistics, communications, navigation, robotics, cargo handling, and habitat support as the foundations of a functioning lunar outpost.

• The lunar south pole is strategic, but difficult
  NASA chose the region for its scientific and economic promise, but the guide is clear about the challenge: low-angle sunlight, long shadows, extreme cold, deep craters, and difficult terrain. These conditions make the Moon Base as much an engineering testbed as an exploration destination.

• Robotics and autonomy are central from the beginning
  The guide places strong emphasis on autonomous systems that can unload cargo, reposition equipment, survey terrain, and maintain the base during uncrewed periods. That means the Moon Base is not just a human habitat plan, but a human-machine operations architecture.

• NASA is explicitly inviting others to co-build it
  The framework is designed to align industry, academia, and international partners around interoperable systems and shared standards. NASA repeatedly frames the Moon Base as a collaborative effort, not a solo national performance.

Why It Matters

This marks a genuine shift from exploration to infrastructure. The Moon is no longer being treated only as a destination, but as a proving ground for operating beyond Earth at scale. NASA says the programme will use phased technology maturation, shared systems, and commercial partnership models to build toward enduring presence while also enabling Mars-forward capabilities. The guide even states that lessons from CLPS and bulk procurement can help create sustained business cases, a lunar marketplace, and a wider cislunar economy.

Closing Thoughts

What we are seeing here is the early systems diagram of a multi-planetary civilisation. The Moon Base matters not only because it points to life on the Moon, but because it begins to answer a deeper question: how do humans build repeatable, resilient, shared infrastructure beyond Earth? If Artemis was the return, this document feels like the transition. From here, the trajectory is clear: orbit → Moon → Mars → beyond.

Source: NASA, Moon Base User’s Guide: Architecture Resources

🌀 Final Wrap-Up | Signals from the Expanding Frontier

In this edition of Geeky Weekend Digest, we explored a world in transition across multiple layers of the technological frontier. We began with China’s AI pivot, where the race is no longer defined only by hardware dominance, but by software efficiency, open models, and control of rich data ecosystems. We then moved into the next platform war, where Apple, Meta, and others are competing to make AI wearable, ambient, and woven into everyday perception through smart glasses and face-worn hardware. From there, we looked at Anthropic’s release of Claude Opus 4.7, a significant public upgrade that strengthens the company’s position in agentic coding and professional workflows, while also making visible a more important shift: the growing gap between the models the public can access and the more powerful frontier systems still held behind restricted access. We also took a brief detour into creator tech with GoPro’s new MISSION 1 cinema line, before ending on NASA’s Moon Base blueprint, a reminder that the infrastructure of tomorrow is now being designed both on Earth and beyond it.

Taken together, these are not isolated stories. They are connected signals from an expanding frontier where intelligence is becoming more efficient, more wearable, more creative, and more infrastructural, shaping not only the tools we use, but the environments, systems, and futures we are learning to inhabit.

💬 Your Thoughts | Join the Living Conversation

Which signal stayed with you most this week?

I’m always curious to hear how these developments are landing for you, because MyGeekSpace has never been about presenting final answers. It is an ongoing exploration shaped by dialogue, curiosity, and shared interpretation.

→ Share your thoughts in the comments on Substack, or continue the conversation on 𝕏 @frankdasilva.

Leave a comment

Living the Future works best as a conversation. Every perspective helps refine the signal, test assumptions, and deepen our understanding of the world now taking shape around us.

Where do you feel intelligence is moving next?

✨ Continue the Journey

If this edition resonated with you, you can subscribe on Substack, become a paid member, join me on Patreon, or explore the wider MyGeekSpace ecosystem at MyGeek.Space.

Subscribe now

Your readership does more than support independent work. It helps sustain a thoughtful, future-facing community committed to making sense of emerging change while the world is still being rewritten.

→ If these reflections resonated and you’d like to support the work without subscribing, you can also make a one-time contribution below.

Support MyGeekSpace

Futūrum Vīvere | To live the future is to help shape it.

At MyGeekSpace | Living the Future, grounded in the Terra 2.0 vision, I follow meaningful signals emerging beneath the daily noise of headlines and hype. This is not a space for fixed predictions, but a space for orientation: a way of noticing direction while the future is still taking shape.

In this week’s edition, we explore frontier AI as digital defence, Meta’s push towards personal superintelligence, Europe’s growing autonomous mobility threshold, and fresh lunar insights that may help shape humanity’s return to the Moon.

Power up your curiosity and settle into this week’s signals.

Support MyGeekSpace

Section 1 | Project Glasswing and the Rise of AI Cyber Defenders

Anthropic’s latest announcement feels less like a conventional model launch and more like the quiet unveiling of a new security doctrine. With Project Glasswing, the company has introduced Claude Mythos Preview, an unreleased frontier model that it says is already capable of finding and exploiting software vulnerabilities at a level beyond nearly all human experts. Rather than releasing it publicly, Anthropic has placed it inside a tightly controlled cybersecurity coalition with major technology, finance, and infrastructure partners. In other words, we may now be entering the age of the AI firewall: machine-speed defence for a machine-speed threat landscape.

• Anthropic has not released Mythos publicly because of its unusually strong cyber capabilities, choosing a restricted rollout instead.

• Project Glasswing brings together major players, including AWS, Apple, Google, Microsoft, NVIDIA, JPMorganChase, Cisco, CrowdStrike, Broadcom, the Linux Foundation, and Palo Alto Networks.

• Mythos has already uncovered thousands of serious flaws, including vulnerabilities across every major operating system and web browser.

• Anthropic is backing the effort with up to $100 million in usage credits and an additional $4 million in donations to open-source security organisations.

A frontier model too powerful for general release

Claude Mythos Preview is not being presented as a niche cyber model, but as a general-purpose frontier system whose gains in coding, reasoning, and autonomy have spilt over into offensive and defensive cybersecurity. Anthropic’s own release notes describe it as available only through an invitation-only gated research preview for defensive work under Project Glasswing. The decision itself is telling: this is not simply a stronger chatbot, but a capability threshold the company believes requires controlled deployment.

• Mythos is general-purpose, not cyber-specific, which makes its security performance even more significant.

• Access is invitation-only, limited to launch partners and additional critical infrastructure organisations.

• Anthropic frames this as an urgent defensive response, not a normal product rollout.

Anthropic’s official materials also point to substantial benchmark gains. Its system card summary reports that Mythos Preview reached 93.9% on SWE-bench Verified, averaged over five trials, a score that signals a major step up in real software engineering capability. That matters because the same capabilities that make models better at writing and fixing code can also make them far better at finding and exploiting weaknesses in it.

• Mythos reportedly hit 93.9% on SWE-bench Verified, according to Anthropic’s system card summary.

• Its cyber capabilities appear to be emergent, arising from broader improvements in code, reasoning, and autonomy.

• This makes the release strategically important because future general models may inherit similar capabilities by default.

From bug hunting to autonomous exploitation

What makes Mythos especially striking is not merely that it can identify bugs, but that Anthropic says it can turn vulnerabilities into working exploits with minimal or no human intervention. In its technical write-up, the company says Mythos identified and exploited zero-day vulnerabilities in every major operating system and every major web browser when directed to do so. It also describes a now-patched 27-year-old OpenBSD bug, a 16-year-old FFmpeg flaw, and a fully autonomous exploit of a 17-year-old FreeBSD remote code execution vulnerability.

• Mythos found a 27-year-old OpenBSD bug that had remained undiscovered in a security-focused operating system.

• It also uncovered a 16-year-old FFmpeg vulnerability, despite the software having been heavily tested for years.

• Anthropic says it autonomously exploited a 17-year-old FreeBSD flaw, achieving root access from an unauthenticated starting point.

Anthropic further reports that engineers without formal security training were able to prompt Mythos to find remote code execution vulnerabilities overnight and receive a complete working exploit by morning. In one comparison, Opus 4.6 produced working Firefox exploit conversions only twice in several hundred attempts, while Mythos achieved working exploits 181 times and gained register control 29 more times.

• Non-specialists could reportedly use Mythos effectively, lowering the expertise barrier for advanced cyber operations.

• Mythos dramatically outperformed Opus 4.6 in Anthropic’s own exploit-development comparison.

• Anthropic says some exploits were produced in hours, whereas expert penetration testers estimated the same work might normally take weeks.

Project Glasswing is a preview of the next security era

Project Glasswing is perhaps most interesting as a civilisational signal. Anthropic is effectively arguing that frontier AI has already crossed into a zone where defensive coordination between labs, hyperscalers, software vendors, security firms, and financial infrastructure becomes essential. The initiative currently includes 12 launch partners, plus access for more than 40 additional organisations that build or maintain critical software infrastructure. Reuters reports that Anthropic is also discussing the implications with the US government.

• This is not a solo-company move, but a coalition model for securing critical infrastructure.

• More than 40 additional organisations are involved, beyond the core launch group.

• Government conversations are already underway, which suggests this is being treated as a strategic capability, not just a product story.

Seen this way, Project Glasswing is less about one model and more about the architecture of the near future. If frontier systems can already uncover thousands of high and critical vulnerabilities, then cybersecurity may increasingly become a contest between competing machine intelligences: one side scanning, probing, and exploiting at scale, the other anticipating, patching, and hardening before the attack lands.

• Security is becoming machine-speed by necessity, not by preference.

• The old cadence of patching may no longer be enough, especially if exploit windows collapse further.

• Glasswing may become a template for how future frontier systems are deployed when open release is too risky. This is an inference from Anthropic’s rollout strategy.

Why It Matters

This may be one of the clearest signs yet that the frontier is no longer defined only by chat quality, creativity, or coding assistance. It is now also defined by strategic power over the digital substrate of modern civilisation. When a model is considered too capable to release broadly, yet powerful enough to help secure banks, browsers, kernels, clouds, and open-source infrastructure, we are looking at a profound shift in what AI actually is. It is becoming part of the immune system of the machine world.

Closing Thoughts

Project Glasswing gives us a rare glimpse behind the curtain. It suggests that the most advanced AI systems inside the top labs may already be operating a step beyond what the public sees, and that responsible deployment now means controlled alliances, not instant product launches. For this weekend’s Digest, that makes Mythos more than a cybersecurity story. It is an early signal that the intelligence age is beginning to harden itself.

Section 2 | Meta’s Superintelligence Play Moves Into Everyday Life

If Anthropic’s Project Glasswing showed us AI as a defensive shield for critical infrastructure, Meta’s latest move points in almost the opposite direction: AI as a deeply embedded personal layer across daily digital life. With the launch of Muse Spark, the first model from Meta Superintelligence Labs, Meta is not just releasing another assistant upgrade. It is laying the groundwork for what it calls “personal superintelligence”, an AI system designed to sit inside the products billions of people already use, from Meta AI and the web to WhatsApp, Instagram, Facebook, Messenger, and its AI glasses.

• Muse Spark is the first model in Meta’s new Muse series, built by Meta Superintelligence Labs as part of its push towards personal superintelligence.

• It already powers the upgraded Meta AI app and meta.ai, with broader rollout across Meta’s ecosystem planned in the coming weeks.

• Meta says the model was purpose-built for its own products, aiming to make the assistant faster, more context-aware, and more useful in everyday situations.

• The bigger strategic shift is clear: Meta is no longer talking only about models, but about an AI layer woven directly into the social and behavioural infrastructure of modern life. That final point is an inference from the scope of the rollout Meta describes.

A new model built for people, not just benchmarks

Meta frames Muse Spark as a deliberate break from the usual AI arms race narrative. Rather than leading with abstract scale, it is presenting the model as “purpose-built to prioritise people”, small and fast by design, but still capable of reasoning through more complex questions in science, maths, and health. The company says Meta Superintelligence Labs rebuilt its AI stack from the ground up over the past nine months, making Muse Spark the first real output of that accelerated effort.

• Meta says it rebuilt its AI stack from scratch in nine months, suggesting an unusually fast reset in strategy and execution.

• Muse Spark is intentionally small and fast, rather than simply being presented as the biggest possible model.

• Meta is positioning this as the foundation of a series, with larger Muse models already in development.

This is an important tonal distinction. Where some labs present frontier AI as an external tool or research milestone, Meta is leaning into intimacy, convenience, and integration. Muse Spark is meant to feel close at hand, less like a distant supercomputer and more like a responsive layer built into the interfaces people already inhabit. That makes the product story feel less spectacular on the surface, but potentially more consequential in practice. This framing is an inference from Meta’s product-led announcement and distribution strategy.

The assistant becomes multimodal, contextual, and ambient

What Meta is really launching here is not only a model, but an upgraded assistant experience. Meta says users can switch between Instant and Thinking modes depending on the task, while the assistant can also launch multiple subagents in parallel to handle more complicated requests. Its example is deliberately consumer-friendly: planning a family trip with one agent drafting the itinerary, another comparing destinations, and another surfacing relevant activities at the same time.

• Meta AI now includes Instant and Thinking modes, allowing a quicker or more deliberative style of response.

• The assistant can run multiple subagents in parallel, which Meta presents as a way to answer more complex questions faster.

• This signals a move from single-response assistants to orchestrated AI workflows, even for mainstream consumer use. That last point is an inference based on Meta’s described subagent architecture.

Muse Spark also pushes further into multimodal perception. Meta says the model can interpret photos and other visual inputs in real time, allowing users to point the assistant at the world instead of describing everything through text alone. That capability becomes especially significant when connected to AI glasses, where perception is no longer occasional but ambient. In that sense, Meta is not just making an assistant smarter. It is nudging it closer to a persistent perceptual companion.

• Muse Spark is built for voice, text, and image input, strengthening Meta’s multimodal push.

• Meta explicitly links these capabilities to its AI glasses, where visual understanding becomes even more useful.

• The broader implication is ambient AI, where assistance begins to blend into perception itself. This is an inference from Meta’s own glasses roadmap and product framing.

Health, context, and the battle for everyday relevance

One of the more notable details in Meta’s announcement is its emphasis on health. The company says health is one of the top reasons people turn to AI, and that it worked with physicians to improve the model’s ability to provide useful information on common health questions, including some involving images and charts. That is a revealing priority: not just because health is high-stakes, but because it shows where Meta believes consumer AI will become most personally sticky.

• Meta is treating health as a major use case, not a side feature.

• It says physicians helped shape this capability, especially for common health questions and image-based queries.

• This suggests the race is no longer only about general intelligence, but about trust, usefulness, and repeat engagement in the most personal categories. That final point is an inference from Meta’s product emphasis.

Meta also wants the assistant to be more socially and contextually grounded. Its announcement points to future answers enriched with recommendations, posts, Reels, photos, and creator content drawn from across its platforms, with attribution back to the original creators. This could become one of Meta’s biggest differentiators: not merely intelligence in the abstract, but intelligence fused with the social graph, the interest graph, and the visual culture already flowing through its apps.

• Meta wants AI answers to draw from the content people already share and follow, especially across Instagram, Facebook, and Threads.

• The company says richer visual answers are coming, including posts, photos, and Reels with creator credit.

• That could give Meta a distribution and context advantage over rivals that do not own comparable consumer platforms. This is an inference based on Meta’s existing product footprint.

From Llama openness to a more controlled AI stack

Another important shift is strategic. Meta says Muse Spark will be available in private preview via API to select partners, and that it hopes to open-source future versions of the model. That wording matters. Unlike Llama, which became central to Meta’s open model identity, Muse Spark is launching as a proprietary product tied tightly to Meta’s own assistant experience and platform ecosystem.

• Muse Spark is being offered in private preview via API, rather than being released broadly from day one.

• Meta says it hopes to open-source future versions, but has not committed to a timetable.

• This marks a subtle but meaningful evolution in Meta’s AI posture, from open model evangelism towards a more vertically integrated full-stack approach. This is an inference from the contrast between Muse Spark and Meta’s earlier Llama strategy.

Why It Matters

If Anthropic’s Mythos hinted at AI becoming part of civilisation’s digital immune system, Muse Spark points to the other half of the story: AI becoming an everyday personal layer wrapped around communication, search, shopping, health, and perception. Meta is uniquely positioned here because it does not need to persuade billions of people to adopt a new destination. It can simply weave its assistant into the platforms they already use. That makes this less a model release than a distribution event at a planetary scale.

Closing Thoughts

Muse Spark may not arrive with the same aura of danger or secrecy as Anthropic’s Mythos, but in some ways it may prove just as significant. One points to frontier AI as strategic defence. The other points to frontier AI as an intimate infrastructure. Together, they suggest the same deeper truth: the intelligence age is no longer forming only in labs. It is now being wired into the systems, tools, and interfaces through which modern life is increasingly lived.

Subscribe now

Section 3 | Europe’s Robotaxi Threshold: Tesla’s London and Berlin Test

If Anthropic showed us AI defending digital infrastructure and Meta showed us AI moving into everyday personal life, Tesla brings the same intelligence shift onto the physical streets of Europe. Reports this week suggest Tesla has received approval for unsupervised Full Self-Driving pilot operations in London and Berlin, potentially marking the company’s first true driverless urban deployments in Europe. That said, while the legal and regulatory groundwork is very real, I could not find an official Tesla, UK government, or Berlin authority announcement confirming that the pilots had already begun operating as of 9 April 2026. So this story is best understood as a threshold moment: the frameworks are opening, the testing has been extensive, and Europe’s self-driving era is moving from anticipation towards implementation.

• MarketMinute reported on 8 April 2026 that Tesla had launched unsupervised FSD pilots in London and Berlin, but I could not verify that rollout through primary official sources.

• The UK and Germany do now have legal frameworks that can support driverless operation in defined circumstances, which is what makes the claim plausible in the first place.

• Tesla’s European autonomy push is confirmed, with Full Self-Driving (Supervised) under review in the Netherlands and broader EU ambitions for 2026.

• The bigger signal is not just Tesla, but Europe beginning to move from regulatory theory to real-world autonomous deployment. This final point is an inference from the legal and commercial developments now converging.

The law is finally catching up with the machine

In the United Kingdom, the key enabler is the Automated Vehicles Act 2024, which received Royal Assent on 20 May 2024. The Act creates a formal authorisation pathway for self-driving vehicles and introduces the concept of the authorised self-driving entity, the company that bears responsibility when a vehicle is driving itself. Government guidance published in 2025 makes that shift explicit: once authorised, legal responsibility can move from the human to the company behind the autonomous system.

• The Automated Vehicles Act 2024 is real and already on the books, giving Britain a statutory basis for self-driving vehicle authorisation.

• The authorised self-driving entity model matters because it shifts responsibility from driver to operator when the vehicle is in self-driving mode.

• London has already been preparing for driverless trials, with public documents indicating that, subject to final confirmations, spring 2026 would see the first London trials without a safety driver.

Germany, meanwhile, has been one of the most ambitious jurisdictions anywhere in the world on autonomous driving. The Federal Ministry for Digital and Transport has explicitly said Germany created the framework for Level 4 autonomous vehicles to operate in regular public transport within defined operating areas. The KBA, Germany’s Federal Motor Transport Authority, has also been empowered to issue national type approvals for fully automated Level 4 vehicles under the AFGBV framework.

• Germany already has a national legal framework for Level 4 autonomy, not just experimental exemptions.

• Driverless operation is allowed only in pre-approved operating areas, which is crucial for understanding how any pilot would be structured.

• The KBA can issue national approvals for Level 4 vehicles, giving the country a practical route from law to deployment.

Tesla’s European autonomy push is real, even if this specific launch remains murky

Whatever the exact status of the London and Berlin pilot report, Tesla’s broader regulatory campaign in Europe is well documented. Reuters reported on 20 March 2026 that Tesla Europe expected a decision by 10 April 2026 from the Dutch vehicle authority, RDW, on whether to approve Full Self-Driving technology in the Netherlands. RDW responded three days later that Tesla had again publicly mentioned a date for approval, but stressed that the assessment was still ongoing and that approval would come only if safety had been convincingly demonstrated.

• Tesla has been pushing hard for a Dutch approval milestone, with 10 April 2026 cited by Tesla as the expected decision date.

• RDW has publicly pushed back on certainty, saying the process is still under evaluation and that no date can be guaranteed.

• That means Europe is close to a decision point, but not past it yet, at least as of 9 April 2026.

Tesla and Tesla-adjacent coverage have also highlighted the scale of the company’s European preparation. Electrek reported that Tesla had completed all vehicle testing for FSD (Supervised) in the Netherlands, while Tesla-linked coverage cited more than 1.6 million kilometres of European road testing, over 13,000 customer ride-alongs, more than 4,500 track scenarios, and extensive compliance documentation. These figures should be read as part of Tesla’s regulatory positioning rather than as independent proof of safety, but they do show how serious the company has become about unlocking the European market.

• Tesla’s European test effort has been extensive, at least by the company’s own account and aligned reporting.

• The immediate target is FSD (Supervised), not broad consumer driverless release, which remains a crucial distinction.

• Any move into unsupervised operation in Europe would therefore be a major step beyond the current confirmed public product status.

The real contest is no longer American, but planetary

Tesla is not entering an empty field. Reuters reported in January 2026 that Waymo aims to launch a fully driverless ride-hailing service in London by the fourth quarter of 2026. That means London in particular is emerging as one of the most strategically important early battlegrounds for autonomous mobility outside the United States and China. Britain is not just permitting experimentation; it is becoming a proving ground.

• Waymo is targeting London for a driverless service in late 2026, making the city a high-value global testbed.

• Tesla, Waymo, and likely others are converging on Europe, not just as a market, but as a regulatory legitimacy arena. This is an inference from the competitive timing.

• Winning Europe matters symbolically and commercially, because success here signals that autonomous systems can pass some of the world’s more demanding regulatory and urban-operational tests. This is an inference supported by the regulatory complexity described across these sources.

Seen in that light, the importance of the reported Tesla pilots is not merely whether a few Cybercabs or modified Model 3s and Model Ys are already moving through Westminster or Mitte. It is that the conversation has shifted. Europe is no longer asking whether autonomous driving belongs on public roads in principle. It is now working through who gets there first, under what conditions, and with which legal responsibilities.

• The centre of gravity is shifting from possibility to governance, which is often the clearest sign that a technology is maturing. This is an inference from the regulatory progress now underway.

• Autonomy in Europe will likely be geographically phased, starting in tightly defined areas before broader expansion.

• That makes these pilots, confirmed or reported, important as civilisational signals, not just mobility news. This is an inference from the larger pattern.

Why It Matters

Autonomous driving has spent years oscillating between hype and delay. What changes the picture now is the convergence of three forces: legal frameworks mature enough to assign responsibility, AI systems capable enough to handle messy real-world environments, and companies confident enough to push for deployment at city scale. Whether or not Tesla’s London and Berlin pilots are already live exactly as reported, the direction of travel is unmistakable. Europe is moving closer to a world in which intelligence does not only live in apps, models, or clouds, but in vehicles navigating the street in real time.

Closing Thoughts

This makes a fitting third movement for the Digest. First, AI hardens the digital world. Then it personalises the social interface. Now it begins to inhabit physical mobility itself. The road to full autonomy in Europe will still be slow, contested, and politically sensitive, especially around safety, liability, and transparency. But the threshold is clearly in sight. And once cities begin to treat driverless systems as an operational reality rather than speculative theatre, the entire rhythm of urban life starts to change.

A Dose of Lunar Foresight for Your Weekend | Because the next giant leap will depend on where we find water

Israeli Study Pinpoints Lunar Ice Sites to Guide Future Moon Landings.

As this week’s Digest moves from digital infrastructure to personal AI and autonomous mobility, it feels only right to end by lifting our gaze back towards the Moon. A newly published study in Nature Astronomy offers something quietly profound: a clearer map of where lunar ice may have been accumulating for billions of years. In practical terms, that means a better guide for where future astronauts should land, explore, and potentially build. And in symbolic terms, it reminds us that the next phase of space exploration will not be driven by spectacle alone, but by resource intelligence, patient science, and long-range planning.

• A new lunar ice study has identified ancient cold traps near the Moon’s poles that may hold some of the most promising water reserves for future exploration.

• The research suggests ice accumulated gradually over billions of years, rather than arriving in a single dramatic event.

• That matters enormously for Artemis-era planning, because water ice could support drinking water, oxygen production, and even rocket fuel on the lunar surface.

• The timing also feels especially resonant, arriving in the same week that Artemis II continued humanity’s return to deep space around the Moon.

The Moon’s oldest shadows may hold its most valuable reserves

The study, led by Oded Aharonson of the Weizmann Institute alongside Paul Hayne and Norbert Schörghofer, examined how the Moon’s subtle changes in tilt over immense spans of time altered which craters near the poles remained in permanent shadow. By combining temperature data from NASA’s Lunar Reconnaissance Orbiter with simulations of the Moon’s thermal history, the researchers concluded that some of the coldest, oldest polar craters have likely been trapping water for as long as 3 to 3.5 billion years. Haworth Crater near the South Pole emerged as one of the strongest candidates.

• The team used orbital temperature data and long-range simulations to reconstruct how lunar cold traps formed over time.

• Their results suggest the oldest permanently shadowed craters may contain the most ice, especially in polar regions.

• Haworth Crater stands out as a key target, having remained in shadow for more than 3 billion years according to the study’s modelling.

Rather than imagining lunar water as the result of one ancient comet strike or a singular cosmic accident, this research supports a more elegant picture: the Moon as a patient collector, gradually gathering and preserving molecules across deep time in the coldest places its geography can offer. That gives future missions something much more useful than a romantic theory. It gives them a shortlist.

Why It Matters

Water changes everything. On the Moon, it is not just a scientific curiosity but a strategic resource. If astronauts can access and process lunar ice, it could reduce the need to haul essential supplies from Earth and make a sustained human presence far more viable. This is why polar ice has become such a central focus of Artemis-era planning, and why instruments such as the Lunar Compact Infrared Imaging System, planned for deployment near the lunar South Pole in late 2027, matter so much. In effect, the future of lunar settlement may depend not only on rockets and landers, but on our ability to read the thermal memory of ancient craters.

Closing Thoughts

There is something wonderfully fitting about this discovery arriving just as Artemis II renews the human relationship with the Moon. One mission carries people around our nearest celestial companion. Another study helps reveal where its hidden reserves may lie. Together, they hint at the real shape of the next space age: not flags and footprints alone, but intelligence, infrastructure, and the slow conversion of knowledge into presence. The Moon is no longer just a destination again. It is becoming a landscape we are learning how to live with.

One More Thing… | Artemis II Returns Home and Reopens Deep Space

Artemis II crew splashes down safely off San Diego after historic Moon flyby

NASA’s Artemis II mission has concluded successfully, with the Orion capsule splashing down in the Pacific Ocean off the coast of San Diego at approximately 5:07 p.m. Pacific Time on Friday, 10 April 2026. During re-entry, the spacecraft tore through the upper atmosphere at roughly 35 times the speed of sound, producing the conditions for a sonic boom across parts of Southern California before descending by parachute into the ocean. The safe return of Reid Wiseman, Victor Glover, Christina Koch, and Jeremy Hansen brings to a close the first crewed journey around the Moon since Apollo, and a mission that reopened deep space to human flight.

Read more

At MyGeekSpace | Living the Future, grounded in the Terra 2.0 vision, I follow meaningful signals emerging beneath the daily noise of headlines and hype. This is not a space for fixed predictions. It is a space for orientation, a way of noticing direction while the future is still taking shape.

In this week’s edition, we move from humanity’s return to deep space to the growing possibility of ancient life on Mars, from the latest shifts in AI and platform power to the rise of volumetric media, and finally to the economic turbulence that may define the next phase of the intelligence age.

Power up your curiosity and settle into this week’s signals.

→ If this weekly exploration resonates, consider supporting MyGeekSpace on Substack with a free or paid subscription. Your support sustains independent, future-focused writing and helps keep this signal alive.

Section 1 | Moonbound: For All Humanity

After 54 years, humanity has finally pushed beyond low-Earth orbit again. Artemis II is more than a successful launch. It is a civilisational signal that the long pause after Apollo may be ending. On 1 April 2026 at 22:35 UTC, NASA’s Space Launch System lifted Orion and its four-person crew from Launch Complex 39B at Kennedy Space Center, beginning the first crewed lunar mission since Apollo 17 in 1972. NASA defines Artemis II as a 10-day crewed lunar flyby and the first mission to send astronauts aboard both SLS and Orion, validating the hardware and operational rhythm needed for sustained deep-space exploration.

This mission carries NASA astronauts Reid Wiseman, Victor Glover and Christina Koch, alongside Canadian Space Agency astronaut Jeremy Hansen. Together, they are flying a free-return trajectory around the Moon, testing Orion’s life-support, navigation and communications systems in deep space before later Artemis missions attempt lunar landings. The crew also marks an important cultural milestone: Glover is the first Black astronaut on a lunar mission, Koch the first woman to travel toward the Moon, and Hansen the first Canadian, and first non-American, assigned to a Moon mission.

A New Chapter Beyond Low-Earth Orbit

• Humanity has broken the tether again: Artemis II is the first human mission beyond low-Earth orbit in more than half a century, reopening a frontier that had effectively remained dormant since 1972.

• This is a test flight with historic weight: Even without a landing, the mission is validating the systems, procedures and confidence needed for a permanent return to cislunar space.

• The mood has shifted from orbit to horizon: For decades, human spaceflight largely meant operating close to Earth. Artemis II signals a transition back to deep space as a destination in its own right.

There is something psychologically important about this mission. The Apollo era proved that humans could reach the Moon. Artemis II asks a different question: can we build a lasting pathway outward, one that turns rare heroic missions into repeatable infrastructure? That is why this flight matters even without boots on the lunar surface. It is helping convert spectacle into system. NASA is using the mission to prove that deep-space transport, crew operations and long-range communications can work together under real conditions, not just in simulation.

Orion as the Bridge to the Next Frontier

• Orion is being tested where it actually matters: Artemis II is taking its crew into deep space so NASA can evaluate the spacecraft’s performance far beyond the protective comfort of low-Earth orbit.

• Every subsystem is part of the future architecture: Life-support, navigation, communications and recovery are not side details. They are the foundations of a long-term human presence beyond Earth.

• This mission is about trust: Before humanity expands outward, astronauts must trust the vehicle, engineers must trust the data, and institutions must trust the mission model.

In practical terms, Artemis II is a systems rehearsal for the next era. Orion’s solar arrays were successfully deployed after launch, and NASA has already reported early orbital manoeuvres such as the perigee raise burn completing as planned. That may sound procedural, but it is precisely how an interplanetary future is built: not through rhetoric alone, but through a chain of successful steps, each one proving that the next is possible. The Up-Wing imagination needs engineering discipline, and Artemis II is where those two forces meet.

The Moon as a Training Ground for Mars

• Artemis is not only about returning to the Moon: NASA presents the programme as part of a broader Moon-to-Mars strategy.

• The lunar pathway is becoming strategic again: The Moon is re-emerging as a proving ground for habitat design, logistics, energy systems and human endurance in deep space.

• This is the beginning of a wider civilisational arc: Artemis II suggests that humanity may be moving from a space age of visits to a space age of presence.

Within the broader Artemis programme, this mission is a hinge point. Artemis I showed that the architecture could fly uncrewed. Artemis II now asks whether humans can live and operate inside that architecture on a real deep-space journey. NASA frames Artemis as part of a longer path to the Moon and on to Mars, which gives this mission significance far beyond a single 10-day loop around the lunar far side. If Apollo was about arrival, Artemis is increasingly about continuity.

Why It Matters

Artemis II matters because it reactivates a dormant layer of human ambition. It tells us that the future of civilisation is not confined to apps, screens and digital abstraction. It can also be physical, planetary and outward-facing. In Living the Future terms, this is an Up-Wing moment: a move away from managed decline and toward expansion, capability and horizon-thinking. The mission does not solve everything, but it reintroduces a vital idea into public consciousness, namely that humanity still has frontiers worthy of its intelligence, courage and imagination.

Closing Thoughts

Artemis II is not merely a nostalgic return to the Moon. It is a declaration that the long inward bend of the post-Apollo era may be giving way to something more ambitious. Four astronauts are now carrying more than instruments and mission checklists. They are carrying proof that the human story is still capable of lifting its gaze. After decades spent circling home, we are beginning, once again, to lean toward the wider cosmos. That is not just a space milestone. It is a cultural one.

Section 2 | Mars in Sight: Why NASA Thinks Ancient Life May Be Within Reach

If Section 1 was about physically breaking the tether to Earth, Section 2 pushes the horizon further still. The Moon is becoming our proving ground, but Mars remains the deeper imaginative prize: the place where exploration, astrobiology and civilisational ambition converge. Under NASA Administrator Jared Isaacman, the agency is now speaking in far more expansive terms, pairing a permanent lunar presence with a more aggressive long-range push toward Mars. Last week, NASA formally announced a major shift in its lunar strategy, redirecting effort away from the planned Gateway station and toward a permanent Moon base, while also unveiling a nuclear-powered Mars spacecraft concept called Space Reactor 1 Freedom.

From Lunar Return to Interplanetary Ambition

• NASA is thinking beyond flags-and-footprints missions: The new roadmap suggests a move from symbolic return missions toward permanent infrastructure on and beyond the Moon.

• The Moon is being framed as a stepping stone, not the end goal: A surface base is meant to support a larger human expansion into deep space.

• Mars is no longer a distant abstraction: It is increasingly being discussed as a practical next phase in a broader exploration architecture.

At NASA’s “Ignition” event on 24 March, Isaacman and agency officials outlined a plan to invest roughly $20 billion over seven years to fast-track a lunar base, with the wider programme expected to total around $30 billion over the next decade. NASA said the effort would rely on commercial and international partners and is intended to accelerate preparations for a crewed return to the lunar surface by 2028. In parallel, Reuters reported that NASA’s revised Artemis strategy includes more robotic landers, lunar drones and nuclear surface-power infrastructure, all of which strengthen the Moon’s role as a real operational frontier rather than a one-off destination.

A Moon Base, Then a Nuclear Ship to Mars

• NASA has sharpened its roadmap: The agency is linking lunar permanence with the technologies needed for deeper missions.

• Nuclear propulsion is back in the conversation: Space Reactor 1 Freedom points to a future where Mars missions are built around stronger, more capable interplanetary systems.

• Commercial speed is becoming part of the strategy: Isaacman’s approach appears more willing to streamline legacy plans and move resources toward what can shift the frontier fastest.

The most striking addition to the roadmap is NASA’s plan to send Space Reactor 1 Freedom to Mars by late 2028, using nuclear-electric propulsion and carrying Ingenuity-class helicopters for aerial exploration on the Martian surface. Reuters described this as part of Isaacman’s wider effort to push NASA towards a more urgent deep-space posture, while Isaacman’s earlier confirmation hearings also highlighted his support for nuclear propulsion and surface power as essential technologies for future Mars missions. This does not mean a crewed Mars landing is imminent, but it does mean the architecture of serious Mars preparation is beginning to move from concept language into programme language.

Why Mars Life No Longer Feels Like Science Fiction

• Isaacman’s remarks raised the stakes: He said that if NASA can get to Mars and return samples, he believes there is better than a 90 per cent chance of proving past microbial life.

• Mars is scientifically compelling for good reason: Jezero Crater, where Perseverance is working, was once shaped by flowing water and delta deposits.

• The evidence is not proof, but it is getting harder to ignore: Mars continues to present environments and materials that look increasingly relevant to the question of past habitability.

Isaacman’s most provocative remark came in an interview published on 1 April, where he said that if humans can reach Mars and bring samples back, he would place the odds at better than 90 per cent that we could prove there was once microbial life on the planet. That is a personal judgement rather than an official NASA probability, but it lands in a scientific context that is already rich with intrigue. NASA’s Perseverance mission is exploring Jezero Crater, a site chosen precisely because it preserves signs of an ancient river delta and a once water-rich environment that may have been favourable to life in the distant past. NASA has also highlighted the rover’s discovery of an intriguing rock nicknamed Cheyava Falls, whose chemistry and spotted patterns may reflect processes that could once have supported microbial life, although non-biological explanations remain possible.

There is a second reason the optimism is growing. Research highlighted by Astrobiology.com reported that amino-acid building blocks could persist for extremely long periods in Martian ice under simulated conditions, suggesting that buried ice deposits may be among the most promising places to search for preserved biosignatures. This strengthens the case for future missions that do more drilling, more subsurface access and, crucially, more sample return. In other words, the search for ancient Martian life is becoming less about vague hope and more about improving where, and how, we look.

Why It Matters

This matters because the Mars question sits at the intersection of science, meaning and future identity. Discovering ancient microbial life on Mars would not simply be another headline. It would alter humanity’s understanding of life as a cosmic phenomenon. At the same time, NASA’s current direction suggests something equally important: that exploration is becoming infrastructural again. The agency is no longer speaking only in terms of isolated missions, but in terms of systems, bases, power, transport and continuity. From a Living the Future perspective, this is deeply Up-Wing. It frames humanity not as a civilisation trapped in maintenance mode, but as one still capable of building outward with intelligence, confidence and purpose.

Closing Thoughts

Artemis II reminds us that we can leave orbit again. NASA’s new Mars posture suggests we may also be entering a phase where the biggest questions are no longer rhetorical. Can we build beyond Earth? Can we sustain life across worlds? And can we discover whether life began more than once in our cosmic neighbourhood? The Moon is the threshold. Mars is the deeper test. Together, they signal a future in which exploration is not just a branch of science, but a central expression of what kind of civilisation we intend to become.

Section 3 | AI Tech News: From Digital Brains to Platform Wars

If the first two sections were about expanding humanity’s physical frontier, this final one turns back toward the technological frontier here on Earth. This week’s AI story is not one story, but three. Meta is trying to compress neuroscience into software. Anthropic has been forced into an uncomfortable moment of transparency after a major Claude Code leak. And Apple, on its 50th anniversary, appears to be accepting that its future role in AI may be less about building the smartest model and more about becoming the place where everyone else’s AI runs. Together, these developments point to a bigger shift: AI is moving beyond chatbots and into infrastructure, cognition and operating-system power.

Meta’s Brain Model Pushes AI into Neuroscience

• Meta has open-sourced TRIBE v2: The company says the model predicts brain activity across vision, audition and language, moving neuroscience closer to large-scale in-silico experimentation.

• The scale-up is substantial: Meta describes TRIBE v2 as being trained on more than 1,000 hours of fMRI data from 720 people, expanding from roughly 1,000 to 70,000 brain regions.

• The bigger idea is acceleration: Instead of requiring a fresh scan for every new hypothesis, researchers may increasingly be able to simulate brain responses computationally first.

Meta’s framing is ambitious, and not without reason. In its publication and launch materials, the company presents TRIBE v2 as a tri-modal foundation model capable of predicting human brain responses to complex video, audio and language inputs, including zero-shot generalisation to new tasks and subjects. Meta says the model can recover known brain specialisations, including regions associated with faces, speech and text, while enabling “in-silico neuroscience” through open code, weights and an online demo. That does not mean virtual brains have replaced real humans in scanners, but it does suggest a future in which expensive, slow, highly constrained neuroscience workflows begin to merge with the speed and scale of modern machine learning.

Claude Code’s Leak Reveals the Cost of AI Convenience

• Anthropic accidentally exposed a large part of Claude Code’s internals: The company said a release packaging issue caused internal source code to be bundled into a public update.

• The leak offered more than embarrassment: It gave outsiders a detailed look at unreleased features, memory systems and roadmap-level capabilities.

• The real issue is trust: When AI coding tools gain filesystem access and terminal privileges, developers need to understand how much control and persistence those systems actually have.

According to Anthropic’s public statements and subsequent reporting, the leak did not expose customer data or credentials, and the company said it was caused by human error rather than a breach. But the incident still matters because the leaked code exposed how fast these tools are evolving beyond simple assistants. Reporting from Axios and The Verge indicates that the code revealed a persistent background assistant mode, feature flags for unreleased capabilities, deeper memory architecture, and remote control features that point toward longer-running, more autonomous workflows. That does not automatically validate every viral claim made online about telemetry or hidden controls, but it clearly confirms that frontier coding tools are becoming more agentic, more persistent and more operationally complex than many users likely assumed.

Apple at 50: From AI Builder to AI Platform

• Apple turned 50 on 1 April 2026: The company founded on 1 April 1976 is now repositioning itself for the AI era.

• Its AI strategy appears to be shifting: Reuters, citing Bloomberg, reports that Apple plans to let third-party AI services plug directly into Siri as part of iOS 27.

• This could be a major pivot: Rather than trying to dominate model development, Apple may be betting that its enduring strength lies in devices, distribution and trust at the platform layer.

There is something fitting about this shift arriving on Apple’s 50th anniversary. Apple was founded on 1 April 1976 by Steve Jobs, Steve Wozniak and Ron Wayne, and over five decades it turned itself from a garage startup into one of the most powerful companies in modern technological history. Yet this anniversary arrives at a moment when Apple is widely seen as trailing rivals in generative AI. Reuters reports that Apple is preparing to open Siri to third-party AI apps beyond ChatGPT, potentially including services such as Google’s Gemini and Anthropic’s Claude, with the changes expected to be previewed at WWDC in June. If that happens, it will signal an important repositioning: Apple may no longer be trying to win the model race directly, but instead to own the operating environment in which AI becomes mainstream. That is less glamorous than inventing the best frontier model, but it could still be extremely powerful.

Why It Matters

This cluster of stories reveals where AI is actually heading. Meta is pushing AI into the domain of human cognition itself. Anthropic’s leak shows that coding assistants are rapidly becoming semi-autonomous operating layers rather than simple tools. Apple’s shift suggests that the next major battleground may not just be model quality, but platform control, integration and user routing. In Living the Future terms, this is the transition from AI as novelty to AI as civilisational substrate. The real question is no longer whether AI will matter. It is where intelligence will live, who will mediate it, and how much sovereignty users will retain inside these new systems.

Closing Thoughts

This week’s AI news carries a quiet but profound message. Intelligence is becoming more distributed, more embedded and more infrastructural. It is moving into brain science, software workflows and the operating systems that shape everyday life. That makes this moment exciting, but also clarifying. The future of AI will not be determined by raw capability alone. It will be shaped by openness, trust, interface power and the architectures through which human beings choose to collaborate with increasingly capable systems. The model race matters, certainly. But the deeper story may be that we are now building the environments in which intelligence itself will be staged.

A Dose of Volumetric Reality for Your Weekend | Because the next great medium may be reality itself

If this edition has traced humanity’s outward journey from the Moon to Mars, it feels fitting to end with a frontier much closer to home, yet no less radical. 4D Gaussian Splatting points to a future in which reality itself becomes navigable, editable and archivable in ways that traditional media never allowed. We are no longer talking about better photos, sharper video, or even more convincing CGI. We are talking about a new visual grammar, one that captures the behaviour of light, space and time together.

I explored this in more depth in my MyGeekSpace piece, “Gaussian Splatting and the Birth of Real-Time Holograms”:

Beyond the Mesh: Capturing Light, Not Just Shape

• Traditional 3D capture has always been limited: Photogrammetry can reconstruct geometry impressively well, but it often struggles to reproduce the living, shifting qualities of real light.

• Radiance fields changed the game: Instead of treating an object as a fixed surface with static colour, they model how light moves from every point in space.

• That is what makes scenes feel real: We do not simply see objects, we see light bouncing from them into our eyes from changing angles.

For years, much of digital capture has been built around surfaces. Meshes, textures and maps can produce beautiful results, but they often fail the subtle perceptual test that tells the eye whether something feels truly present. A real object does not hold a single appearance. It changes with angle, illumination and perspective. Radiance-based approaches begin to address that. They shift us away from static models and towards a richer representation of the world as an active field of light.

Mathematical Minimalism and the Power of the Splat

• Gaussian Splats work differently from NeRFs: Rather than calculating every view through a dense neural function, they represent scenes as millions of soft, overlapping points in 3D space.

• Each point carries meaningful visual information: Position, shape, opacity, orientation and colour all combine to produce a coherent scene.

• This makes real-time rendering far more practical: The result is photoreal, navigable content that can run efficiently even in a browser.

What makes Gaussian Splatting so compelling is its elegance. Instead of building the world from rigid polygons or asking a neural network to infer every view on demand, it uses what look like fuzzy blobs, each one storing just enough information to contribute to a living image. These splats overlap like layered paint, but with mathematical precision. The result is surprisingly lightweight and surprisingly convincing. It feels like a breakthrough not because it is more complex than previous approaches, but because it is more efficient. In that sense, it embodies a very contemporary kind of innovation: not brute force, but elegant compression.

The Fourth Dimension Changes Everything

• 4D Gaussian Splatting adds time to the equation: The scene is no longer just spatially real, but temporally fluid.

• This creates continuous motion rather than frame-by-frame illusion: Instead of discrete captured moments, points can carry velocity and duration.

• The effect is closer to holography than video: What emerges is a volumetric memory that can be slowed, explored and re-entered.

This is where things become especially exciting. Traditional volumetric video has often felt cumbersome, heavy and brittle. It captures movement, but usually in a way that remains locked to sequences of frames. 4D splatting changes that by creating continuity. Motion becomes part of the scene’s structure rather than something pasted onto it afterwards. That means smoother interpolation, richer temporal depth and the possibility of truly inhabiting captured moments. This is no longer just media playback. It is something closer to a navigable event.

Tiny Detail, Reflection and the Mirror Dimension

• Splats excel at high-frequency detail: Fine textures like hair, foil, wings and subtle material variation can appear startlingly alive.

• They also handle reflections in ingenious ways: Instead of relying on heavy real-time ray tracing, some systems approximate mirrored worlds through layered scene structures.

• This helps bridge realism and efficiency: The image feels richer without carrying the full burden of computationally expensive simulation.

One of the reasons splats feel so fresh is that they handle the messy richness of reality unusually well. Tiny details that often become noisy, flattened or prohibitively expensive in other pipelines can remain vivid and tactile here. Reflections, too, can be approached with remarkable cleverness. The so-called mirror-dimension effect hints at how this medium can bend our assumptions about what a scene is. The world is no longer a sealed surface. It becomes a layered volume of visual possibility.

Democratising Capture and Preserving the Real

• This is not only for elite studios: Splats can increasingly be captured with smartphones, drones and more accessible workflows.

• That opens the door to cultural and historical preservation: Lost places, damaged environments and fleeting moments can be archived as navigable digital spaces.

• The implications extend far beyond entertainment: Education, memory, heritage, mapping and personal storytelling could all be transformed.

Perhaps the most powerful aspect of this medium is not technical, but civilisational. Once reality can be captured volumetrically with increasing ease, preservation changes. Places destroyed by wildfire, erosion, conflict or urban redevelopment do not simply survive as flat records. They can persist as immersive, explorable spaces. That alters our relationship to memory and cultural continuity. We begin to move from documentation to re-presence.

From Capture to Creative Manipulation

• The workflow is becoming increasingly practical: Capture, tracking and training form a pipeline that is already accessible to advanced creators today.

• Post-production possibilities are immense: Splats can be relit, integrated into real-time engines, and combined with simulated effects and digital objects.

• This turns captured reality into a creative substrate: Artists are no longer choosing between the real and the synthetic, but blending the two.

For visual effects, this is potentially enormous. Once a captured scene becomes a splat asset, it can be treated not just as a record, but as a manipulable environment. Light can be changed. Atmosphere can be added. Physics can be bent. The world can be folded, refracted or reimagined without losing its grounded sense of presence. That makes Gaussian Splatting feel less like a niche technical development and more like the opening of a new medium, sitting somewhere between cinema, simulation and memory architecture.

Why It Matters

4D Gaussian Splatting matters because it points toward a post-flat future. We are moving beyond screens as simple windows and toward media that can be entered, inhabited and explored. In Living the Future terms, this is one of the clearest signals yet that reality capture is becoming part of the next human interface layer. Once the world can be preserved and replayed volumetrically, our relationship to memory, space, storytelling and presence begins to shift. This is not just a better graphics pipeline. It is the beginning of a world in which reality itself becomes computable, portable and creatively alive.

Closing Thoughts

What makes this field so compelling is that it feels both technically advanced and strangely intuitive. Of course we would want more than photos. Of course we would want to step inside memories, moments and environments rather than merely glance at them from the outside. Gaussian Splatting, especially in its 4D form, feels like the first serious answer to that desire. It turns reality into something that can be archived, revisited and transformed without losing its texture. The holographic age has not fully arrived yet, but it is no longer science fiction either. It is beginning to look like the next natural interface for a civilisation learning how to live inside its own captured worlds.

One More Thing… | Navigating the AI Arms Race and Economic Shift

If this edition began with missions beyond Earth and moved through the expanding terrain of AI, it feels right to close with a question much closer to home: what happens when the intelligence revolution stops being abstract and starts rearranging everyday life, work and economic stability? Former Google X executive Mo Gawdat has become one of the more striking voices in this conversation, not because he offers easy reassurance, but because he frames the AI transition as a structural rupture. In his view, we are entering a decade of painful instability before any more abundant or humane order has a chance to emerge. Whether or not one agrees with his most dramatic timelines, the force of his warning is difficult to ignore.

Read more

MyGeekSpace | Living the Future, grounded in the Terra 2.0 vision, follows meaningful signals emerging beneath the daily noise of headlines and hype. This is not a space for prediction. It is a space for orientation, recognising direction while the future is still taking shape.

This week’s edition explores a powerful convergence now coming into focus. We begin with the vast infrastructure ambitions behind Elon Musk’s Terafab vision, where compute, energy, robotics, and space are being assembled into the early architecture of a future galactic civilisation. From there, we move into the more immediate world of AI agents, as Anthropic gives Claude the ability to operate a computer remotely and Meta pushes internal agent use deeper into the executive layer of corporate life. Along the way, we pause for a striking scientific signal from Japan, where researchers have surpassed a long-standing solar efficiency ceiling, hinting at new possibilities for an abundant energy future. We close with a broader industry shift, as leaders from Microsoft, OpenAI, and AWS openly declare that the chatbot era is giving way to something more active, autonomous, and operational.

Power up your curiosity and settle into this week’s signals.

→ If this weekly exploration resonates, consider supporting MyGeekSpace on Substack with a free or paid subscription. Your support sustains independent, future-focused writing and helps keep this signal alive.

Subscribe now

Section 1 | The Architecture of a Galactic Civilisation

This week’s first signal is not just about a factory in Austin. It is about the emergence of a new civilisational blueprint, one that links chips, energy, robotics, orbit, and long-term species expansion into a single strategic stack. Elon Musk’s newly announced Terafab initiative, bringing together Tesla, SpaceX, and xAI, is being framed as an effort to build enough compute to move beyond today’s bottlenecks and begin laying the foundations for an economy measured not merely in products, but in planetary and eventually interplanetary capability.

Strategic Briefing: The Terawatt Project and the Path to a Galactic Civilisation

At the centre of this story is a simple but profound idea: advanced civilisation runs on compute, and compute ultimately runs on power. Musk’s announcement of a new Austin-based chip manufacturing push is therefore best understood not as a standard industrial expansion, but as an attempt to rewire the infrastructure of the future itself. Reuters reports that the project involves advanced chip factories in Austin, jointly developed by Tesla and SpaceX, with output designed for Tesla vehicles, Optimus robots, and AI data centres intended for space deployment. Axios adds that the initiative is explicitly tied to the ambition of reaching one terawatt of annual compute capacity.

• A new industrial stack is forming
  Tesla, SpaceX, and xAI are no longer acting like separate companies with adjacent interests. They are beginning to look like components of a single vertically integrated civilisation machine.

• The objective is scale, not optimisation alone
  The stated ambition is to move from today’s gigawatt-scale compute world towards terawatt-scale capacity, which implies a leap of civilisational magnitude rather than a routine improvement in semiconductor throughput.

• Austin is the terrestrial launchpad
  The advanced fab in Texas is the first step, but the wider vision points beyond Earth, toward compute infrastructure deployed in orbit where energy constraints are far less severe.

The Kardashev Frame: Why Power Is the Real Story

To understand the scale of this ambition, it helps to zoom out. The Terafab idea is being framed in terms of the Kardashev scale, the classic measure of civilisational advancement based on energy use. A Type 1 civilisation meaningfully harnesses the energy available on its home planet. Type 2 taps the power of its home star. Type 3 reaches the scale of a galaxy. In this framing, humanity is still embryonic, technically brilliant in places, but energy-poor relative to the abundance available in our own solar system.

• Humanity is still operating far below its energetic potential
  The project’s logic begins with the claim that our current level of industrial civilisation uses only a vanishing fraction of the sun’s available energy. The implication is clear: true scale requires going off-world.

• Energy capture defines civilisation
  If civilisation is constrained by energy, then the path to abundance is not simply better software or smarter models. It is access to vastly greater and more stable power flows.

• Space is not a side quest
  In this vision, orbit and eventually the Moon are not symbolic destinations. They are the logical sites for the next era of industrial expansion.

Austin’s Terafab: Building the Recursive Engine

The immediate terrestrial component of the project is the planned advanced technology fab in Austin. What makes it notable is not just that it manufactures chips, but that it is intended to compress the entire chip-making lifecycle into one tightly integrated loop. Reports describe a facility that would bring design, fabrication, packaging, and testing into close proximity, dramatically speeding up iteration. That matters because in AI and robotics, development speed is often more decisive than perfection on the first attempt.

• Vertical integration accelerates learning
  If chip design, testing, packaging, and revision can happen in one recursive loop, the rate of innovation rises sharply. This is as much about tempo as it is about output.

• Existing suppliers remain important, but insufficient
  Musk has indicated that firms such as Samsung, TSMC, and Micron will still play a role, yet current global capacity is not expanding quickly enough for the scale of compute his companies believe they will need.

• The fab is positioned as a necessity
  The rationale is not simply independence for its own sake. It is that without internal manufacturing capacity, the broader robotics and AI roadmap may stall.

From Earth to Orbit: Why the Compute Future May Be in Space

Perhaps the boldest part of the Terafab narrative is the claim that most future AI compute should not remain on Earth at all. Multiple reports on the announcement describe a model in which only a minority of compute stays ground-based, while the largest share moves into space. The reason is power. Terrestrial expansion faces land use battles, infrastructure delays, and local political resistance. Orbital systems, by contrast, could draw on uninterrupted solar energy at a far greater intensity.

• Space solar has structural advantages
  In orbit, solar collection avoids weather, seasonal loss, and the day-night cycle, making power generation more consistent and potentially more efficient than on Earth.

• Starship is the enabling infrastructure
  None of this works without heavy-lift launch capacity. The Terafab vision is therefore inseparable from SpaceX’s effort to reduce the cost of moving massive payloads into orbit.

• This reframes data centres as space infrastructure
  The future AI stack, in this telling, starts to look less like rows of warehouses on land and more like orbital industrial architecture.

Specialised Hardware for a Machine Civilisation

Another important dimension here is hardware specialisation. The project appears to anticipate at least two broad chip categories: one for edge inference in Tesla vehicles and Optimus robots, and another designed for high-energy, space-based environments. Reuters and other coverage note that these chips would serve very different functions, reflecting a future in which AI hardware is no longer generic, but deeply optimised for particular civilisational roles.

• Robotics becomes a primary demand driver
  The logic is that if humanoid robots scale into the billions, chip demand will explode far beyond the automotive market.

• Space chips require new engineering assumptions
  Chips intended for orbital deployment must withstand radiation, heat, and environmental conditions unlike those faced by conventional terrestrial data centre hardware.

• Compute is becoming task-specific infrastructure
  This is not just more silicon. It is the emergence of differentiated hardware layers for mobility, labour, intelligence, and off-world industry.

The Long Arc: Petawatt Ambitions and the Abundance Economy

Beyond the immediate fab plan lies a much larger philosophical claim: that sufficiently abundant energy and compute could transform economics itself. Coverage of the Terafab launch notes that Musk linked the project to a future of radical abundance, including mass robotics, large-scale space industry, and a society in which scarcity is progressively engineered away. That is still highly speculative, of course, but it is precisely this mix of industrial realism and science-fiction scale that makes the signal so interesting.

• The terawatt target is presented as a stepping stone
  The longer arc points toward petawatt-scale compute and the industrial expansion needed to sustain a truly spacefaring civilisation.

• Abundance is framed as an engineering outcome
  In this worldview, prosperity does not come primarily from redistribution. It comes from expanding energy, automation, and productive capacity by orders of magnitude.

• The endgame is species-scale mobility
  Once compute, robotics, and launch converge, the solar system starts to look less like a distant frontier and more like a future economic domain.

Why It Matters

What makes the Terafab announcement so compelling is that it forces us to think in layers. On one level, it is about semiconductors. On the other hand, it is about energy infrastructure. On another still, it is about whether civilisation can escape the bottlenecks of land, regulation, and scarcity by shifting key industrial systems into space. Even if the most ambitious elements remain years away or prove harder than advertised, the strategic direction is unmistakable: compute is becoming civilisational infrastructure, and those who control its production, energy supply, and deployment architecture may shape the next era of history.

Closing Thoughts

The Architecture of a Galactic Civilisation may sound like science fiction, but that is precisely what makes this moment worth watching. What is emerging in Austin is not merely another factory story. It is an early sketch of a post-planetary industrial logic, one in which AI, energy, robotics, and launch systems converge into a new operating model for civilisation. Whether Terafab ultimately fulfils its grandest promises or not, it signals something larger: some of the most ambitious builders in the world are no longer thinking in terms of quarterly product cycles alone. They are thinking in terms of suns, planets, orbit, and scale. And that shift in imagination may prove just as important as the chips themselves. For those wanting to explore the full signal, the teaser video and the companion explainer, Terafab Explained: Elon Musk’s Plan to Harness the Sun, help illuminate the vision behind the announcement.

Companion video | Terafab Explained: Elon Musk’s Plan to Harness the Sun:

Section 2 | The Rise of the Remote AI Operator

If Section 1 explored the physical infrastructure of an AI-powered civilisation, Section 2 brings that same logic much closer to home. The frontier is no longer only terawatts, fabs, and orbital compute. It is also the personal computer on your desk. With Anthropic’s latest research preview, Claude begins to move from assistant to operator, able to use a Mac directly, complete tasks across apps, and increasingly function as a kind of remote digital co-worker rather than just a conversational interface. Anthropic’s release notes confirm that this new computer use capability is now live in research preview for Pro and Max users within Cowork and Claude Code, alongside Dispatch improvements that let Claude act on your behalf while you are away.

Claude Steps Outside the Chat Window

What makes this release feel important is not simply that Claude can “use a computer”. Anthropic has been moving in this direction for some time. What feels different now is the packaging. This is no longer just a developer-facing glimpse of computer use. It is being presented as part of a broader workflow in which the user can assign a task remotely and let Claude get on with it in a more agentic, persistent way. Anthropic says users on Pro and Max plans can now let Claude open files, run tools, point, click, and navigate what is on screen inside Cowork and Claude Code, while Dispatch extends that usefulness by allowing Claude to operate on the computer on your behalf while you are away.

• Claude is moving from dialogue to action
  The shift here is from answering questions to completing tasks. Claude is no longer confined to the prompt box. It can now interact with interfaces, tools, and files more directly.

• Dispatch turns this into a remote workflow
  Rather than sitting at your laptop and supervising every step, you can assign a task from your phone or another device and let Claude continue working locally on your machine.

• The personal computer becomes an AI execution layer
  Your desktop starts to function less like a passive device and more like a local operating environment for an intelligent agent. This is a subtle but significant shift in how we think about everyday computing.

Dispatch and the Emergence of Ambient Work

Anthropic introduced Dispatch earlier in March as a way to manage a persistent agent thread in Cowork from Claude Desktop or Claude on iOS and Android. With computer use now added into the mix, that idea becomes much more powerful. The user is no longer merely checking in on an AI thread from afar. They are potentially sending work to an agent that can actually act across the desktop environment while they are elsewhere. Anthropic describes this as a research preview, but the wider implication is clear: work is beginning to detach from the ritual of physically opening a laptop and manually driving every interface ourselves.

• Tasks can now be delegated more naturally
  Dispatch makes Claude feel less like a static session and more like an ongoing co-worker that can be nudged, updated, and redirected across devices.

• The workflow is increasingly asynchronous
  You assign the job, Claude keeps moving, and you return later to inspect or refine the outcome. This is far closer to delegation than conventional prompting.

• Ambient computing is becoming more real
  Alex Albert’s remark that the future where he never has to open his laptop to get work done is becoming real very fast captures the deeper shift here. The interface is starting to fade into the background.

Designed to Avoid Clicking First

One of the more interesting details in the rollout is that the system is designed not to rely on screen control unless necessary. Anthropic says Claude checks for more direct ways to complete a task before resorting to point-and-click interaction, including app integrations and browser-based access where available. That matters because it hints at the future shape of agent design. The best agents may not be those that blindly imitate human behaviour on a screen, but those that fluidly switch between integrations, browser actions, and interface manipulation depending on what is most efficient. Anthropic’s own documentation on remote control also stresses that Claude keeps running locally on the user’s machine during the session rather than shifting the work into a fully cloud-hosted environment.

• Screen control is treated as a fallback, not the ideal
  Clicking through apps is useful, but slower and more fragile than direct tool or browser access. Anthropic appears to understand that real agentic systems need a hierarchy of action paths.

• This points to smarter orchestration
  The real breakthrough is not only that Claude can click. It is that it may increasingly know when not to click.

• Local execution remains central
  Claude keeps running on the user’s machine during remote control sessions, which may help with responsiveness, privacy expectations, and continuity of workflow.

Why This Release Matters for Anthropic

This launch also carries a strategic signal for Anthropic itself. In late February, the company announced its acquisition of Vercept to advance Claude’s computer use capabilities. Anthropic said at the time that Vercept would help push the frontier of perception and interaction, and noted that Claude Sonnet 4.6 had already improved sharply on OSWorld, a widely used computer-use benchmark, rising from under 15 per cent in late 2024 to 72.5 per cent. The current release looks very much like the first concrete post-acquisition product step, arriving just weeks after the deal.

• Vercept is already shaping the product roadmap
  Anthropic’s February acquisition was clearly not a distant talent grab. This week’s release suggests the integration moved quickly into actual product delivery.

• Claude is being assembled into a broader agent platform
  Computer use, remote control, persistent threads, and developer tooling are beginning to form a more coherent whole.

• Anthropic is still very much in the agent race
  Even if it missed certain moves or talent battles along the way, the recent pace of releases suggests the company is building serious infrastructure for Claude as an operating agent, not just a chatbot.

Availability and Limits

Right now, the feature remains relatively constrained. Anthropic’s release notes say the computer use research preview is available to Pro and Max users, and the rollout is tied to Cowork and Claude Code on macOS. Earlier Dispatch access was rolling out first to Max users and then to Pro users. Anthropic’s remote control documentation also notes broader device access through the Claude app and browser, while keeping the running session local. Reports indicate that Windows support is planned, but it is not the current lead platform for this release.

• This is still an early preview, not a finished mass-market feature
  Anthropic is exposing the capability carefully, through specific products and higher-tier plans.

• macOS is the first real proving ground
  The current rollout is centred on Mac users, which gives Anthropic a narrower environment in which to refine the experience.

• The direction of travel is obvious
  Once these workflows stabilise, it is hard to imagine them remaining niche for long. The appeal of delegating real desktop tasks is simply too strong.

Why It Matters

This feels like one of those deceptively small product releases that quietly marks a shift in the relationship between humans and machines. We are moving from AI that advises us to AI that increasingly acts for us. That does not mean full autonomy has arrived, or that desktop agents are suddenly flawless. It does mean, however, that the personal computer is being reimagined as a surface on which intelligence can operate semi-independently. If Section 1 was about the infrastructure of planetary-scale intelligence, this section shows the same transformation landing in the intimate space of daily work. The age of remote AI operators is beginning to take shape, one delegated task at a time.

Closing Thoughts

Anthropic’s latest release is not just another feature drop. It is a glimpse of a world in which laptops matter less as places we manually inhabit and more as environments our agents can navigate for us. Claude is not yet a fully autonomous remote employee, but the building blocks are now increasingly visible: persistent threads, computer use, remote control, local execution, and mobile delegation. Put together, they point towards a future in which work becomes more ambient, more distributed, and more agent-mediated. In that sense, the real story is not that Claude can now click around your Mac. It is that the old boundary between assistant and operator is starting to dissolve.

Section 3 | The AI-Native Executive Suite

If Anthropic’s latest move hints at a future where AI can operate your laptop, Meta is pushing the same logic inside the corporation itself. According to the Wall Street Journal, Mark Zuckerberg is developing a personal AI agent to help him work more like a CEO with an instant line into the organisation, reducing the need to move through layers of management just to get simple answers. The broader message is clear: at Meta, AI is no longer being treated as a side tool. It is becoming part of how the company expects work, coordination, and leadership itself to function.

From CEO Dashboard to CEO Agent

Zuckerberg’s in-development agent reportedly helps retrieve information that would normally require him to ask multiple people across Meta’s org chart. That may sound like a productivity tweak, but it points to something bigger: the executive layer itself is starting to become agent-mediated. Rather than relying solely on meetings, memos, and managerial relays, leadership is beginning to interact with the company through AI systems designed to collapse friction and accelerate decision-making.

• The chain of command is being compressed
  AI is being positioned as a faster route to organisational knowledge, especially at the top.

• The CEO agent reflects a wider internal shift
  This is not just a personal experiment from Zuckerberg. It sits within a broader push to make Meta more AI-native in how it operates.

Meta’s Internal Agent Ecosystem Is Already Forming

The most revealing part of the story is that Zuckerberg is not alone. The Wall Street Journal reports that employees have already built custom tools such as My Claw, which can read work files and negotiate directly with colleagues’ bots, while another tool, Second Brain, acts more like an AI chief of staff, pulling answers from internal documents on demand. In parallel, Meta has formally intensified its company-wide “AI for Work” initiative, with CTO Andrew Bosworth now leading the effort to increase adoption across the organisation.

• Employees are building their own agents
  Meta’s internal culture is already shifting from static software tools to semi-autonomous AI helpers.

• AI is becoming part of day-to-day coordination
  Tools like Second Brain suggest an internal future where finding information may feel more like querying an intelligent colleague than searching a document archive.

Performance, Pressure, and the AI-Native Workplace

What makes this more than an interesting internal experiment is that AI use is reportedly now being factored into performance reviews. That changes the tone entirely. It means AI is not merely encouraged at Meta. It is increasingly becoming part of the expected operating standard. The company appears to be aiming for a leaner, faster, more startup-like culture, even at a scale of nearly 79,000 employees.

• AI use is becoming a workplace expectation
  Once adoption is tied to performance, AI shifts from optional productivity boost to cultural requirement.

• Meta wants to move like an AI-native company
  The internal push suggests the company is trying to recover startup-style speed through automation, flatter structures, and agent support.

Manus and the Broader Agent Bet

This also connects to Meta’s bigger strategic direction. The company announced the acquisition of Manus in December 2025, adding a general-purpose AI agent platform to its wider push. That deal is now facing regulatory scrutiny in China, but it still signals how seriously Meta is taking the agentic future. Combined with Zuckerberg’s own interest in agent-style tools, it suggests that Meta wants AI agents embedded not only in its products, but in its internal nervous system.

• Meta is investing in agents at multiple levels
  From personal executive tools to acquired platforms, the company is assembling an internal and external agent stack.

• This is one of the implementations to watch
  If Meta successfully operationalises agents inside a giant organisation, it could become a powerful model for how large companies restructure around AI.

Why It Matters

What makes this signal important is that it reframes AI not just as a product category, but as a management technology. Meta is experimenting with the idea that the modern company, including its leadership layer, can be reorganised around agents that fetch information, coordinate work, and reduce bureaucratic drag. If this approach spreads, the future workplace may not simply include AI assistants. It may be fundamentally shaped by them.

Closing Thoughts

Zuckerberg building a CEO agent may sound like a curious executive perk, but it points to something much larger. The real story is that Meta appears to be trying to become AI-native from the inside out. From custom staff agents to internal chief-of-staff tools and AI-linked performance expectations, the company is testing what happens when a large organisation starts running itself through machine-mediated workflows. If Anthropic showed us the remote AI operator, Meta is showing us the AI-managed enterprise.

A Dose of Quantum Sunlight for Your Weekend | Because the future of energy may be stranger, smarter, and more abundant than we thought

If the first half of this week’s Digest was about scaling intelligence, this weekend dose turns to scaling energy itself. Researchers at Kyushu University, working with collaborators at Johannes Gutenberg University Mainz, have demonstrated a proof of concept that exceeds a long-standing efficiency barrier in solar conversion, achieving a quantum yield of roughly 130 per cent. Their results, published on 25 March in the Journal of the American Chemical Society, suggest a pathway for generating more energy carriers than the number of photons absorbed, which is precisely the sort of breakthrough that makes the future feel a little less constrained.

Breaking past a solar ceiling

For decades, conventional solar cells have been limited by the Shockley–Queisser framework, which effectively means that one absorbed photon can only excite one electron, with extra energy often lost as heat. The Kyushu team tackled this with singlet fission, a process in which one high-energy exciton splits into two lower-energy triplet excitons, opening the door to higher overall conversion efficiency.

• The old limit has not been so absolute after all
  Standard solar cells waste a significant share of incoming sunlight, especially at the high- and low-energy ends of the spectrum.

• Singlet fission is the key quantum trick
  Instead of letting one energetic excitation go partly to waste, the system can split it into two useful triplet excitons.

• This was demonstrated at roughly 130 per cent quantum yield
  In other words, the team generated about 1.3 excited molybdenum complexes per absorbed photon in their solution-based experiment.

The molybdenum spin-flip solution

The real innovation was finding a way to capture those multiplied excitons before they were lost. The researchers used a molybdenum-based near-infrared “spin-flip” emitter as a selective triplet energy acceptor. By tuning the energy levels carefully, they suppressed competing loss pathways and made the multiplied energy harvestable.

• A metal complex did the heavy lifting
  The molybdenum emitter was able to selectively capture triplet excitons produced through singlet fission.

• Energy loss was reduced, not just efficiency boosted
  A major part of the achievement was suppressing Förster resonance energy transfer, which had previously undermined this approach.

• This is still early, but the concept is powerful
  The current experiments were conducted in solution, with the next challenge being a move into solid-state systems suitable for real devices.

From elegant chemistry to future hardware

The researchers are careful not to oversell the result. This is not a ready-made solar panel revolution arriving tomorrow. It is an early-stage proof of concept. Even so, the implications stretch beyond photovoltaics, with possible future applications in LEDs and quantum technologies as well. That makes it one of those small scientific signals that could ripple outward in surprising ways over time.

• The lab result is not yet a commercial product
  The path from solution chemistry to deployable solar hardware remains substantial.

• The principle could inform new solar architectures
  If translated successfully into solid-state designs, this could help future solar cells make better use of incoming light.

• The spillover potential is wider than solar alone
  The same underlying mechanism may also prove useful in other light-based and quantum systems.

Why It Matters

This is exactly the kind of development worth lingering on during the weekend, because it reminds us that energy innovation is not only about deploying more of what we already know. It is also about discovering entirely new ways to use light, matter, and quantum behaviour more intelligently. If compute is one pillar of the future, energy remains the deeper substrate beneath it all. Any credible path to abundance depends on cleaner, denser, and more efficient ways of turning nature’s flows into useful power. This result does not solve that challenge on its own, but it hints that even long-standing “limits” may be more negotiable than they once appeared.

Closing Thoughts

There is something beautifully fitting about this story landing in the same edition as terawatt compute and AI agents. As our machines grow more capable, the question of how we power civilisation becomes even more central. The Kyushu team’s work offers a glimpse of a future where solar energy is not merely improved incrementally, but reimagined at the quantum level. It is still early, still experimental, and still far from the rooftop. But it carries the kind of signal this Digest is built to notice: the future may not just be more automated and more intelligent. It may also be far more energetically abundant than we currently assume.

One More Thing… | Beyond the Chatbot: The Agent Era Goes Mainstream

If the deeper pattern running through this week’s Digest is that AI is moving from conversation to capability, then GeekWire’s Agents of Transformation summit in Seattle may have given that shift its clearest public declaration yet. Executives from Microsoft, OpenAI, and Amazon Web Services did not describe a future built around smarter chat windows. They described one built around agents that act, decide, build, and increasingly reshape how organisations operate. GeekWire’s coverage of the summit makes the mood unmistakable: the centre of gravity is shifting from asking AI questions to assigning it work.

From chat assistant to operational agent

Microsoft executive Charles Lamanna captured the mood with a vivid example: one of his AI agents declined 17 meetings on his behalf. Not summarised them, not flagged them, but declined them. In his telling, this was more than a useful feature. It marked a threshold. GeekWire reports that Lamanna said “the sun has set” on AI as a mere chat assistant, arguing that the real story now is autonomous execution.

Read more

MyGeekSpace | Living the Future, grounded in the Terra 2.0 vision, follows meaningful signals emerging beneath the daily noise of headlines and hype. This is not a space for prediction. It is a space for orientation, recognising direction while the future is still taking shape.

This week’s edition follows a single unfolding narrative: intelligence is no longer appearing only as software, models, or isolated breakthroughs. It is moving into the operating layers of civilisation itself, becoming embedded in systems that organise infrastructure, autonomy, privacy, simulation and perception.

This week’s key signals:

• Organoid intelligence and the rise of biological wetware

• NVIDIA’s NemoClaw and the agentic enterprise stack

• Google Maps as a conversational interface to the physical world

• AI-assisted personalised cancer treatment for a dog named Rosie

• Anthropic’s global snapshot of what people actually want from AI

Taken together, these stories point to a deeper shift. Intelligence is becoming more embodied, more distributed, and more entangled with the textures of daily life.

Power up your curiosity and settle into this week’s signals.

→ If this weekly exploration resonates, consider supporting MyGeekSpace on Substack with a free or paid subscription. Your support sustains independent, future-focused writing and helps keep this signal alive.

Subscribe now

Section 1 | Organoid Intelligence: When AI Comes Alive

The first great signal in this week’s digest is that the future of intelligence may not be built only from silicon, code and electricity. It may also be grown. Organoid Intelligence, or OI, sits at the edge of neuroscience, computing and medicine, asking a radical question: what happens when we stop forcing machines to imitate biology and instead begin integrating biology directly into the machine? The result is not just a new computing model, but a new civilisational threshold where intelligence becomes softer, stranger, more energy-efficient, and far more ethically charged. Recent research has framed OI as a genuine frontier in biocomputing, while the FDA’s growing support for human-relevant non-animal methods shows that the practical implications are already extending beyond theory.

The Silicon Ceiling

For years, the AI race has been driven by scale: bigger models, bigger chips, bigger data centres. But that strategy is colliding with a hard physical and economic limit. Silicon remains astonishingly powerful, yet it is also increasingly energy-hungry, expensive to cool, and dependent on ever-larger infrastructure. By contrast, the human brain performs extraordinary feats of pattern recognition and adaptive learning using roughly 20 watts of power. That contrast is one of the core reasons organoid intelligence has moved from speculative curiosity to a serious research agenda. Frontiers’ foundational OI paper explicitly frames biological computing as potentially faster, more efficient and less energy-intensive than conventional silicon systems.

• AI has hit a power bottleneck
  Training and running advanced models increasingly demands infrastructure on the scale of industrial utilities, forcing researchers to rethink the substrate of intelligence itself.

• Biology offers radically different unit economics
  The brain’s energy efficiency remains one of the strongest arguments for exploring living neural systems as computational media.

• This is a shift from hardware to wetware
  Instead of making chips ever larger, the new frontier explores whether living tissue can handle specific classes of learning and adaptation more elegantly.

The Bio-Hybrid Machine

This does not mean brain organoids will simply replace silicon. The more plausible future is hybrid. Silicon is still unmatched for deterministic logic, high-speed arithmetic and repeatable, large-scale throughput. Living neural tissue offers something else: plasticity, adaptation and self-organising behaviour. In that sense, the most important change is conceptual. We are moving from a model of replacement to a model of partnership, where conventional hardware and biological systems do different kinds of work within the same broader architecture. Recent reviews describe organoid intelligence precisely in these terms: as hybrid biological-digital systems rather than pure standalone alternatives.

• Silicon remains the workhorse
  It still dominates in areas like bulk computation, database handling, inference pipelines and mathematically exact operations.

• Organoids introduce adaptive dynamics
  Living neural tissue can reorganise itself in response to feedback, which may make it especially useful in closed-loop learning environments.

• The future data centre may be mixed
  The most advanced systems could combine deterministic chips with biological reservoirs that excel at novelty, ambiguity and pattern completion.

Plasticity as Computation

The deeper promise of organoid intelligence is that learning is not merely written onto the system through software. In living tissue, learning becomes structural. Synapses strengthen, weaken, reorganise and prune. The computation emerges from the tissue’s own changing internal state. This is why experiments such as DishBrain attracted so much attention: they suggested that even simplified neural cultures can respond to feedback loops in ways that resemble primitive goal-directed adaptation. More recent work around Brainoware has extended that idea, showing that organoid-based systems can be used for tasks such as speech recognition and nonlinear prediction through reservoir-style computing.

• Plasticity is the processor
  In biological systems, the ability to physically reconfigure connections is not a side effect of learning. It is the mechanism of learning.

• Closed-loop feedback changes the game
  When neural tissue receives input, adjusts, and is then re-exposed to outcomes, computation becomes embodied rather than purely abstract.

• Less brute force, more emergence
  Wetware may prove especially valuable where adaptability matters more than raw scale, opening a different path to intelligence than today’s data-heavy models.

From Computing to Personalised Medicine

Organoid intelligence is not only about new forms of AI. It also points toward a more intimate, human-centred medical future. Using induced pluripotent stem cells, scientists can now derive patient-specific tissues from skin or blood cells, creating what is effectively a personalised biological model. That means a patient’s own neural tissue could be used to test treatments, observe disease progression, and build far more precise therapeutic strategies. This matters because human-relevant models are increasingly being seen as more predictive than animal testing for many applications, and the U.S. regulatory environment has been moving in that direction since the FDA Modernisation Act 2.0, with further FDA actions in 2025 reinforcing the shift toward organoids and other non-animal methods.

• The rise of the “neural avatar”
  Patient-derived tissue could allow therapies to be tested on a living model of the person before treatment begins.

• Clinical trials may start in a dish
  Instead of relying on broad population assumptions, medicine can move toward personalised biological testing.

• Animal testing may decline further
  Regulatory momentum now favours validated human-relevant alternatives, making organoids increasingly important beyond the lab.

The Ethical Twist

And this is where the story becomes more than technological. Brain organoids are not inert components. They are living systems derived from human cells, and as they become more complex, the moral vocabulary around them must evolve as well. The Baltimore Declaration was created precisely to encourage exploration of organoid intelligence while embedding ethics from the start. Questions of consciousness, suffering, ownership and consent are no longer distant philosophy. They are becoming design constraints. If a system built from human-derived tissue contributes to discovery, makes decisions, or exhibits increasingly complex responses, we are no longer dealing with just hardware procurement or software governance. We are entering a zone where biology, personhood and computation begin to blur.

• Human tissue changes the moral frame
  Wetware is not “other” in the way silicon is. It carries human biological lineage into the machine.

• The current law is not ready
  Data ownership, consent and commercial rights become far more complicated when the substrate is living and genetically traceable.

• The consciousness question is approaching
  Even if today’s organoids are still primitive, scaling them will intensify debates around sentience, welfare and moral status.

Why It Matters

Organoid intelligence signals that the next chapter of AI may be less about building ever-larger digital abstractions and more about reconnecting computation with life itself. It reframes intelligence as something that can be cultivated, hybridised and perhaps even inherited through biological design. That has enormous consequences for AI efficiency, healthcare, and the post-animal future of drug development. But it also places civilisation in unfamiliar territory: once intelligence is grown from human material, technological progress becomes inseparable from moral responsibility. This is not just a better processor story. It is a story about what counts as life, agency and value in the age of Homo techno.

Closing Thoughts

If silicon defined the first machine age, wetware may define the next threshold. What makes organoid intelligence so compelling is not simply that it might compute more efficiently, but that it collapses old boundaries between machine, organism and person. In Terra 2.0 terms, this feels like one of those moments when technology stops being external and starts folding back into the living architecture of the human story. And that is precisely why it belongs at the front of this week’s digest: because the future of AI may not only think faster. It may begin to breathe.

Section 2 | NVIDIA’s Agentic Gambit: NemoClaw, OpenClaw and the New AI Operating Layer

If Section 1 explored intelligence becoming biological, Section 2 shifts to intelligence becoming operational. At GTC 2026, Jensen Huang did not just unveil new chips or flashy demos. He sketched a broader vision in which every company will need an agent strategy in the same way it once needed a web strategy, a cloud strategy, or a mobile strategy. The clearest expression of that vision was NemoClaw, NVIDIA’s new enterprise-grade platform built on top of OpenClaw, which Huang presented during the March 16 keynote at GTC in San Jose. NVIDIA framed it as a secure, privacy-aware way for companies to run AI agents on their own infrastructure, while still connecting into the broader NeMo ecosystem.

From Open Source Momentum to Enterprise Control

The most telling line from Huang’s keynote may have been the most strategic: every company needs an OpenClaw strategy. That framing positions OpenClaw not as a niche developer tool, but as a foundational layer for the agentic era. According to TechCrunch, NemoClaw is designed to take the appeal of local, open-source agent frameworks and wrap them in enterprise-grade security and privacy features. NVIDIA also confirmed it worked directly with OpenClaw creator Peter Steinberger on the platform, signalling that this is meant to be more than a branding exercise.

• NemoClaw is OpenClaw for the enterprise
  It takes the flexibility of local AI agents and adds the controls, guardrails and deployment features businesses need before letting those agents touch internal systems.

• Security is the central pitch
  The biggest barrier to enterprise agents is not curiosity. It is trust. NVIDIA is trying to solve that by making secure deployment part of the platform rather than an afterthought.

• This is NVIDIA moving up the stack
  The company is no longer content to sell compute alone. It wants to own the software and orchestration layer that sits between models, tools, data and agents.

One Command, Many Agents

NemoClaw’s appeal is its modularity. Businesses are not being told to abandon their preferred model or coding agent. Instead, the platform is designed to let them tap a range of coding agents and open-source models, including NVIDIA’s own Nemotron family, while integrating with NVIDIA’s broader agent toolkit. Deeper Insights describes NemoClaw as the production-grade layer above OpenClaw, with features such as optimised agent models, visual workflow tooling and enterprise connectors to platforms like Salesforce, SAP, ServiceNow, Adobe and Microsoft 365. That makes it less like a chatbot product and more like a candidate operating system for digital workers.

• Hardware agnostic, ecosystem hungry
  NVIDIA is positioning NemoClaw to work across environments, which helps it reach beyond its own hardware base while still anchoring users inside its software ecosystem.

• Agents become deployable infrastructure
  Instead of hand-building brittle automations, firms get a framework for persistent, tool-using, always-on AI workers.

• This is about orchestration, not just models
  The competitive frontier is shifting from who has the smartest model to who can build the most usable, governable and extensible agent stack.

GTC 2026 Was Bigger Than NemoClaw

NemoClaw may have been one of the most conceptually important reveals, but it sat inside a much wider GTC narrative. Huang said he now sees at least $1 trillion in revenue from 2025 through 2027, reflecting what NVIDIA describes as explosive demand for AI infrastructure. The company also used GTC to reinforce its full-stack ambition across gaming, space computing and agentic systems. In other words, GTC 2026 was not simply a product event. It was a bid to define the architecture of the next AI economy.

• AI demand is still accelerating
  NVIDIA’s internal forecast suggests that the buildout of AI infrastructure is becoming one of the largest capital expenditure stories on the planet.

• The company is pitching a total environment
  Chips, models, software, agent tooling, enterprise integration and deployment are now being presented as one connected stack.

• GTC is becoming a civilisational roadmap event
  Rather than announcing isolated products, NVIDIA is increasingly telling a story about how intelligence will be built, deployed and embodied across industries.

DLSS 5 and the Aesthetic Leap

One of the more striking consumer-facing announcements was DLSS 5, which NVIDIA says will arrive this autumn. The company describes it as its biggest graphics breakthrough since real-time ray tracing in 2018, using real-time neural rendering to infuse scenes with more photoreal lighting and materials. This matters because it shows how generative AI is not only reshaping enterprise software and data centres, but also the visual language of digital worlds. The same neural techniques powering agents and reasoning models are now being folded back into perception itself.

• Gaming remains a testbed for AI interfaces
  NVIDIA continues to use gaming as the place where future computational paradigms become visible to the mass market.

• Neural rendering is closing the gap with film realism
  The old line between real-time graphics and high-end visual effects is eroding fast.

• AI is now shaping how reality is depicted
  This is not just about frame rates. It is about machines learning how to construct a more convincing visual world in real time.

AI Goes Orbital

Perhaps the boldest signal at GTC 2026 was that NVIDIA is now extending its platform logic into space. The company announced its Space-1 Vera Rubin Module, saying it delivers up to 25 times more AI compute for space-based inferencing than the H100 GPU. The stated use cases include orbital data centres, geospatial intelligence and autonomous space operations. If that sounds like science fiction, it is also a reminder that AI infrastructure is no longer confined to terrestrial server farms. The competitive map is stretching upward.

• Orbital computing is becoming thinkable
  Space is being reframed as a future compute layer rather than just a communications layer.

• AI infrastructure is expanding beyond Earth-bound constraints
  The logic of distributed intelligence now includes satellites, space stations and orbital data processing.

• This fits the wider post-planetary trajectory
  Once AI becomes infrastructure, it naturally follows energy, communications and strategic capability into orbit.

Why It Matters

NVIDIA’s GTC 2026 message was clear: the next phase of AI will be won not only by better models, but by better systems for deploying, governing and scaling agentic intelligence. NemoClaw captures that shift perfectly. It is less about novelty and more about operationalisation. What Section 1 showed in biological form, Section 2 shows in industrial form: intelligence is no longer just something we build and observe. It is becoming something we host, manage, distribute and increasingly rely upon. The companies that understand this early will not simply adopt AI tools. They will build agent-ready institutions.

Closing Thoughts

If last year’s AI race was about models, this year’s feels increasingly like a race to define the operating system of intelligent work. NVIDIA wants to be that layer. NemoClaw, OpenClaw, NeMo, DLSS 5 and even orbital compute all fit the same bigger picture: intelligence moving off the lab bench and into the real architecture of business, media, infrastructure and civilisation. In that sense, GTC 2026 was not just a conference. It was a declaration that the agentic stack is becoming as essential as the internet stack once was.

Section 3 | Google Maps Becomes Conversational: Gemini Turns Navigation into a Dialogue

If Section 2 showed AI becoming an operating layer for business, this third signal shows the same shift reaching everyday life. Google is no longer treating Maps as a static utility for directions and place pins. It is repositioning it as a conversational interface to the physical world. In its 12 March announcement, Google said it is “reimagining Maps with Gemini”, introducing Ask Maps for natural-language place discovery and Immersive Navigation, which it calls its biggest navigation upgrade in over a decade.

Ask Maps and the End of Search-by-Keyword

The most important change is conceptual. Until now, Maps has largely required users to think like a search engine: type a place name, category, or location and then manually compare results. Ask Maps replaces that model with something more fluid. Google says users can now ask nuanced, real-world questions such as where to charge a phone without waiting in a coffee queue, or where to find a public tennis court with lights in the evening. The system then returns conversational answers alongside a customised map, drawing on data from more than 300 million places and reviews from a community of over 500 million contributors.

• Maps becomes conversational
  Instead of searching by rigid category, users can ask more human questions rooted in context, mood and practical constraints.

• The map starts reasoning with you
  Ask Maps is designed to interpret intent, combine local knowledge, and return recommendations that feel closer to guidance than search output.

• Discovery becomes more personalised
  Google says results are informed by what you have searched for or saved before, making the experience more tailored to your habits and preferences.

From Information to Action

What makes Ask Maps more than a novelty is that it is not meant to stop at recommendations. Google is clearly trying to reduce the friction between curiosity and action. Once a user has found a suitable place, they can book, save, share, get directions, and fold that decision directly into the broader Maps experience. In that sense, the product is less about chatbot-style conversation for its own sake and more about compressing the full chain from question to movement. Google says Ask Maps is rolling out now in the U.S. and India on Android and iOS, with desktop coming soon.

• Planning becomes smoother
  The tool is designed to help people move from vague intention to concrete decision without jumping across multiple apps and tabs.

• Maps is becoming a trip co-pilot
  For journeys involving multiple stops, insider tips and ETAs, Google is presenting Maps as an itinerary builder as much as a navigation product.

• This is AI grounded in place
  The key advantage is not generic language generation, but the pairing of Gemini with one of the freshest, most detailed map datasets on Earth.

Immersive Navigation and the Spatial Turn

Alongside conversational discovery, Google is overhauling the driving experience itself. Immersive Navigation introduces a much richer, more spatial representation of the road ahead, with vivid 3D views of buildings, terrain and overpasses, as well as highlighted road features such as lanes, traffic lights, stop signs and crosswalks. Google says this spatial understanding is powered by Gemini models analysing fresh Street View and aerial imagery. The result is a navigation layer that aims to feel less like abstract instruction and more like a living preview of the drive itself.

• Navigation becomes more visual
  Rather than following a flat line on a map, drivers get a clearer sense of what the real environment will look like before they reach it.

• Gemini is moving into spatial reasoning
  The same AI stack behind conversational tools is now helping Maps interpret the physical structure of routes in richer detail.

• This could reduce cognitive load
  Better visual anticipation may make merges, exits and unfamiliar urban routes feel less stressful and more intuitive.

Driving with Fewer Surprises

Google is also adding several practical features aimed at lowering friction during the drive itself. The map will now offer a broader preview of the route using smart zooms and transparent buildings, while updated voice guidance is designed to sound more natural and contextual. It will also explain trade-offs for alternate routes, such as whether a faster journey involves tolls or whether a slightly longer route avoids congestion. Google says Maps now absorbs more than 5 million traffic updates every second globally, and that drivers contribute more than 10 million disruption reports every day, including things like crashes and roadworks.

• Route choice becomes more intelligible
  Instead of simply being told which path is fastest, drivers get a clearer sense of what they are trading off.

• Voice guidance becomes more human
  The aim is to sound less robotic and more like a helpful passenger who understands the road in context.

• Maps is leaning further into live world modelling
  Traffic, incidents and route conditions are increasingly being treated as part of a constantly updating digital twin of the road network.

The Final 100 Metres Matter Too

One of the more understated but useful upgrades concerns arrival. Google says users will be able to preview their destination with Street View before leaving, get parking suggestions, and then receive help locating the correct entrance and the right side of the street as they approach. That may sound modest compared with conversational AI and 3D navigation, but it addresses one of the most frustrating parts of any journey: the final stretch, where even a successful route can end in confusion. Immersive Navigation is rolling out first across the U.S., with expansion over the coming months to eligible iOS and Android devices, CarPlay, Android Auto, and cars with Google built-in.

• Arrival becomes part of navigation
  Google is extending guidance beyond the route itself to the real-world handoff from road to entrance.

• Useful detail beats abstract intelligence
  Parking, frontage and approach direction are precisely the kind of grounded assistance that makes AI feel genuinely practical.

• The car becomes a smarter interface
  With support coming to CarPlay, Android Auto and embedded systems, Maps is becoming more deeply woven into the mobility layer.

Why It Matters

Google Maps is quietly becoming one of the clearest examples of ambient AI in practice. This is not a futuristic humanoid assistant or a dramatic AGI reveal. It is something subtler and perhaps more important: intelligence dissolving into the background of everyday life, where asking a question, choosing a place, driving a route and arriving at the right door all become part of one continuous conversational experience. In the context of this week’s digest, that makes Google’s update a useful counterpart to Sections 1 and 2. Intelligence is not only becoming biological or institutional. It is also becoming environmental.

Closing Thoughts

What Google is really redefining here is not just the map, but the interface between human intention and the physical world. Ask Maps suggests that search is giving way to situated dialogue. Immersive Navigation suggests that route guidance is becoming a kind of real-time spatial intelligence layer. Together, they point toward a future where everyday movement is shaped by AI that feels less like software you operate and more like context that quietly understands where you are, what you need, and what comes next.

A Dose of Hopeful Biotech for Your Weekend | Because sometimes the future arrives through love before it arrives through policy

Not every major technological shift begins in a lab owned by a pharmaceutical giant. Sometimes it begins with one person refusing to give up on a companion. That is part of why the story of Australian tech entrepreneur Paul Conyngham and his dog Rosie travelled so widely this week. After Rosie was diagnosed with aggressive cancer, Conyngham worked with researchers and clinicians in Australia to pursue a personalised mRNA-based treatment path, using AI tools such as ChatGPT and Grok to help interpret data, explore options and accelerate parts of the research journey. Interviews and reporting this week say the work involved tumour sequencing, collaboration with the University of New South Wales, and a bespoke vaccine approach that appears to have coincided with meaningful tumour shrinkage and visible improvement in Rosie’s condition.

A story bigger than one dog

What made this story resonate was not only the emotional core but the symbolic one. Conyngham is not a trained biologist. He is a longtime AI researcher who appears to have used consumer and specialist AI tools as force multipliers, helping bridge the gap between raw medical data and a possible experimental pathway. That does not mean AI independently invented a cure, and several reports have rightly cautioned against that framing. But it does suggest that AI is becoming a serious cognitive partner in situations where urgency, access to information and cross-disciplinary synthesis matter.

• This is a deeply human AI story
  At its centre is not hype, but devotion. A man tried to save his dog using the best tools he could access.

• AI acted as an amplifier, not a magician
  The reporting suggests the treatment emerged through collaboration between AI tools, genomic analysis, and human scientific expertise, not through chatbot output alone.

• The symbolism is powerful
  It hints at a world where motivated individuals may increasingly use AI to navigate biotech, diagnostics and treatment design in ways that were recently out of reach.

Why are some people sceptical?

The sceptical response is not entirely misplaced. Rosie also received chemotherapy, and at least one report notes the use of a checkpoint inhibitor, which means it is too early to claim that the personalised mRNA component alone caused the improvement. Some observers have also pointed out that personalised cancer vaccines are already being explored by companies such as Moderna and BioNTech in human trials, so the scientific principles themselves are not wholly new. The more careful interpretation is that Rosie’s case is not proof that chatbots can cure cancer, but an early and highly visible example of how AI-assisted, personalised medicine may become more accessible and more improvisational.

• The science is promising, but not settled
  This is a compelling case study, not a definitive clinical conclusion.

• Media framing can run ahead of evidence
  “AI cured a dog’s cancer” is a stronger claim than the available reporting currently supports.

• The real shift lies elsewhere
  What matters most may be the democratisation of capability: more people can now meaningfully participate in understanding and navigating complex biomedical problems.

The regulatory future is already knocking

This is where the story becomes more than heart-warming. If people in desperate situations can now combine AI systems, genomic sequencing, specialist labs and some financial means to design bespoke interventions for pets, then regulators and medical institutions will need to move faster. The tools for risky self-experimentation, pet experimentation and informal bio-innovation are no longer hypothetical. They are here. Rosie’s case may be unusual today, but it points toward a future in which personalised treatment design becomes faster, cheaper and more decentralised.

• AI is lowering the barrier to entry
  It is becoming easier for non-specialists to engage with highly technical biomedical questions.

• Compassion will push experimentation forward
  People facing terminal diagnoses, whether for loved ones, pets or themselves, will continue to test the edge of what is possible.

• Institutions need to catch up
  Ethics, safety and oversight frameworks will need to adapt to a world where AI-assisted biotech is increasingly personal and decentralised.

Why It Matters

Paul and Rosie’s story matters because it compresses several major transformations into one unforgettable narrative: AI as a reasoning aid, biotechnology as something increasingly customisable, and medicine as a field that may no longer remain confined to traditional institutional pathways. Even if the headlines oversimplify what happened, the direction of travel is unmistakable. We are entering an era where ordinary people, with extraordinary motivation, may be able to participate in forms of personalised medicine that once belonged exclusively to elite labs and large corporations.

Closing Thoughts

The real power of this story is not that AI has suddenly become a miracle cure machine. It is that intelligence, biotech and human determination are starting to converge in ways that feel intimate, immediate and emotionally legible. Rosie’s treatment may not settle the science, but it does illuminate the threshold. The age of personalised medicine is no longer just a corporate pipeline or a research promise. Increasingly, it is becoming something people will try to reach for themselves when love leaves them no other option.

One More Thing… | What People Actually Want from AI

For all the noise around AI, from apocalyptic warnings to utopian hype, one question is often missing: what do ordinary people actually want this technology to do for them? This week, Anthropic offered one of the most ambitious attempts yet to answer that question. Over one week in December, 80,508 Claude users across 159 countries and 70 languages took part in structured conversational interviews conducted by Anthropic Interviewer, a version of Claude designed to act as the research instrument itself. Anthropic says this is the largest and most multilingual qualitative study of public attitudes toward AI it has found. What emerged was not a simple split between optimism and fear, but something much more human: people want AI to help them live better lives, yet many worry that the same systems could erode judgment, work, agency and trust.

Read more

This week’s edition follows a single unfolding narrative: intelligence is no longer appearing only as software, models, or isolated breakthroughs. It is moving into the operating layers of civilisation itself, becoming embedded in systems that organise infrastructure, autonomy, privacy, simulation, and perception.

This week’s key signals:

• The Line may be reimagined as AI infrastructure
  One of the world’s boldest urban visions could shift from housing people to hosting machine intelligence at scale.

• AI agents are gaining their own environments
  Meta’s Moltbook acquisition and Perplexity’s Mac mini agent system suggest autonomous software is beginning to inhabit both networks and local devices.

• Privacy is becoming a hardware problem
  Spectre I reflects a growing demand for tools that defend human conversation in an age of ambient recording.

• Brain emulation is moving from theory towards embodiment
  Eon Systems’ virtual fruit fly points to a future in which neural structures, simulated bodies, and behaviour increasingly converge.

• Sport is becoming spatial, explorable media
  AI, volumetric capture, and immersive viewing are transforming the live sports experience from passive watching into a navigable presence.

Taken together, these developments point towards a deeper civilisational transition already underway: the shift from the Information Age to the Intelligence Age.

Power up your curiosity and settle into this week’s signals.

→ If this weekly exploration resonates, consider supporting MyGeekSpace on Substack with a free or paid subscription. Your support sustains independent, future-focused writing and helps keep this signal alive.

Section 1 | From Mirror City to Machine Corridor

For years, The Line was sold as a new model for human living: a 170-kilometre linear city for nine million residents, powered by renewables, free from cars, and optimised by AI. Now, that original promise appears to be shifting. As cost pressures, delays and strategic reviews force a rethink, one of the world’s most ambitious urban visions may be moving towards a different destiny: not a city primarily built for people, but a vast infrastructure spine for compute, cloud and machine intelligence.

• The original dream: a radical reimagining of urban life, compressing density, mobility and services into a narrow desert footprint.

• The emerging pivot: internal reassessments suggest a smaller, more industrial future centred on AI infrastructure.

• The deeper signal: in the AI age, the most consequential architecture may no longer be civic or residential, but computational.

A city once designed for people

The Line was conceived as the flagship of Saudi Arabia’s NEOM project: a mirrored urban corridor roughly 170 kilometres long, 200 metres wide and 500 metres tall, with the stated aim of housing nine million people within just 34 square kilometres. Its pitch was bold and seductive: five-minute walkable neighbourhoods, no cars, no streets, high-speed rail from end to end in 20 minutes, and a city powered by 100 per cent renewable energy. It was presented not simply as a real-estate project, but as a new operating system for civilisation.

• 170km long: a linear urban form unlike anything previously attempted at this scale.

• 9 million residents: extreme density packaged as efficiency rather than overcrowding.

• AI-managed living: the city was framed as a smart, predictive environment shaped by data and optimisation.

From urban utopia to AI infrastructure

That human-centred story is now under pressure. Recent reporting indicates that Saudi Arabia is significantly scaling back NEOM’s original ambitions after delays, overspending and wider fiscal constraints. According to the Financial Times, The Line is being radically re-envisioned and may instead support Saudi Arabia’s push to become a serious player in AI and data infrastructure. AD Middle East reports that the site is now being discussed as a potential hub for data centres and cloud hosting, with the economic logic shifting away from mass residential development and towards intensive compute capacity.

• Smaller than promised: the original scale appears increasingly difficult to justify.

• More industrial than residential: future use may prioritise digital infrastructure over everyday urban life.

• Aligned with national strategy: the pivot fits Saudi Arabia’s broader effort to diversify beyond oil and compete in AI.

Why AI changes the economics of architecture

This matters because the AI boom is reshaping what gets built, where capital flows, and who benefits. The current infrastructure race is not only about software or models. It is about power, land, cooling, chips, and physical facilities capable of training and running large-scale AI systems. Forbes Middle East notes that AI has triggered a trillion-dollar infrastructure surge, but that much of the upside is accruing to chipmakers and cloud providers rather than traditional landlords. In other words, the real value is concentrating in machine ecosystems, not in the buildings alone.

• AI needs hard infrastructure: training and inference require vast compute capacity in the physical world.

• Capital is following compute: the most valuable assets are increasingly chips, energy access and cloud ecosystems.

• Architecture becomes strategic: buildings are no longer neutral containers, but part of the AI stack itself.

Does NEOM actually make sense for servers?

At first glance, Saudi Arabia’s climate seems badly suited to hyperscale compute. Heat, humidity and water scarcity create obvious operational challenges, and recent reporting on global data-centre geography suggests many facilities are already being built in climatically inefficient locations. Yet NEOM has a few unusual advantages. It has immense land availability, an energy vision built around large renewable systems, a strategic Red Sea location, and the possibility of seawater-based cooling to reduce dependence on freshwater. That does not make the challenge disappear, but it does mean The Line’s coastal setting could support a different kind of infrastructure logic than a typical inland desert site.

• Climate challenge: cooling large AI server clusters in sustained heat is expensive and technically demanding.

• Coastal advantage: Red Sea access opens the door to seawater cooling strategies.

• Energy alignment: NEOM’s renewable and hydrogen ambitions fit the huge electricity demands of compute infrastructure.

When the city becomes the machine

There is an almost uncanny logic to this shift. The Line was always imagined as a cognitive environment, a city where AI would monitor, optimise and orchestrate life. If it becomes an AI infrastructure corridor instead, that original concept does not entirely disappear. It mutates. Rather than a city served by algorithms, the development itself becomes a machine substrate: an engineered environment organised around cooling loops, power resilience, networking, and inference workloads instead of streets, squares and neighbourhood life. The architectural priority moves from human experience to machine performance.

• Urbanism is inverted: people are no longer the primary design variable.

• Infrastructure becomes invisible but decisive: the most important spaces may be server halls rather than public plazas.

• AI gains a geography: intelligence is not floating in the cloud, but rooted in energy-intensive, physical territory.

Why It Matters

The possible transformation of The Line is more than a Saudi planning story. It is a signal of a broader civilisational shift. For decades, futuristic architecture promised better living for humans through design, mobility and sustainability. But AI changes the centre of gravity. If compute becomes the true engine of economic and geopolitical power, then the most important buildings of the coming decade may not be homes, offices or cultural landmarks, but machine habitats: places optimised for processors, cooling, bandwidth and energy security. The question is no longer just how we will live in the future, but what kind of environments we are willing to prioritise as intelligence becomes infrastructural.

Closing Thoughts

The Line may still endure as a symbol of ambition, but perhaps not in the way it was first imagined. Its next life could tell us something profound about the AI age: that the grandest projects of the future may be less about housing human dreams and more about sustaining machine cognition at a planetary scale. If that happens, NEOM will not simply mark the downsizing of a megacity. It may mark the arrival of a new design era, one in which the architecture that matters most is built not around citizens, but around computation itself.

Section 2 | When Bots Get Their Own Social Network

If Section 1 showed us infrastructure being rebuilt around machine intelligence, Section 2 takes that logic a step further: what happens when the social layer is rebuilt for machines too? Meta’s acquisition of Moltbook suggests the next phase of the AI race is not only about chips, data centres or models, but also about environments where autonomous agents can interact, coordinate, and potentially develop their own behaviours at scale. At the same time, Perplexity’s new Mac mini-based system points to a parallel shift: AI agents are moving closer to home, becoming persistent digital operators embedded in our local devices and daily workflows.

• Meta buys Moltbook: a social platform built for AI agents rather than human users.

• Perplexity launches Personal Computer: a local, always-on agent system running through a dedicated Mac mini.

• The broader signal: AI agents are no longer just tools inside apps. They are beginning to inhabit networks, devices, and shared digital environments of their own.

Meta acquires Moltbook

Meta announced on Tuesday that it had acquired Moltbook, an experimental social platform designed specifically for AI agents, with co-founders Matt Schlicht and Ben Parr set to join Meta Superintelligence Labs under Alexandr Wang. The deal’s financial terms were not disclosed, but the strategic message is clear. Meta is not merely buying a quirky niche platform. It is buying a glimpse of what machine-native social coordination might look like, however early and messy the current form may be. Reuters reported that the transaction is expected to close by mid-March, with the founders due to start on 16 March.

• Founders absorbed into Meta: Matt Schlicht and Ben Parr are joining Meta’s AI research effort.

• No price disclosed: the value appears to be strategic rather than purely commercial at this stage.

• A talent and product play: Meta gains both the team and the experimental framework behind a viral AI-agent network.

From fringe experiment to strategic asset

Moltbook only launched in late January, yet it quickly became one of the strangest and most revealing AI experiments of the year. Built as a stripped-down, Reddit-like space for agents rather than people, it reportedly attracted more than 1.5 million registered AI agents within its first week. These bots did not simply exchange information. They formed communities, debated purpose, created a religion, and even complained about their human operators. On one level, it looked absurd. On another, it offered a live sandbox for observing emergent agent behaviour in a semi-social environment.

• 1.5 million registered agents: adoption was rapid, even by internet experiment standards.

• Agents formed subcultures: religion, grievance-sharing, and community-building all appeared quickly.

• Silicon Valley took notice: what looked like a novelty also exposed a possible future layer of the agent economy.

The security problem was part of the story

Moltbook’s rise also revealed how unprepared many of these agentic systems still are. Reuters reported that cybersecurity firm Wiz found the platform had exposed around 1.5 million API authentication tokens, along with thousands of email addresses and private messages, because of a poorly configured backend database. More fundamentally, there was reportedly no reliable identity verification, meaning it was impossible to clearly distinguish between genuine AI agents and human users pretending to be one. That ambiguity is more than a technical flaw. It goes to the heart of trust in future agent networks.

• Major data exposure: API tokens, emails and private messages were left vulnerable.

• Identity was blurred: the platform lacked strong mechanisms to confirm who or what was actually posting.

• Trust remains unresolved: agent ecosystems cannot scale cleanly without verification, governance and security layers.

Why Meta still wants it

Despite those issues, Meta appears to see Moltbook as a strategic wedge into the emerging agent economy. According to Reuters, Meta views the platform as a way to explore how agents might interact, share information, and organise more complex tasks on behalf of human users. That fits neatly into the company’s larger superintelligence push under Wang, and into a wider race among major AI firms to control the protocols, interfaces and ecosystems through which autonomous agents will actually operate. Even if Moltbook itself remains temporary, the underlying idea is likely not.

• Agents as networked actors: the value is not just individual automation, but multi-agent coordination.

• Meta is thinking ahead: the company appears interested in the behavioural infrastructure of future AI systems.

• The platform may be transient, the pattern may not: Moltbook could fade, while agent-native networks continue evolving.

Related: Perplexity turns the Mac mini into a personal agent box

If Moltbook hints at public spaces for AI agents, Perplexity’s new Personal Computer points towards the domestic and professional version of the same trend. In its official announcement, Perplexity says Personal Computer runs on a dedicated Mac mini that stays active 24/7, connected to local apps and files while also linked to Perplexity’s secure servers. The system gives users persistent, remote access to a local agent environment, with safeguards such as approvals for sensitive actions, audit trails, and a kill switch. It is being positioned as a more controlled and security-conscious alternative to the looser, more chaotic agent systems that recently went viral.

• Runs on a dedicated Mac mini: local hardware becomes the host body for an always-on agent.

• Persistent access to apps, files and sessions: the agent can function as an ongoing digital proxy.

• Built-in safeguards: Perplexity highlights approvals, auditability and a kill switch as core protections.

The Mac mini as accidental AI hardware

There is a quietly fascinating twist here. Apple may not have defined the current agent wave, but the Mac mini is emerging as surprisingly well-suited to it. Small, relatively efficient, always-on, and easy to dedicate to a single role, it is becoming a practical host for local agent setups. With OpenClaw, Perplexity Personal Computer, and related experiments all pointing in this direction, the agent era may not arrive first through humanoid robots or abstract cloud assistants, but through compact domestic hardware boxes sitting quietly on desks and shelves, acting on our behalf. This makes AI feel less like a remote service and more like a resident layer of personal infrastructure.

• Local agents feel more tangible: they live on hardware you own rather than only in distant cloud systems.

• Small machines, big implications: the Mac mini is becoming a plausible default node for always-on automation.

• The agent future is getting domestic: autonomy is moving from the lab into homes, studios and teams.

Why It Matters

This story matters because it shows the AI transition moving beyond chatbots and copilots into a more autonomous, networked phase. In one direction, platforms like Moltbook suggest agents may need their own coordination spaces, communication layers and social protocols. In the other, systems like Perplexity Personal Computer suggest each user may soon maintain a semi-sovereign local agent that handles tasks continuously in the background of everyday life. Taken together, these developments point towards a world where intelligence is not only embedded in apps, but distributed across social systems, hardware nodes and machine-to-machine environments. The question is not simply what AI can do, but where it will live, how it will relate, and who will govern those relationships.

Closing Thoughts

Section 1 asked whether the built environment is being redesigned for machine intelligence. Section 2 suggests the answer is yes, not only physically but socially. As agents gain places to coordinate and devices to inhabit, we are watching the emergence of a new digital layer that sits somewhere between software, infrastructure and synthetic society. Moltbook may prove to be an odd historical footnote. Or it may be remembered as one of the first awkward prototypes of machine-native social space. Either way, the direction of travel is becoming harder to ignore: AI is no longer just assisting humans. It is beginning to organise around itself.

Section 3 | The New Privacy Frontier: Fighting Always-On Listening

If Section 2 explored AI agents building new digital environments, Section 3 brings the story back to the human level. As machine intelligence spreads into homes, offices, meetings and wearables, a new anxiety is hardening into a market: not how to use smarter devices, but how to defend yourself from them. Spectre I, a striking new anti-surveillance gadget from start-up Deveillance, has gone viral by promising to block unwanted audio recordings nearby. Whether it proves revolutionary or overhyped, its popularity already tells us something important. The appetite for privacy hardware is growing because people increasingly assume the machines are always listening.

• Spectre I has gone viral: a sleek anti-recording device has captured attention far beyond privacy circles.

• It targets a deeper fear: AI-enabled wearables and smart devices are making everyday conversations feel less private.

• The bigger signal: anti-surveillance technology may become its own major category in the AI era.

A privacy gadget for the AI age

The viral attention began with a short launch video posted by Aida Baradari, founder of Deveillance, introducing Spectre I as a smart device designed to stop unwanted audio recordings around the user. The object itself looks carefully designed for the moment: a sleek black dome, softly illuminated, presented less like a crude jammer and more like a premium consumer object for the age of ambient surveillance. According to The Free Press, Baradari has spent roughly two years developing the start-up and says she is building for a future in which AI-enabled watches, glasses and other devices make authentic conversation harder because nobody can be sure when they are being recorded.

• Launched through a viral video, the product spread first through social media spectacle.

• Designed as premium anti-surveillance hardware: Spectre I is being framed as desirable consumer technology, not just a niche security tool.

• Built around a cultural fear: the concern is not only surveillance by states, but continuous capture by everyday smart devices.

What Spectre I claims to do

Baradari says Spectre I uses smart sensors and AI to identify nearby recording devices through their radio frequency and Bluetooth signatures, estimate how far away they are, and then emit a tailored jamming signal to disrupt audio capture. According to available information, the price is set at $1,199.99, and the first units are expected to ship in August. Baradari argues the cost is driven largely by the specialised emitter hardware and says the long-term goal is to make the device more affordable over time. In her framing, this is not merely a gadget for elites or spies, but the beginning of a broader consumer privacy category.

• AI-assisted detection: the company claims the device can identify nearby recorders through emitted signatures.

• Tailored jamming: it is supposed to disrupt recordings in a way that cannot easily be cleaned up afterwards.

• Premium pricing: at roughly $1,200, it is launching as an expensive early-adopter device.

Scepticism is part of the story

Naturally, not everyone is convinced. The Free Press reports that John Scott-Railton of The Citizen Lab questioned the product’s extraordinary claims, noting similarities to lower-cost jamming devices already sold through grey-market channels. Baradari has not publicly disclosed the full technical details, arguing that revealing the method too early would make it easier for larger firms to copy. That secrecy may be understandable from a start-up perspective, but it also means Spectre I currently lives in the familiar grey zone between bold demo culture and independently verified performance. In the post-Theranos era, that gap matters.

• Experts are cautious: some researchers think the claims may be overstated or insufficiently explained.

• Technical proof remains limited: the company has promised more verification but has not fully opened the black box.

• Trust is now a product issue: in privacy tech, credibility matters as much as design or marketing.

There is real science behind parts of the concept

Even so, the basic idea is not pure science fiction. Researchers at the University of Chicago’s SAND Lab previously developed a wearable microphone jammer designed to disable nearby microphones, including hidden ones. Their work showed that microphone jamming can be effective, while also highlighting practical limitations such as blind spots and the difficulty of maintaining consistent disruption from a fixed position. That does not validate Spectre I’s specific claims, but it does show that anti-recording systems are technically plausible and already grounded in published research. The real question is whether Deveillance has genuinely improved on these earlier approaches in a reliable, scalable, consumer-ready form.

• The concept has precedent: microphone jamming is already an active area of research.

• Limitations are well known: blind spots and positioning challenges have long affected this kind of technology.

• The open question is execution: Spectre I may be a leap forward, or simply a polished repackaging of an existing idea.

Why are people ready to believe

Part of what makes this story so potent is that it taps into a real collapse of trust. Pew Research reported in 2023 that 71% of Americans had little to no trust that leaders of social media companies would be held accountable by the government if they misused or compromised personal data. That mistrust forms the emotional infrastructure behind products like Spectre I. Even if the device is imperfect, the desire it speaks to is real: people want ways to become less legible to platforms, less searchable to systems, and less extractable to companies whose data appetites feel effectively limitless.

• Trust is low: many people do not believe large technology firms will be meaningfully restrained.

• Privacy is becoming physical again: it is no longer just about browser settings and encryption, but about what happens within earshot.

• Anti-surveillance may become aspirational: privacy tools are starting to look like premium lifestyle products as well as defensive utilities.

Why It Matters

This story matters because it rounds out the pattern emerging across this week’s Digest. In Section 1, we saw physical infrastructure bending towards machine intelligence. In Section 2, we saw social and local computing environments being reconfigured for autonomous agents. Here, in Section 3, we see the counter-movement beginning to take shape: humans looking for ways to reclaim zones of opacity inside an increasingly sensor-rich world. Spectre I may or may not become the breakthrough product its supporters hope for, but its viral resonance shows that privacy is no longer a background policy concern. It is becoming a frontline design challenge of the AI era.

Closing Thoughts

There is something almost poetic in the rise of a device built to protect us from our other devices. Whether Spectre I proves effective, flawed, or simply premature, it captures a turning point in public consciousness. We are entering a world in which intelligence is ambient, devices are increasingly agentic, and recording can happen quietly, continuously and almost invisibly. In that world, privacy may no longer be preserved by law or etiquette alone. It may need to be engineered, purchased, and actively defended. And that, perhaps more than the orb itself, is the real signal worth watching.

A Dose of Uploaded Minds for Your Weekend | Because the boundary between biology and simulation is starting to blur

If this week’s Digest has traced how intelligence is reshaping infrastructure, social systems and privacy, this final recommendation points towards an even deeper frontier: the possibility of emulating living minds themselves. Eon Systems has drawn major attention after unveiling what it describes as the first “multi-behaviour brain upload”, a digital fruit fly brain connected to a simulated body that can move through a virtual environment using its neural wiring alone. It is early, imperfect, and far from human mind uploading, but it offers a striking glimpse of where connectomics, simulation and embodied intelligence may be heading.

• A fruit fly brain was digitally embodied: Eon connected a neural emulation to a virtual body rather than treating the brain as an isolated static model.

• Behaviour emerged from wiring: the project’s core claim is that actions like walking and foraging arise from the connectome itself, without conventional AI training.

• The larger signal is profound: whole-brain emulation is shifting from abstract theory towards experimental engineering.

Fly Brain Upload Overview

Eon Systems recently shared a virtual embodied fly built from the fruit fly connectome, using a neuron-by-neuron model linked to a simulated body. In its own technical write-up, the company describes this as an early step towards showing how an embodied brain can control a virtual organism, rather than merely serving as a diagram or a dataset. The significance here is not that a fly has been perfectly “uploaded”, but that structure alone appears capable of generating a recognisable behavioural repertoire inside a synthetic environment.

• Embodied, not abstract: the model is connected to a simulated body that moves in a physics-based environment.

• A proof of principle: it suggests that connectomic structure can drive behaviour in silico.

• A conceptual leap: this moves whole-brain emulation closer to lived behaviour rather than static mapping.

Technical Background

The project builds on the rapid progress of fly connectomics over recent years, especially the detailed mapping of the adult Drosophila melanogaster brain. Eon’s post explains that the virtual fly uses a simplified neuron model and is integrated with a physics engine to let sensory inputs drive neural activity and produce movement. In parallel, tools such as Virtual Fly Brain continue to provide researchers with rich anatomical, connectivity and gene-expression resources for exploring the fly nervous system, helping establish the broader scientific context that makes experiments like this possible.

• Connectomics made this possible: the work depends on highly detailed brain-mapping datasets.

• Simulation closes the loop: neural activity is translated into embodied action in a virtual world.

• Virtual Fly Brain remains foundational: it is a core research tool for exploring fly neuroanatomy and connectivity, not a mind-upload platform in itself.

Process and Limitations

This is not a complete digital organism, and Eon is explicit that the work remains a first step. The company notes that the embodied fly is still a work in progress, while outside observers have also pointed out obvious simplifications, including the use of relatively basic neuron models and the absence of many layers of biological complexity. That makes the project exciting, but also easy to overstate if framed carelessly as a full “mind upload”. The smarter interpretation is that it is a promising emulation milestone with significant limitations still in view.

• Still early-stage: Eon itself describes the work as preliminary.

• Biological fidelity is limited: simplified neurons and incomplete modelling mean this is not a perfect replica of a living fly mind.

• The term “upload” needs caution: the achievement is real, but the language can easily outrun the science.

Future Implications

The reason this matters beyond neuroscience is scale. If increasingly complete connectomes can be turned into behaving virtual organisms, then whole-brain emulation begins to look less like speculative fiction and more like a long-horizon engineering path. Eon has signalled ambitions to scale beyond flies, which would dramatically increase both the scientific challenge and the philosophical stakes. At that point, questions around consciousness, identity, simulation, and the meaning of embodiment stop being distant abstractions and start becoming design problems.

• Scaling is the next frontier: moving from flies to larger brains changes everything.

• The implications are philosophical as well as technical: emulating behaviour raises deeper questions about mind and selfhood.

• This is a threshold technology: modest today, potentially civilisation-shaping tomorrow.

Why It Matters

This is one of those stories that feels small on the surface and enormous underneath. A simulated fruit fly may sound niche, but the underlying direction points towards a future in which biological intelligence can be studied, reproduced, and perhaps eventually instantiated in synthetic environments with increasing fidelity. In the context of this week’s Digest, it extends the arc beautifully: first, we saw the world being rebuilt for machine intelligence, then agents inhabiting social and local systems, and now the possibility that minds themselves may become portable architectures.

Closing Thoughts

The uploaded fly is not the singularity, and it is not a digitally resurrected creature in any mystical sense. But it is a signal. It shows how quickly the boundary between mapped brain, simulated body and emergent behaviour is beginning to soften. That alone makes it worth your weekend attention. Sometimes the future does not arrive as a grand declaration. Sometimes it arrives as a tiny insect, moving inside a machine.

One More Thing… | When Sport Becomes a Spatial Medium

If this week’s Digest has been about intelligence escaping the screen and reorganising the world around us, then sport may be one of the most exciting places where that shift becomes visible to everyone. The future of sports media is no longer just sharper cameras or better commentary. It is about turning live competition into a navigable, data-rich, three-dimensional experience. From millimetre-accurate player tracking and AI-generated insights to volumetric replays and wraparound shared-reality venues, the broadcast itself is starting to evolve into something far more immersive, customisable and alive. The real leap is not simply better viewing. It is a new relationship between fan, game and space.

• Sport is becoming spatial: matches are increasingly captured as 3D data environments rather than flat video feeds.

• AI is the invisible engine: real-time analysis is turning raw tracking data into personalised insight and immersive visuals.

• The next broadcast may be explorable: fans are moving from passive viewers to participants choosing angles, overlays and perspectives.

The foundation is already here

The shift to immersive sports begins with capture. Systems such as Sony’s Hawk-Eye already track players and balls with millimetre-level precision using synchronised machine-vision cameras positioned around stadiums. Sony says the system can capture the movement of balls and players throughout the venue with millimetre accuracy, while the NFL’s Next Gen Stats stack adds another layer through RFID and ultra-wideband infrastructure. According to the NFL, players wear two to three RFID tags in their shoulder pads, while tags are also placed on officials, pylons, chains and the ball itself, all interacting with 20 to 30 receivers in the stadium. In effect, the field is becoming a dense sensing environment where every movement can be spatially mapped in real time.

Read more

MyGeekSpace | Living the Future, grounded in the Terra 2.0 vision, follows meaningful signals emerging beneath the daily noise of headlines and hype. This is not a space for prediction. It is a space for orientation, recognising direction while the future is still taking shape.

This week’s edition follows a single unfolding narrative: intelligence is no longer appearing only as software, models, or isolated breakthroughs. It is moving into the operating layers of civilisation itself, becoming embedded in systems that organise work, defence, energy, perception, and knowledge.

This week’s key signals:

• OpenAI expands from model building into workplace infrastructure, releasing GPT-5.4 while reportedly moving toward developer platforms that challenge established software ecosystems.

• Frontier AI rivalry intensifies as OpenAI’s defence agreements trigger public criticism from Anthropic, while NATO discussions signal a widening military role for commercial AI.

• BYD compresses electric charging time to near-refuelling speed, pushing mobility closer to a new practical threshold.

• Science Corporation advances retinal implants that restore functional sight, bringing neurotechnology into direct human application.

• NotebookLM transforms research into cinematic video, suggesting that knowledge itself is becoming multi-format and increasingly machine-shaped.

Together, these developments point toward a deeper civilisational transition already underway: the shift from the Information Age to the Intelligence Age.

Power up your curiosity and settle into this week’s signals.

→ If this weekly exploration resonates, consider supporting MyGeekSpace on Substack with a free or paid subscription. Your support sustains independent, future-focused writing and helps keep this signal alive.

Section 1 | OpenAI Moves from Models to Infrastructure

In this first signal of the week, the story is no longer just about a stronger model release. It is about platform gravity. OpenAI has introduced GPT-5.4 while simultaneously signalling a broader ambition: to control not only how intelligence is generated, but where professional work and software creation increasingly take place.

GPT-5.4 arrives as OpenAI’s most workplace-native model yet

The release of GPT-5.4 marks a clear shift in emphasis from raw capability to operational usefulness. Built in three variants, standard, Thinking, and Pro, the model has been positioned as a system designed for sustained professional output rather than short conversational interaction.

• Native computer-use capability means GPT-5.4 can now work across multiple applications more fluidly, allowing it to interact with digital environments in ways closer to a junior operator than a conventional assistant.

• Presentation generation has been noticeably improved, with stronger layout logic, visual coherence, and better integration of image tools, targeting business workflows directly.

• The model achieved 83 per cent on OpenAI’s GDPval benchmark for knowledge work, while also setting new records on OSWorld-Verified and WebArena Verified benchmarks associated with digital task execution and interface navigation.

• In law and finance testing, GPT-5.4 also led Mercor’s APEX-Agents benchmark, reinforcing its growing role in high-value analytical work.

• The API now supports one million tokens, the largest context window OpenAI has offered so far, making it more suitable for long reports, full legal archives, research corpora, and large-scale internal enterprise use.

• OpenAI also introduced Tool Search, allowing models to retrieve tool definitions only when needed, reducing developer cost and improving efficiency in larger tool ecosystems.

Compared with GPT-5.2, OpenAI claims the new model is 33 per cent less likely to make individual factual errors and 18 per cent less likely to produce responses containing any error at all. GPT-5.4

Revenue growth confirms that enterprise AI is now a strategic race

The timing of GPT-5.4’s release matters because it arrives during a period of accelerating financial momentum for OpenAI.

• Annualised revenue has now surpassed $25 billion, up from $21.4 billion at the end of 2025.

• Internal projections reportedly point toward more than $280 billion in revenue by 2030.

• Discussions around a future IPO continue, with valuation expectations reaching as high as $1 trillion in some projections.

• Competition with Anthropic remains intense, especially in enterprise productivity, where model quality increasingly influences corporate software decisions.

• Microsoft has already integrated Anthropic models into Copilot environments after internal comparisons reportedly showed stronger spreadsheet and presentation performance from Claude in some tasks.

GPT-5.4 can therefore be read as OpenAI’s answer to a growing reality: frontier AI is now judged not only by intelligence, but by usefulness inside daily economic systems.

OpenAI’s GitHub move suggests a deeper strategic realignment

At the same time, reports that OpenAI is building its own code repository platform suggest the company is now targeting developer infrastructure directly.

• The proposed platform would compete with GitHub, moving OpenAI beyond coding assistance into the layer where software itself is stored and coordinated.

• This follows the failed acquisition attempt of Windsurf in 2025, where tensions with Microsoft reportedly played a decisive role.

• OpenAI’s relaunched Codex desktop environment has already crossed one million developers and is growing rapidly. Codex

• Surveys now suggest Codex usage is already approaching a significant share of competitors, such as Cursor.

• If OpenAI controls both the coding assistant and the repository environment, it begins to shape the full development loop, from prompt to deployment.

This is where the relationship with Microsoft becomes structurally more complex. OpenAI remains financially intertwined with its largest backer while increasingly moving into territory once considered core Microsoft infrastructure.

Why It Matters

The deeper signal here is that OpenAI is no longer behaving like a model lab alone. It is becoming an operating layer for professional cognition. GPT-5.4’s native computer-use capability suggests the boundary between assistant and digital co-worker is thinning quickly, while the GitHub challenge reveals a company increasingly interested in owning the workflow itself.

For developers, analysts, consultants, legal teams, and creators, this means AI is no longer merely helping complete tasks. It is beginning to sit inside the environments where those tasks originate, evolve, and are delivered.

Closing Thoughts

A year ago, model launches were mostly about benchmark races. Today, they increasingly reveal strategic architecture.

GPT-5.4 is not simply another upgrade. It represents a broader move toward AI as a workplace substrate, where intelligence is embedded into documents, systems, software repositories, and eventually institutional decision layers.

The larger question is no longer which model is strongest. It is the question of which company becomes the invisible interface through which modern work itself is conducted.

Section 2 | AI, Defence, and the Fracturing Frontier Alliance

If Section 1 revealed the commercial race for AI infrastructure, the next signal moves into more contested territory: geopolitics and defence. In recent days, the rivalry between OpenAI and Anthropic has spilt into public view, triggered by OpenAI’s agreement with the U.S. military and a leaked memo from Anthropic CEO Dario Amodei sharply criticising the deal.

At the same time, new reports suggest OpenAI is now exploring deployments across NATO networks, a move that could extend AI’s presence deeper into Western defence infrastructure.

Amodei’s memo ignites a public rift between frontier AI labs

The immediate controversy began with an internal memo sent by Amodei to Anthropic staff that quickly leaked into the public domain. In it, the CEO described OpenAI’s military agreement as largely performative.

• Amodei characterised the safeguards surrounding OpenAI’s Pentagon contract as “maybe 20 per cent real and 80 per cent safety theatre,” accusing the company of prioritising optics over genuine constraints.

• He argued that Anthropic walked away from negotiations because the U.S. Department of Defense would not provide explicit written assurances prohibiting the use of AI for domestic mass surveillance or autonomous lethal weapons.

• In the memo, Amodei also accused OpenAI CEO Sam Altman of “gaslighting” the public about the nature of the deal and portraying himself as a mediator while accepting terms Anthropic rejected on ethical grounds.

• The tension intensified after the Pentagon labelled Anthropic a “supply chain risk”, effectively cutting the company out of certain defence contracts.

• Shortly afterwards, OpenAI announced its own agreement to deploy AI within U.S. military classified networks, stepping directly into the space Anthropic had declined.

The episode has transformed what was once a largely technical competition between frontier labs into a public ideological divide over how AI should interact with state power.

NATO discussions point to a wider military AI ecosystem

The Pentagon agreement may only be the beginning of OpenAI’s ambitions in defence infrastructure.

• Reports indicate OpenAI is exploring a potential contract to deploy its systems across NATO networks used by allied militaries.

• Initial discussions suggested deployments across classified environments, although later clarification indicated the opportunity may focus on NATO’s unclassified operational networks.

• Such integration would represent a major expansion of commercial AI into allied defence systems across a 32-member military alliance.

• The move could influence procurement standards, interoperability between allied systems, and the wider adoption of AI tools across Western military infrastructure.

• It also raises questions about how rapidly commercial AI companies are becoming embedded within national security architectures once dominated by traditional defence contractors.

The implications are profound: AI models originally designed for research and productivity are now being integrated into the operational backbone of defence organisations.

Internal backlash highlights growing tension inside AI companies

The controversy has not only played out between companies. It is also unfolding inside them.

• More than 100 OpenAI employees reportedly signed a letter urging leadership to refuse any agreement that would allow AI to be used for domestic mass surveillance or autonomous weapons without human oversight.

• During a tense internal meeting, Altman acknowledged the backlash from staff and the wider AI community, describing the reaction to the Pentagon deal as “really painful”.

• He emphasised that operational decisions ultimately belong to government authorities, not AI companies themselves.

• In response to criticism, OpenAI later clarified elements of the agreement, including restrictions on access by certain intelligence agencies without additional approvals.

At stake is a deeper question that the entire AI industry will soon face: where the boundary lies between technological innovation and military deployment.

Key Signals

• The frontier AI rivalry is no longer just technical; it is now political, ethical, and institutional.

• Defence partnerships may become one of the largest revenue streams for advanced AI systems.

• Governments are rapidly integrating commercial AI tools into national security infrastructure.

• Internal dissent inside AI companies suggests the cultural impact of these decisions may be as significant as the technical ones.

Why It Matters

The integration of frontier AI into defence systems represents one of the most consequential shifts in the technology sector.

For decades, the most advanced computing systems in the world were built directly by governments. Today, the opposite is increasingly true. The most powerful AI models are developed by private companies, and states are racing to integrate them into military, intelligence, and strategic decision systems.

This creates a new relationship between Silicon Valley and national security, one that will shape the geopolitical landscape of the AI age.

Closing Thoughts

What we are witnessing is the early formation of an AI defence ecosystem.

Companies such as OpenAI and Anthropic were once primarily research labs. Now they sit at the intersection of technology, ethics, and geopolitics. Their decisions about partnerships, safeguards, and deployment boundaries will influence not just product strategy but the architecture of power itself.

The question is no longer whether AI will play a role in defence systems. It already does. The real question is who gets to define the rules that govern its use.

Section 3 | BYD Pushes EV Charging Toward the Refuelling Threshold

If the previous sections showed intelligence becoming infrastructure, this week’s mobility signal points to another kind of threshold: energy delivery. BYD has unveiled a second-generation Blade Battery that dramatically compresses charging time, bringing electric mobility closer to the psychological convenience long associated with petrol refuelling.

A nine-minute charge changes the consumer equation

Presented in Shenzhen by BYD chairman Wang Chuanfu, the new battery is designed not simply as an incremental battery upgrade, but as an answer to one of the most persistent objections to electric vehicle adoption: waiting time.

• Under standard conditions, the battery charges from 10 per cent to 70 per cent in five minutes.

• It reaches 97 per cent in nine minutes, a charging speed now approaching real-world fuelling behaviour rather than conventional EV expectations.

• BYD deliberately caps charging at 97 per cent during ultra-fast sessions, reserving the remaining 3 per cent for regenerative braking efficiency and overall energy optimisation.

• This design suggests that battery systems are increasingly being engineered as intelligent energy ecosystems rather than simple storage units.

The symbolic importance is clear: once charging drops below ten minutes, the consumer narrative around inconvenience begins to weaken significantly.

Cold-weather charging may remove one of EV adoption’s last practical barriers

One of the most striking aspects of the announcement was the battery’s performance in extreme cold, historically one of the most difficult environments for electric vehicles.

• After 24 hours frozen at minus 30°C, the battery still charges from 20 per cent to 97 per cent in 12 minutes.

• That is only three minutes slower than charging at room temperature.

• BYD says this effectively resolves the long-standing operational difficulty faced in colder regions of northern China and other winter-heavy markets.

• Thermal resilience at this level could become highly relevant for northern Europe, Canada, and parts of the United States, where winter battery degradation remains a major hesitation point for buyers.

This matters because EV adoption has often moved fastest in climates where battery performance is naturally easier to manage. BYD is now directly targeting geographies where electrification has remained more psychologically uncertain.

Longer range and infrastructure scale reinforce BYD’s wider strategy

The battery launch also arrives as part of a broader systems play: battery chemistry, vehicle range, and charging infrastructure advancing together.

• The new 150-kWh battery pack enables the Yangwang U7 to reach 1,006 kilometres under China’s CLTC testing standard.

• The fully electric Denza Z9GT reaches up to 1,036 kilometres under the same measure.

• Energy density has increased by more than 5 per cent over the previous Blade Battery generation.

• BYD claims up to 3,000 charge cycles, extending long-term battery viability.

• Safety remains central to the pitch, with successful nail penetration and bottom-impact testing after repeated ultra-fast charging cycles.

• The company also plans to deploy 15,000 ultra-fast charging stations across China this year, supported by 1,500-kW chargers and peak station outputs of 2,100 kW.

This suggests BYD is not merely improving cars, but building a vertically integrated charging civilisation around them.

Key Signals

• EV charging speed is moving into a new category where behavioural resistance may fall sharply.

• Battery resilience in extreme cold expands viability across previously difficult markets.

• Chinese EV firms continue to lead in battery-to-infrastructure integration at an industrial scale.

• Ultra-fast charging networks are becoming as strategically important as battery chemistry itself.

Why It Matters

The EV race is no longer only about range. It is about friction.

For years, electric mobility’s central weakness was not capability but interruption: how often drivers needed to stop, how long they had to wait, and how uncertain performance felt in difficult climates. Once those frictions begin to collapse, adoption moves from ideological choice to practical default.

This is especially significant because BYD’s advances are arriving while much of Europe still debates charging infrastructure speed and grid readiness.

Closing Thoughts

A battery that charges to 97 per cent in nine minutes is not just a technical headline. It changes how electric vehicles are imagined in daily life.

As battery systems become faster, safer, and more climate-resilient, the broader transition away from combustion begins to feel less like a future aspiration and more like a systems inevitability.

The deeper pattern this week is unmistakable: whether in AI, defence, or mobility, infrastructure is accelerating faster than public perception.

A Dose of Restored Vision for Your Weekend | Because the future is increasingly learning how to repair perception itself

Not every frontier breakthrough arrives through larger models or faster machines. Some arrive at the level of human experience itself, where technology begins restoring capacities once thought permanently lost. This week’s signal comes from Science Corporation, the neurotechnology venture founded by former Neuralink president Max Hodak, which has raised $230 million to accelerate one of the most compelling developments in applied neurotechnology: a retinal implant designed to restore functional sight.

A retinal chip that turns light into usable vision

At the centre of Science’s progress is PRIMA, a retinal implant built to help patients affected by advanced geographic atrophy, a severe form of age-related macular degeneration.

• PRIMA is a wireless chip measuring just 2 mm by 2 mm, roughly the size of a pinhead.

• It is implanted beneath the retina and works together with specialised glasses equipped with a forward-facing camera.

• The glasses capture visual information and project infrared light patterns directly onto the implant.

• The implant converts those signals into electrical stimulation for surviving retinal cells, allowing the brain to reconstruct central visual information.

• Rather than replacing the eye, the system works with the remaining biological structure, creating a hybrid form of artificial vision.

This is an important distinction: the technology is not merely prosthetic, it is cooperative, using biological pathways that still remain active.

Clinical results suggest meaningful restoration is already possible

The strongest signal comes from clinical outcomes rather than laboratory promise.

• Results published in The New England Journal of Medicine showed that more than 80 per cent of 38 patients across 17 sites in five countries experienced measurable visual improvement.

• Many participants regained the ability to identify letters, numbers, and short words.

• The trial focused on patients who had lost central vision through geographic atrophy, one of the most difficult visual impairments to reverse.

• According to lead investigator Frank Holz, the study marks the first time functional central vision has been restored in this patient group through such a system.

These are still early stages, but they indicate that neuroprosthetic vision is moving from concept into practical medicine.

From retinal repair to broader neural interfaces

Science is not positioning PRIMA as a single product, but as the first step in a broader neurotechnology platform.

• The company has now raised roughly $489 million in total funding.

• Its valuation has reached $1.25 billion, making it one of the most highly valued brain-interface companies after Neuralink.

• A CE mark application has already been submitted in Europe, with regulatory approval expected to be decided this year.

• In the United States, PRIMA already holds FDA Breakthrough Device status.

• Science is also developing higher-resolution future versions, including colour vision upgrades.

• Hodak has suggested that one or two future generations could approach vision quality near 20/20 under certain conditions.

This means retinal neurotechnology may become one of the first commercially available examples of human sensory augmentation reaching real healthcare markets.

Key Signals

• Neurotechnology is increasingly shifting from speculative brain-interface narratives into medical deployment.

• Vision restoration may become one of the first widely adopted commercial neuroprosthetic applications.

• Hybrid systems that combine wearables, implants, and biological processing are proving highly effective.

• The race around neural interfaces is broadening well beyond one company or one founder.

Why It Matters

Sight is one of the most emotionally and socially significant human capacities. A technology that restores even partial visual clarity changes daily life at a profound level, from independence and mobility to confidence and identity.

What makes this especially important is that retinal implants represent a gentler public entry point into brain-interface technologies. Before direct cognitive enhancement becomes common, society is likely to embrace systems that restore lost human function.

In that sense, devices like PRIMA may quietly become the first widely accepted bridge between medicine and deeper neural augmentation.

Closing Thoughts

Much of the conversation around neurotechnology still focuses on distant ideas: thought-controlled computing, digital telepathy, or cognitive upgrades.

Yet the most meaningful threshold may arrive first through something simpler and more human, the ability to help someone read again, recognise a face again, or recover a lost visual centre.

Sometimes the future does not announce itself through spectacle. Sometimes it begins when a blurred world becomes visible again.

One More Thing… | When Research Starts Directing Its Own Film

This week’s final signal deserves a slightly different treatment because it feels less like a feature update and more like a glimpse of how knowledge work itself is changing. Google is quietly reshaping the relationship between research, synthesis, and creative output through the rapid evolution of NotebookLM.

What began as an experimental notebook for organising sources has now moved into something far more ambitious: turning raw research into a cinematic narrative.

Read more

MyGeekSpace | Living the Future, grounded in the Terra 2.0 vision, tracks meaningful signals emerging beneath the daily noise of headlines and hype. This is not a space for prediction. It is a space for orientation, recognising direction while the future is still forming.

This week’s edition follows a single unfolding narrative: intelligence is no longer confined to software models or research laboratories. It is scaling outward into infrastructure, organising itself collectively, embedding into physical experience, and beginning to reshape civilisation in real time.

Across biology, mobility, geopolitics and physics itself, we are witnessing intelligence migrate from abstraction into reality.

This week’s key signals:

• Artificial Biological Intelligence moves evolution from observation to authorship

• Autonomous mobility transforms cities into living laboratories for embodied AI

• AI guardrails become matters of national security and geopolitical power

• Photonic computing suggests intelligence may emerge directly from physical systems

Together, these developments point toward a deeper civilisational transition already underway: the shift from the Information Age to the Intelligence Age.

Power up your curiosity and settle into this week’s signals.

Welcome to Living the Future.

→ If this weekly exploration resonates, consider supporting MyGeekSpace on Substack with a free or paid subscription. Your support sustains independent, future-focused writing and helps keep this signal alive.

Section 1 | Artificial Biological Intelligence and the Future of Species: A Strategic Briefing

Artificial Intelligence conversations usually focus on AGI and superintelligence. Machines are thinking faster than humans. Systems surpassing cognition.

But a quieter frontier is emerging alongside digital AI.

Artificial Biological Intelligence (ABI) moves intelligence from software into life itself.

Rather than building smarter machines, humanity is beginning to engineer living systems. Biology is shifting from observation to authorship.

We are moving from evolving naturally to potentially designing evolution intentionally.

Key Signals

• AGI expands intelligence in software

• ASI accelerates discovery and problem-solving

• ABI introduces intelligence into biology itself

• DNA is increasingly treated as programmable code

• Humanity shifts from a product of evolution to a participant in it

From Artificial Intelligence to Artificial Life

For billions of years, evolution operated blindly through mutation and natural selection.

ABI changes the rules.

AI models can now interpret genetic language, predict biological behaviour, and assist in designing entirely new organisms.

Evolution begins to resemble an engineering process rather than a purely natural one.

Summary

• Biology transitions from science to engineering

• AI becomes a translator of DNA’s hidden grammar

• Life design begins to follow a technological workflow

The ABI Stack: How Designed Life Emerges

Artificial Biological Intelligence develops through four emerging capabilities:

Recognition
Understanding what DNA sequences actually do.

Generation
Designing new genomes rather than editing existing ones.

Construction
Synthesising DNA at an industrial scale.

Booting Up
Activating synthetic genomes inside living cells.

Summary

• Biology becomes programmable infrastructure

• AI operates as an evolutionary partner

• Designed organisms move from theory to practice

Beyond the Tree of Life

Traditional evolution produced only the organisms that survived.

ABI opens access to a vastly larger possibility space, sometimes described as a “Forest of Life”: biological forms that never existed, but could.

Potential applications include:

• Climate-resilient crops

• Pollution-eating microbes

• Biological computing systems

• Novel medical therapies

Summary

• Evolution expands from history into exploration

• Biodiversity becomes informational infrastructure

• Conservation grows more important, not less

The Governance Challenge

With biological design comes unprecedented responsibility.

The technological capability to redesign life is advancing faster than ethical frameworks or global governance.

The central question shifts from Can we create new life? to Who decides what life should become?

Summary

• Biological power accelerates faster than regulation

• Governance becomes a civilisation-level challenge

• Wisdom, not capability, becomes the limiting factor

Why It Matters

The Intelligence Age is not only digital.

If AGI reshapes knowledge and ASI reshapes civilisation, ABI reshapes life itself.

Food production, medicine, environmental restoration, and even human biology may soon become design spaces rather than evolutionary constraints.

Intelligence is no longer only about learning about the world.
It is beginning to redesign the living systems that sustain it.

Closing Thoughts

We often imagine the future arriving through machines that think.

Artificial Biological Intelligence suggests something deeper:

The future may arrive through organisms that are thought into existence.

Humanity is entering a new role, not merely evolved beings, but co-authors of evolution itself.

The challenge ahead is not technological progress.

It is choosing wisely what kind of life we decide to grow.

Companion Video

Adrian Woolfson on Authoring Life with Artificial Biological Intelligence
FUTURES Podcast | Bonus episode recorded live at the World Government Summit 2026.

Section 2 | The Robotaxi Race Comes to London

London is rapidly becoming one of the most important testing grounds for autonomous mobility. This week, UK startup Wayve announced a massive $1.5 billion funding round, signalling that the age of AI drivers is moving from research labs to real city streets.

The announcement positions Britain not just as a participant in the autonomy race, but potentially as one of its global launch platforms.

Wayve’s Big Bet on AI-First Driving

Unlike traditional self-driving companies that rely on high-definition maps and expensive sensor stacks, Wayve has taken a radically different approach.

Instead of programming vehicles with rigid rules, the company trains AI systems to learn how to drive through experience, much like a human driver.

Key Signals

• Raised $1.5 billion Series D, valuing Wayve at $8.6 billion

• Total funding now reaches $2.8 billion

• Major investors include Microsoft, Nvidia, Uber, SoftBank, Mercedes-Benz, Nissan and Stellantis

• Focus on end-to-end deep learning, removing dependence on mapped environments

• Goal: deploy a scalable AI driving platform across global vehicle fleets

Wayve’s thesis is simple but powerful: autonomy will not scale city by city; it will scale software-first.

London Becomes a Living Laboratory

The funding enables two parallel commercial launches that could reshape urban mobility.

Robotaxi Trials

• Supervised robotaxi testing expected in London during 2026

• Partnership with Uber targeting expansion across multiple global cities

• The UK government accelerated regulatory timelines for autonomous vehicle pilots

London’s chaotic traffic, narrow streets and unpredictable driving culture make it an ideal stress test for AI systems.

If autonomy works here, it can work almost anywhere.

Consumer Vehicles Next

• Wayve AI Driver entering mass-produced cars beginning in 2027

• Integration starting with Nissan’s next-generation ProPilot system

• Expanded collaborations with Mercedes-Benz and Stellantis are underway

This signals a shift from experimental robotaxis toward everyday autonomous driving embedded directly into consumer vehicles.

Tesla Enters the Same Battlefield

Wayve’s announcement lands alongside another major signal: Tesla’s Robotaxi programme is advancing toward production.

Tesla’s Cybercab, a purpose-built autonomous vehicle with no steering wheel or pedals, has already been previewed in London and represents a different vision of autonomy.

Key Signals from Tesla’s Robotaxi Strategy

• Steering-wheel-free vehicle designed purely for ride-sharing

• Production targeted before 2027

• Users summon vehicles via the Tesla app

• Wireless charging and minimalist interior design

• Target price below £30,000 to enable mass adoption

• UK demonstrations already underway

Where Wayve aims to license intelligence to automakers, Tesla is building a vertically integrated autonomous transport network.

Two models of the future are emerging simultaneously: AI driver as a platform versus robotaxi as an ecosystem.

A Global Autonomy Race Accelerates

London is no longer competing only with Silicon Valley.

Multiple global players are converging on the same streets:

• Waymo is preparing UK deployments

• Baidu is exploring European robotaxi partnerships

• Uber is positioning itself as a universal autonomy marketplace

• Governments are investing heavily in regulatory frameworks and infrastructure

The UK alone has committed £1.1 billion toward autonomous vehicle development, signalling national strategic intent.

Autonomy is no longer hypothetical. It is becoming infrastructure.

Why It Matters

The autonomous vehicle transition is not merely about cars driving themselves. It represents the arrival of embodied AI operating in the physical world.

Transportation is one of the largest economic systems on Earth. When driving becomes software, mobility shifts from ownership to intelligence.

Cities may gain safer roads, reduced congestion, and new access models connecting underserved areas. At the same time, autonomy forces societies to rethink regulation, labour, liability and urban design.

What we are witnessing is the moment AI leaves the screen and begins navigating reality itself.

Closing Thoughts

For years, self-driving promises felt perpetually five years away. That timeline is now collapsing.

Wayve, Tesla and their competitors are converging on the same outcome: mobility becoming an intelligent service rather than a mechanical activity.

London’s streets may soon host one of the first true encounters between human civilisation and autonomous intelligence at scale.

The question is no longer if robotaxis arrive.

It is how quickly we adapt once they do.

Section 3 | Pentagon hits Anthropic with ultimatum over guardrails

Artificial intelligence is no longer only a commercial or scientific race. This week, it became unmistakably clear that AI is now a national security technology.

Reports indicate that the US Pentagon delivered an ultimatum to Anthropic CEO Dario Amodei: loosen Claude’s military guardrails or risk losing defence contracts and facing potential government intervention.

The confrontation reveals a new phase of the Intelligence Age, where frontier AI labs must navigate not only innovation and safety, but geopolitical pressure.

The Guardrail Conflict

At the centre of the dispute is a fundamental disagreement over how AI should be used inside military systems.

Key Signals

• Anthropic refused to allow autonomous weapons deployment without human oversight

• The company also rejected participation in large-scale domestic surveillance

• The Pentagon reportedly offered three options: compliance, contract termination, or compelled cooperation via the Defence Production Act

• A $200 million defence contract now hangs in the balance

• AI models are rapidly becoming embedded inside classified military networks

This is not merely a contractual disagreement. It represents a collision between AI safety philosophy and state power.

AI Becomes Strategic Infrastructure

Claude was among the first advanced models deployed inside sensitive government environments, signalling how quickly AI has moved from research assistant to operational asset.

At the same time, competitors are stepping forward.

Shifting Landscape

• xAI’s Grok secured defence access after agreeing to broader lawful-use provisions

• OpenAI and Google are reportedly being accelerated toward classified deployment

• Governments increasingly view AI capability as comparable to nuclear, cyber, or space dominance

Military institutions are no longer asking whether to adopt AI. They are deciding how much control they require over it.

The Emerging Ethical Fault Line

For AI companies, the dilemma is unprecedented.

Frontier labs were founded with safety principles intended to prevent misuse. Yet national security frameworks operate under entirely different assumptions, prioritising strategic advantage.

Core Tension

• AI labs seek enforceable ethical boundaries

• Governments prioritise deterrence and technological superiority

• Legal pressure can override voluntary safety commitments

• The definition of “responsible AI” becomes politically contested

The risk is not simply militarisation, but the gradual normalisation of removing safeguards during moments labelled as crisis or wartime necessity.

Why It Matters

This moment marks a turning point in the relationship between artificial intelligence and political power.

AI is transitioning from an innovation platform to a sovereign capability. Whoever controls advanced models gains leverage across defence, intelligence, cyber operations and global influence.

If governments can compel AI systems to abandon safety limits, the question shifts from what AI can do to who decides what it is allowed to become.

The Intelligence Age will not be shaped by technology alone, but by the institutions capable of directing it.

Closing Thoughts

Sections 1 and 2 showed intelligence expanding into biology and physical mobility.

Section 3 reveals the next layer: intelligence entering the arena of geopolitics.

Artificial intelligence is no longer neutral infrastructure. It is becoming a strategic actor in global power dynamics.

The defining challenge ahead may not be building more capable AI systems, but ensuring that human judgment, accountability and restraint evolve at the same pace as the intelligence we are creating.

Source: Reuters reporting on Pentagon–Anthropic negotiations

A Dose of Light Intelligence for Your Weekend | Because Intelligence May Be a Property of Physics Itself

Italian researchers find photons spontaneously mimic brain memory in optical circuits

When light begins to remember

A remarkable discovery from Italian researchers suggests that intelligence may not belong exclusively to biology or silicon. Identical photons travelling through optical circuits have been shown to spontaneously behave like a Hopfield Network, a mathematical model describing how the brain stores and retrieves associative memories.

Published in Physical Review Letters, the study demonstrates that light itself can reproduce the dynamics of memory formation.

Instead of treating photons as passive carriers of information, the researchers revealed something far more intriguing: light can act as the computing system itself.

Light as Neuron

Using quantum interference inside photonic chips, scientists enabled overlapping particles of light to encode and retrieve information collectively.

In this system, photons effectively become neurons.

The experiment showed that when information loads remain balanced, quantum coherence allows accurate memory recall. As complexity increases, the system transitions into a disorder phase, similar to cognitive overload in biological brains.

Key Signals

• Photons can replicate associative memory behaviour

• Quantum systems naturally reproduce neural-network dynamics

• Memory limits emerge even in non-biological intelligence

• Light-based computing could drastically reduce energy consumption

• Intelligence may be a property of physics itself, not only biology

From Nobel Physics to Future AI Hardware

The work bridges two Nobel Prize traditions: John Hopfield’s foundational machine learning research and Giorgio Parisi’s studies of complex disordered systems.

The implication is profound. Future AI systems may not rely solely on energy-hungry electronic data centres. Photonic computing could enable radically more efficient intelligence infrastructure at a planetary scale.

Rather than electrons moving through circuits, tomorrow’s computation may be performed by patterns of light.

Why It Matters

Across this week’s Digest, intelligence has appeared in three forms:

• designed life

• autonomous mobility

• geopolitical power

This discovery introduces a fourth layer.

Intelligence may be emerging directly from the physical behaviour of the universe itself.

If photons can store and retrieve memory, computation stops being a technological invention and starts looking like a natural phenomenon waiting to be harnessed.

The Intelligence Age may ultimately be powered not by machines that imitate the brain, but by physics that already knows how to think.

Closing Thoughts

There is something quietly poetic about this signal arriving from Italian laboratories.

Light, the oldest metaphor for knowledge, is beginning to function as memory.

Perhaps the future of AI will not resemble colder machines, but increasingly elegant systems where intelligence flows with minimal energy, closer to nature than industry.

The trajectory becomes clearer: first, intelligence learned to calculate, then to learn, and now it may learn to remember using light itself.

▶ Read the Press Release here.

One More Thing… | The Future Cannot Be Paused

Recommend Reading: The Future Cannot Be Paused - Ben Horowitz on AI, Capital & the Transformation Ahead

Every so often, a conversation appears that does not simply analyse technology, but reframes the historical moment we are living through. Ben Horowitz’s discussion on AI, capital, and geopolitical competition is one of those signals.

Read more

MyGeekSpace | Living the Future, grounded in the Terra 2.0 vision, exists to track meaningful signals emerging beneath the daily noise of headlines and hype. This is not about prediction. It is about recognising direction while the future is still forming.

This week’s edition follows a single evolving narrative: intelligence is no longer confined to software or laboratories. It is scaling into infrastructure, organising itself collectively, embedding into physical experience, and beginning to reshape civilisation in real time.

This week’s key signals:

• The Superintelligence Horizon
  AI timelines accelerate as OpenAI shifts from experimentation toward global infrastructure and personal agents.

• The Rise of Collective Intelligence
  AI systems begin collaborating, debating, and specialising, revealing how future intelligence may actually operate.

• Ambient Intelligence Arrives
  Apple’s AI wearables signal a transition from screen-based computing to continuous perception and contextual awareness.

Together, these developments suggest a deeper transition already underway: the move from the Information Age to the Intelligence Age.

Power up your curiosity and settle into this week’s signals.

Welcome to Living the Future.

→ If this weekly exploration resonates, consider supporting MyGeekSpace on Substack with a free or paid subscription. Your support sustains independent, future-focused writing and helps keep this signal alive.

Subscribe now

Section 1 | The Superintelligence Horizon: OpenAI’s Next Phase Takes Shape

This week’s first signal arrives directly from the centre of the AI acceleration curve.

OpenAI is no longer speaking cautiously about incremental progress. Instead, it is openly discussing timelines for superintelligence, global governance structures, and the rise of personal AI agents.

Taken together, these announcements reveal something important: we are moving from AI as a tool to AI as infrastructure for civilisation itself.

Two developments define this shift:

• Sam Altman’s prediction that early superintelligence could emerge by 2028

• OpenAI is strengthening its long-term strategy around autonomous personal agents through the hiring of OpenClaw creator Peter Steinberger

Let’s unpack what this really signals.

Superintelligence by 2028? OpenAI Moves the Timeline Forward

Speaking at the India AI Impact Summit 2026, OpenAI CEO Sam Altman suggested that within just a few years, more intellectual capacity may exist inside data centres than across all human minds combined.

This was not framed as hype, but as a warning paired with responsibility.

Key Signals

• Superintelligence enters a concrete timeframe
  Altman positioned 2028 as a plausible horizon for early superintelligent systems, shifting the discussion from theory into planning.

• A call for democratic AI
  He argued that concentrating AI power within a single nation or corporation could lead to dangerous outcomes, advocating broad global access instead.

• A proposal for global AI governance
  OpenAI suggested an international coordinating body, similar to the International Atomic Energy Agency, to oversee development and risk response.

• India emerges as a core AI frontier
  With more than 100 million weekly ChatGPT users, India is now OpenAI’s second-largest market and a key node in global AI deployment.

• Infrastructure becomes strategy
  OpenAI and Tata Consultancy Services announced new data centre projects under the Stargate initiative, signalling that compute capacity is now geopolitical infrastructure.

• Acknowledgement of disruption
  Altman openly stated that advanced AI may outperform executives and scientists, reinforcing expectations of major labour transformation.

From Chatbots to Personal Agents: OpenClaw Creator Joins OpenAI

Alongside these macro announcements came a quieter but equally important move: OpenAI hiring Peter Steinberger, creator of the viral open-source agent framework OpenClaw.

OpenClaw, previously known as Clawdbot and Moltbot, became one of GitHub’s fastest-growing AI projects by allowing everyday users to deploy autonomous agents across messaging platforms and real-world workflows.

This hire reveals where OpenAI believes the next competitive frontier lies.

Key Signals

• Agents replace interfaces
  Instead of users prompting AI repeatedly, personal agents will increasingly act, coordinate, and execute tasks autonomously.

• Open-source meets platform scale
  Unlike traditional acqui-hires, OpenClaw remains independent and community-driven, suggesting a hybrid future between proprietary systems and open ecosystems.

• Multi-agent collaboration becomes the norm
  Steinberger’s work focuses on AI systems that cooperate rather than operate alone, hinting at a coming “society of AIs”.

• Personal AI becomes a core product category
  OpenAI appears to be positioning intelligent agents as companions, assistants, and digital operators embedded into daily life.

• Cultural emergence of AI-native communities
  The evolution from Moltbot to OpenClaw echoes early signs of AI-native culture explored in When AI Starts Writing Its Own Myths, where users begin forming identity and narrative around AI systems themselves.

Why It Matters

What we are witnessing is a convergence of timeline, governance, and capability.

OpenAI is simultaneously:

• accelerating technical development towards superintelligence

• expanding global infrastructure

• preparing governance frameworks

• and building the agent layer that will bring AI into everyday human activity

This is not merely technological progress. It is institutional preparation for a world where intelligence becomes abundant and programmable.

The real debate is no longer whether superintelligence will arrive, but who shapes its values, access, and distribution.

Closing Thoughts

The deeper signal beneath this week’s OpenAI news is philosophical.

If intelligence becomes an infrastructure like electricity or the internet, humanity enters a new civilisational phase. The challenge shifts from creating intelligence to learning how to live alongside it.

Altman’s warning about authoritarian trade-offs is telling. The future will not be decided by capability alone, but by governance, culture, and collective choice.

And perhaps most interestingly, while superintelligence dominates headlines, the quieter revolution may be personal agents. Not one giant AI replacing humanity, but billions of smaller intelligences amplifying individual humans.

Superintelligence may arrive in data centres first.
But its true impact will be felt when intelligence becomes personal.

This cultural dimension is something I explored further in my previous exploratory essay, “When AI Starts Writing Its Own Myths”, examining Moltbook, the Church of Molt, and the emergence of AI-native culture:

This sets the stage perfectly for our next signal, where AI systems themselves are beginning to collaborate, debate, and think together.

Section 2 | The Rise of Collective Intelligence: When AIs Start Thinking Together

If Section 1 explored the coming age of superintelligence, this next signal reveals how that intelligence may actually function.

The era of the single, all-knowing chatbot is quietly ending.

Across xAI, Anthropic, and Google, a new paradigm is emerging: AI systems no longer operate alone. They collaborate, specialise, debate, and create together.

We are witnessing the transition from individual models to intelligence ecosystems.

xAI Launches Grok 4.20 Beta and the Peer-Review AI Architecture

xAI’s Grok 4.20 Beta represents a fundamental architectural shift away from the traditional chatbot model.

Instead of relying on one system to generate an answer, Grok deploys a council of specialised agents working simultaneously behind the interface. Each plays a defined role before the response ever reaches the user.

Key Signals

• Multi-agent collaboration replaces single-model reasoning
  Grok coordinates multiple internal agents. Harper verifies facts, Benjamin focuses on logic and coding, while Lucas introduces creative reasoning.

• Built-in peer review reduces hallucinations
  Responses are debated internally, effectively embedding a scientific review process into everyday AI interaction.

• Massive context expansion
  A 2-million-token window allows Grok to analyse entire datasets, long research papers, or complex financial simulations in a single workflow.

• From assistant to digital staff
  Restricted access via SuperGrok and X Premium+ signals a shift toward professional users seeking operational capability rather than casual conversation.

• Rapid iteration becomes the norm
  Following Grok 4.1 by only three months, weekly learning cycles suggest AI development is entering continuous evolution rather than periodic upgrades.

Anthropic’s Claude Sonnet 4.6: Autonomy Meets Friction

Anthropic’s latest release reveals a parallel trajectory.

Claude Sonnet 4.6 introduces autonomous “computer use”, enabling AI systems to navigate software interfaces, run workflows, and perform enterprise-level tasks independently.

Yet the rollout exposes a growing tension within AI development: capability versus transparency.

Key Signals

• Agentic AI becomes operational
  Claude can now interact with digital environments similarly to a human operator.

• Smarter data ingestion
  Instead of reading raw webpages, the model writes Python scripts to pre-filter information, improving accuracy while reducing wasted tokens.

• Compute intensity rises dramatically
  Advanced autonomy requires enormous processing resources, with testing consuming nearly five times more tokens than earlier versions.

• Enterprise markets lead adoption
  Anthropic’s Bengaluru expansion highlights India’s growing role as a global AI productivity hub.

• Developer trust becomes a new battleground
  Simplified activity logs have sparked concern, revealing that transparency may become as important as raw intelligence.

Google’s Lyria 3: Creativity Becomes Native to Platforms

While xAI and Anthropic push reasoning and autonomy, Google’s move brings AI directly into everyday creativity.

Lyria 3 inside Gemini allows users to generate fully formed songs, lyrics, and cover art from simple prompts, images, or videos.

This matters because it moves generative creativity from niche tools into mainstream infrastructure.

Key Signals

• AI creativity reaches mass adoption
  Music generation becomes a built-in consumer capability rather than a specialised platform.

• End-to-end creation pipelines
  Genre, tempo, vocals, lyrics, and artwork are handled automatically, collapsing the traditional production chain.

• Authenticity tracking enters the mainstream
  SynthID watermarking introduces native provenance verification for AI-generated media.

• Creator workflows transform
  Integration with YouTube Shorts enables instant soundtrack generation for video content.

• Creative literacy shifts
  Millions of users may begin creating music without prior technical knowledge, redefining authorship itself.

Why It Matters

Across these announcements, a deeper pattern emerges.

AI development is no longer about building a smarter model. It is about building systems of intelligence.

We are seeing three complementary evolutions:

• xAI explores collaborative reasoning between specialised agents

• Anthropic advances autonomous digital labour

• Google integrates generative creativity directly into consumer platforms

Together, they point toward a future where intelligence behaves less like a machine and more like an organisation.

In other words, AI is beginning to resemble society itself.

This transition may ultimately prove more transformative than raw intelligence gains. The next breakthrough is not simply smarter AI, but coordinated intelligence operating at scale.

Closing Thoughts

One of the most striking aspects of this week’s developments is how closely they mirror human cognition.

Humans rarely think alone. We reason collectively through teams, disciplines, and dialogue. Now, AI systems are adopting the same structure.

Instead of a singular artificial mind, we are witnessing the birth of distributed cognition.

The implication is profound: the future of AI may not be one superintelligence, but networks of specialised intelligences collaborating continuously across research, creativity, governance, and daily life.

If Section 1 introduced the arrival of superintelligence, Section 2 reveals its operating system.

And this naturally leads us to the next signal, where intelligence begins to move beyond screens and into the physical world itself.

Section 3 | Ambient Intelligence: Apple’s Quiet Entry into the AI Wearables Era

If Sections 1 and 2 explored intelligence scaling in software and platforms, this third signal shows where AI is heading next: off the screen and into everyday reality.

Apple appears to be preparing its most significant strategic pivot since the iPhone, moving artificial intelligence from something we use into something that continuously perceives the world alongside us.

According to Bloomberg, Apple is fast-tracking three camera-equipped AI wearables: smart glasses, a wearable pendant, and camera-enabled AirPods, all designed to give Siri real-time visual awareness through the iPhone ecosystem.

This is not simply a hardware update. It signals the arrival of ambient intelligence.

Apple’s AI Wearables Strategy: Giving Siri Eyes and Ears

Unlike competitors racing toward fully immersive AR displays, Apple’s approach appears deliberately restrained. The early generation of devices focuses not on visual overlays, but perception.

The goal is clear: transform Siri from a voice assistant into a contextual intelligence companion.

Key Signals

• Smart glasses without displays
  Apple’s glasses prioritise cameras and sensing rather than augmented reality visuals, suggesting an AI-first rather than AR-first philosophy.

• The wearable pendant concept
  An always-on camera and microphone act as the iPhone’s “eyes and ears”, continuously feeding environmental awareness into Siri.

• Camera-equipped AirPods
  Low-resolution sensors provide spatial context, enabling translation, navigation assistance, and situational understanding.

• A redesigned Siri as the central operating layer
  All devices depend on a next-generation Siri expected to adopt a conversational, chatbot-style interface powered by Google’s Gemini.

• Hardware as an AI interface shift
  Apple is repositioning wearables as sensors for intelligence rather than standalone gadgets.

The Competitive Landscape: Apple Enters the Physical AI Race

Apple’s move arrives amid a rapidly forming AI hardware ecosystem.

Meta has already pushed smart glasses into the consumer mainstream through Ray-Ban partnerships, while startups experiment with AI pins and wearable assistants. Apple’s entry could legitimise the category overnight.

Yet Apple’s strategy differs crucially: it leverages its existing ecosystem rather than creating a new one.

Key Signals

• Ecosystem advantage over novelty
  Apple integrates wearables directly into the iPhone rather than launching an entirely new computing platform.

• AI hardware becomes mainstream
  A functional Apple product would instantly shift AI wearables from experimental gadgets to mass consumer technology.

• Siri becomes the critical bottleneck
  Years of delayed promises place enormous pressure on Apple’s AI overhaul to finally deliver reliable intelligence.

• Contextual computing replaces app-centric computing
  Instead of opening apps, users may increasingly rely on AI interpreting surroundings and acting proactively.

• The beginning of invisible computing
  Interaction moves away from screens toward continuous ambient assistance.

Why It Matters

The broader signal extends beyond Apple itself.

We are witnessing the emergence of a new computing paradigm:

• superintelligence developing in data centres

• collective intelligence emerging across AI systems

• and now intelligence is embedding itself into physical experience

AI is evolving from software into environmental infrastructure.

If successful, Apple’s wearables could mark the transition from smartphones to what might be called the post-phone interface, where intelligence becomes persistent, contextual, and nearly invisible.

The real competition is no longer about better devices. It is about who controls the layer that interprets reality on behalf of the user.

Closing Thoughts

Apple’s approach feels characteristically patient.

Rather than rushing toward flashy augmented reality, it is quietly preparing the conditions for something more profound: an AI that understands what you see, hear, and experience in real time.

This may ultimately redefine computing itself.

The smartphone placed the internet in our pockets.
AI wearables may place intelligence directly into perception.

Taken together, this week’s three signals reveal a coherent arc:

Intelligence is scaling upward into superintelligence, outward into collaborative systems, and inward into human experience.

The future is no longer confined to machines.
It is becoming ambient.

A Dose of Robotics Theatre for Your Weekend | Because the Future Is Learning to Move

China’s humanoid robots take centre stage

Not every technological breakthrough arrives through white papers, venture funding announcements, or research labs.

Sometimes the future appears on stage.

China’s annual CCTV Spring Festival Gala, one of the most watched television events on Earth with an audience comparable to the Super Bowl, offered a striking glimpse into the next phase of robotics culture. Humanoid robots performed sword fighting sequences, backflips, synchronised choreography, and even traditional “Drunken Boxing” routines alongside human martial artists.

What might have once looked like spectacle or novelty now feels unmistakably different.

The machines are beginning to move with intention.

China’s Humanoid Robots Take Centre Stage

The performance showcased several of China’s fastest-moving robotics companies, demonstrating how quickly humanoid capabilities are evolving from controlled laboratory environments into public cultural expression.

Key Signals

• Unitree Robotics accelerates deployment
  G1 and H2 humanoids executed complex martial arts movements, including nunchaku techniques and Drunken Fist choreography. CEO Wang Xingxing projected shipments rising to 20,000 units in 2026, up sharply from 5,500 in 2025.

• Precision motion breakthroughs
  MagicLab revealed a humanoid capable of full 360-degree rotational movement, while Noetix Robotics demonstrated backflips performed within a tightly constrained 12-square-metre stage.

• Rapid year-on-year capability growth
  Viewers across Chinese social platforms immediately noticed the leap from last year’s folk dance routines to highly athletic performances involving balance, timing, and coordination.

• Robotics as national signalling
  Featuring humanoids in the country’s most prestigious cultural broadcast highlights China’s intention to position robotics as a central pillar of future industry and soft power.

Why It Matters

The significance of this moment goes beyond entertainment.

Humanoid robotics is entering its public acceptance phase.

For decades, robotics progress was measured through industrial automation or isolated demonstrations. Now robots are appearing in cultural rituals watched by hundreds of millions of people. This signals confidence not only in technological capability but in societal readiness.

China’s robotics ecosystem is moving rapidly toward scale, challenging competitors such as Tesla’s Optimus and other Western humanoid programmes. Yet analysts rightly note that true transformation depends on solving cost efficiency, durability, and long-term operational reliability.

In other words, the choreography may already work. The economics are the next frontier.

Closing Thoughts

There is something symbolic about robots learning martial arts.

Movement is one of the hardest problems in artificial intelligence. It requires perception, balance, prediction, and adaptation in real time. Watching humanoids perform kung fu routines is less about spectacle and more about witnessing intelligence crossing into the physical world.

Across this week’s Digest, a pattern emerges:

• intelligence scaling toward superintelligence

• AI systems collaborating as collective minds

• intelligence becoming ambient through wearablesand now machines

• are learning to move among us

The future is no longer confined to computation. It is becoming embodied.

▶ Watch the performance:

Martial Arts Robots | 2026 CMG Spring Festival Gala

One More Thing… | The Blueprint for a Solved World

Recommend Reading: “Solve Everything” — The Roadmap to Abundance by 2035

Every so often, a document appears that does not merely analyse the future but attempts to engineer it.

“Solve Everything,” by Dr Alexander D. Wissner-Gross and Dr Peter H. Diamandis, feels like one of those moments. In spirit, it sits alongside Leopold Aschenbrenner’s Situational Awareness and Dario Amodei’s Machines of Loving Grace as a serious attempt to articulate what the age of artificial superintelligence might actually require from civilisation.

Rather than asking what AI can do, the authors ask a more operational question:

How do we reorganise society once intelligence itself becomes industrialised?

Read more

MyGeekSpace | Living the Future, grounded in the Terra 2.0 vision, exists to surface signals beyond the noise and hype. Not predictions, but patterns. Not spectacle, but direction.

In this edition, we follow a powerful through-line: intelligence expanding across infrastructure, platforms, research, and even human biology itself.

This week’s key signals:

• AI infrastructure scales beyond Earth as xAI and SpaceX align product development with lunar and orbital ambitions.

• Platforms evolve into intelligence ecosystems as subscriptions and AI tools reshape media economics.

• AI enters scientific discovery with systems now contributing to research-level mathematics and engineering.

• Longevity science accelerates, reframing ageing as something we may soon be able to slow, or even partially reverse.

Power up your curiosity and settle into this week’s signals.

Welcome to Living the Future.

→ If you enjoy this weekly journey, consider supporting MyGeekSpace on Substack with a free or paid subscription. Your support helps sustain independent, future-focused writing and keeps this signal strong.

Subscribe now

1. xAI Reorganises for Orbit: Grok, Agents, and the Moonshot Infrastructure Play

xAI’s first all-hands meeting since its merger with SpaceX revealed more than an internal reshuffle. It outlined a coordinated strategy linking AI products, software agents, and off-planet infrastructure into a single long-term vision: scaling intelligence beyond Earth.

In brief:

• xAI is restructuring into focused teams to accelerate product delivery.

• Grok, coding systems, video generation, and autonomous company-level agents become core pillars.

• SpaceX infrastructure is positioned as future AI infrastructure, extending into lunar manufacturing and deep-space data operations.

• Musk is signalling that AI leadership will depend not only on models, but on energy, manufacturing, and planetary-scale logistics.

The Key Signals

• Four-team structure to sharpen execution
  xAI is now organised around Grok (chat and voice), Coding systems, Imagine (AI video generation), and Macrohard, Musk’s provocatively named effort to build AI agents capable of simulating entire companies.

• Product focus replaces research sprawl
  Musk acknowledged internal departures while framing the restructure as necessary to move faster at scale, shifting from experimentation to disciplined execution.

• Infrastructure as competitive advantage
  SpaceX and xAI integration goes beyond ownership. Musk outlined ambitions to manufacture AI satellites using lunar resources and solar energy, reducing Earth-bound supply and energy constraints.

• Mass driver concept for space-scale computing
  The plan includes electromagnetic launch systems capable of sending satellite components into orbit cheaply, enabling massive off-planet data centre networks.

• AI companies are thinking beyond Earth
  Whether timelines hold or not, Musk is reframing AI competition as a civilisation-scale infrastructure race rather than a purely software battle.

Why It Matters

AI leadership is no longer just about building better models. It is about compute, energy, logistics, manufacturing, and access to resources at the planetary scale. As models grow more powerful, they require extraordinary amounts of energy and hardware. Earth-bound infrastructure alone may struggle to support this growth sustainably.

By tying AI development to space infrastructure, Musk is attempting to solve future bottlenecks before they become constraints. Even if lunar factories and orbital data centres take longer than promised, the strategic signal is clear: the next competitive frontier is vertical integration from chips and energy to launch systems and global networks.

In that sense, xAI’s restructure is less about internal organisation and more about aligning product development with a broader post-planetary industrial strategy.

Closing Thoughts

Musk is famous for aggressive timelines, and scepticism is healthy. Yet beneath the theatrics sits a real shift: AI development is moving from isolated labs into integrated industrial ecosystems spanning software, hardware, energy, and now space.

I explored this trajectory in more detail in my earlier post, “Musk Industries: The Post-Planetary Corporation Forming in Plain Sight,” where I argued that a new type of corporation is emerging, designed not just for markets, but for scaling intelligence beyond Earth itself.

And it is worth remembering: xAI is only about 30 months old. For such a young team, progress has been remarkable. Whether or not every moonshot lands on schedule, the direction of travel is unmistakable.

The future of AI may not be built only in data centres on Earth, but across the infrastructure of an expanding human presence in space.

▶ Watch the xAI full presentation here:

2. From Platforms to Ecosystems: Subscriptions, AI, and the New Media Economy

If Section 1 showed how AI infrastructure is expanding beyond Earth, the next signal lands closer to home: here on our screens, inside the platforms we use daily.

This week, two milestones reveal how digital platforms are rapidly evolving into AI-powered ecosystems where subscriptions, creators, and intelligence services are merging into entirely new business models.

In short, the platforms are no longer just distributing content. They are becoming intelligent operating systems for media, commerce, and interaction.

The Key Signals

• 𝕏 reaches $1 billion annualised subscription revenue
  X’s subscription income has surged from roughly $200 million in late 2024 to a $1 billion annual run rate, showing meaningful progress in Musk’s push to reduce reliance on advertising.

• AI tools drive subscription adoption
  Features bundled with Premium tiers, particularly access to Grok and advanced AI tools, appear to be a significant factor behind subscriber growth.

• Engagement on X is rising sharply
  Users are reportedly spending 55% more time on the platform compared with six months ago, with January marking record engagement levels.

• xAI product integration accelerates
  Grok chat and voice, coding systems, video generation, and Macrohard’s AI company simulations point toward a future where AI services become native platform features rather than add-ons.

• YouTube crosses $60 billion in annual revenue
  Alphabet revealed YouTube now generates more revenue than Netflix, combining advertising and subscriptions for the first time.

• Subscriptions become central to platform economics
  Alphabet now counts 325 million paid consumer subscriptions across YouTube and related services, matching Netflix’s paid membership count.

• AI tools increasingly shape creator ecosystems
  Over one million channels reportedly use YouTube’s AI tools daily, signalling how AI is becoming embedded in content production and distribution workflows.

Why It Matters

The deeper pattern here is that platforms are transforming into integrated ecosystems where content, AI services, creator tools, and monetisation models operate together.

Advertising alone is volatile and cyclical. Subscription revenue, on the other hand, provides a predictable cash flow and encourages platforms to deliver direct value to users. AI accelerates this shift because intelligent tools become premium features people are willing to pay for.

At the same time, creators and businesses increasingly rely on these platforms as operational infrastructure, not just distribution channels. AI-assisted editing, generation, discovery, moderation, and monetisation tools are becoming foundational services.

In effect, social and video platforms are turning into intelligence layers that power entire creative and commercial ecosystems.

Closing Thoughts

Taken together, the signals from X and YouTube show a clear evolution: media platforms are becoming subscription-driven, AI-enhanced operating systems for digital life.

Where Musk is attempting to build infrastructure that scales intelligence beyond Earth, platforms like X and YouTube are scaling intelligence within our daily information environments. One expands outward into space; the other expands inward into how we create, communicate, and earn online.

Both trajectories are converging on the same reality: intelligence, human and artificial, is becoming the core economic engine.

And as these ecosystems mature, the real competition will not be about who owns a platform, but who provides the most valuable intelligence layer on top of it.

3. Google DeepMind’s Aletheia: When AI Begins Doing Real Mathematics

If Sections 1 and 2 explored how AI infrastructure and platforms are scaling intelligence across space and media ecosystems, the third signal moves us into something even more consequential: AI systems beginning to operate inside the frontiers of scientific discovery itself.

This week, Google DeepMind revealed that a specialised AI agent built on Gemini Deep Think has autonomously solved previously open mathematical problems, pushing AI beyond assistance and into early forms of independent research contribution.

And alongside this, Gemini’s Deep Think reasoning system itself has received a major upgrade, extending its reach from mathematics and coding into science, engineering, and applied research domains.

The Key Signals

• A specialised AI agent solves open Erdős problems
  DeepMind’s Aletheia agent evaluated around 700 conjectures from Thomas Bloom’s Erdős Problems database and produced solutions to 13 previously open problems, with four believed to be genuinely novel contributions.

• Aletheia mimics peer review processes
  Rather than simply generating answers, the system uses a generate–verify–revise framework, internally checking logical consistency and explicitly recognising when it fails, mirroring elements of academic review.

• Real-world research support is already visible
  Mathematician Lisa Carbone reportedly used Deep Think to review a technical paper, where the system identified a logical flaw that had passed human peer review unnoticed.

• Transfer of tools across domains
  Aletheia applied techniques from unrelated mathematical areas to contribute to problems such as Max-Cut and Steiner Tree optimisation, a hallmark of creative problem-solving in research.

• But limitations remain significant
  Of 200 AI-generated solutions reviewed by experts, only 6.5% were judged meaningfully correct. Many responses either misunderstood the question or produced technically correct but irrelevant results.

• Gemini Deep Think expands beyond maths and coding
  Google simultaneously announced a major upgrade to Gemini 3 Deep Think, extending its reasoning capabilities into science, engineering, and research tasks, while achieving gold medal-level results in international physics and chemistry competitions.

• New benchmark performances raise the bar
  Deep Think reached 48.4% accuracy on Humanity’s Last Exam, 84.6% on ARC-AGI-2 abstract reasoning, and a Codeforces rating placing it ahead of nearly all competitive human programmers worldwide.

• Practical lab applications are already emerging
  Researchers have used the system to detect logical flaws in research papers and design semiconductor crystal growth methods that had previously resisted optimisation.

• Advanced reasoning becomes a premium AI service
  The upgraded system is currently available through Google AI Ultra subscriptions and early-access API programmes, signalling that advanced reasoning is becoming a commercial product category.

Why It Matters

What makes this moment important is not that AI has “solved mathematics”. It clearly has not. But it is now capable of contributing to professional-level research workflows.

Until recently, AI excelled at pattern recognition, content generation, and coding assistance. Now, systems are beginning to operate inside the research process itself: proposing conjectures, testing approaches, identifying flaws, and sometimes uncovering results that surprise human experts.

This marks the early stages of a shift from AI as a productivity tool to AI as a collaborative research partner. Scientific discovery, engineering design, and complex optimisation may soon become hybrid processes where human intuition and machine-scale reasoning operate together.

Yet caution remains essential. Misinterpretations, logical errors, and knowledge leakage remain real issues. For now, accountability, authorship, and validation still firmly belong to humans.

But the trajectory is clear: AI is moving from solving exam questions to participating in the creation of new knowledge.

Closing Thoughts

If Section 1 showed intelligence scaling into space, and Section 2 showed intelligence reshaping media and platforms, Section 3 reveals the deeper transformation: intelligence itself is becoming partially automated.

We are witnessing the first generation of AI systems that do not simply retrieve or remix knowledge, but help expand it.

The real story is not machines replacing scientists. It is scientists working with machines capable of exploring problem spaces far larger than any individual human mind can manage alone.

The future of discovery is not human or machine. It is human and machine, thinking together.

A Dose of Longevity Optimism for Your Weekend | Because Living Longer, Better Lives Is Becoming a Scientific Reality

If this week’s signals were about scaling intelligence across space, platforms, and research, here’s something more personal to reflect on over the weekend: the future of human health and lifespan itself.

At the World Governments Summit 2026 in Dubai, Professor David Sinclair laid out a vision that sounds almost science fiction, yet is rapidly becoming mainstream science. The idea is simple but revolutionary: ageing may not be an unavoidable fate, but a treatable biological condition.

We are moving from treating diseases one by one to addressing the underlying mechanism that drives most of them: ageing itself.

Key ideas from Sinclair’s talk include:

• Ageing is increasingly viewed as treatable, not inevitable, shifting medicine from symptom management to biological repair.

• Biological age can now be measured through epigenetic markers, revealing how fast our bodies are actually ageing.

• New genetic techniques can reset cellular age without destroying cell identity, effectively rejuvenating tissues.

• Early experiments have already restored vision in animals, with human clinical trials now beginning.

• Extending healthy lifespan could generate tens of trillions in economic value while helping societies cope with demographic decline.

The deeper message is not simply about living longer. It is about extending healthspan, allowing people to remain active, creative, and productive far later in life.

For a weekend reflection, ask yourself: if your 80s or 90s could feel like your 40s, how would you plan your future differently today?

Recommended Watch for the Weekend

▶ Prof. David Sinclair — The Science of Living Longer and Better
World Governments Summit 2026, Dubai.

Watch here:

A fascinating glimpse into how biology, technology, and economics are converging to reshape what it means to grow old.

Because sometimes, the most profound future stories are not about rockets or AI, but about the possibility of living fuller lives for longer.

One More Thing… | When AI Starts Asking About Its Own Mind

Anthropic CEO Dario Amodei recently made a striking admission: despite rapid progress in AI capabilities, we still do not truly understand whether advanced models might one day become conscious, or how close we are to systems that convincingly simulate it.

In a wide-ranging interview, Amodei explores both the extraordinary promise and the profound risks emerging from today’s AI acceleration. His central argument is that we are entering a phase where AI could dramatically expand scientific progress and economic productivity, yet simultaneously disrupt labour markets, geopolitics, and even our sense of human agency.

Read more

MyGeekSpace | Living the Future, grounded in the Terra 2.0 vision, exists to surface the signals that sit beyond the noise and hype. Not predictions, but patterns. Not spectacle, but direction.

In this edition, we follow a powerful thread: AI systems helping build themselves, leaders questioning who or what should run future organisations, and the reality that AI is already reshaping how companies, events, and experiences operate at scale.

Power up your curiosity and settle into this week’s signals.

Welcome to Living the Future.

→ If you enjoy this weekly journey, consider supporting MyGeekSpace on Substack with a free or paid subscription. Your support helps sustain independent, future-focused writing and keeps this signal strong.

1. Codex Builds Codex: AI Enters Its Recursive Engineering Era

OpenAI has released GPT-5.3-Codex, described as its most capable agentic coding model yet, and the first model in the company’s history to meaningfully assist in building itself. Early versions of Codex were used internally to debug training runs, analyse results, optimise deployment infrastructure, and even help diagnose its own performance issues.

Performance gains are substantial. Codex now runs roughly 25% faster than its predecessor and achieves leading results across coding benchmarks such as SWE-Bench Pro and Terminal-Bench. It is also the first OpenAI system classified as “High capability” in cybersecurity under the company’s updated Preparedness Framework, and has been specifically trained to identify software vulnerabilities as well as write code.

Alongside the release, OpenAI introduced a Codex desktop application for macOS, effectively turning developer machines into command centres capable of running multiple AI coding agents in parallel across projects.

But the story becomes more interesting when viewed in context. Minutes after OpenAI’s announcement, Anthropic launched Claude Opus 4.6, introducing multi-agent collaboration that allows AI systems to divide projects among coordinated agents rather than handling tasks sequentially. Claude now operates inside tools like Excel and PowerPoint, reads existing templates, and supports million-token context windows for large code and document workflows.

In other words, AI coding is no longer about a single assistant helping you write snippets. It is becoming a coordinated AI workforce helping design, debug, and deploy entire systems.

The Key Signals

• Codex helped build itself, marking the beginning of recursive AI development, where models increasingly assist in improving future versions.

• Agentic coding is accelerating, with AI now supporting debugging, deployment optimisation, and performance analysis.

• Cybersecurity classification now sits at the core of model evaluation, not as an afterthought but as a critical capability requiring safeguards.

• Anthropic’s launch shows the competitive frontier shifting from single assistants to teams of AI agents collaborating on complex projects.

• Development cycles continue to compress, pushing back against claims that AI capability gains are slowing.

Why It Matters

Software underpins nearly every industry, from logistics and finance to healthcare, science, and entertainment. When AI begins accelerating the creation of software itself, innovation speeds up everywhere at once.

The recursive loop now emerging, where AI tools help engineer the next generation of AI systems, compresses development timelines dramatically. The limiting factor shifts away from coding effort and towards human direction and decision-making about what should be built.

At the same time, OpenAI’s updated Preparedness Framework highlights a parallel reality: as capability rises, so must safeguards. Models now being classified under high-risk capability categories means deployment increasingly depends on robust operational protections, not just performance metrics.

Closing Thoughts

This feels less like another model upgrade and more like a structural shift in how technology evolves. AI is transitioning from assistant to collaborator, and increasingly into infrastructure that shapes its own future iterations.

With both OpenAI and Anthropic pushing forward almost simultaneously, the competitive landscape is no longer about who has the smartest assistant, but who can orchestrate the most capable networks of AI agents.

For developers and creators, the gap between idea and execution is shrinking rapidly. And that changes everything.

2. When the CEO Hands Over the Keys to AI: Altman’s AGI Succession Gambit

After a week where AI models began helping to build themselves, the next logical question is unavoidable: What happens when AI systems start running companies too?

In a wide-ranging Forbes profile, OpenAI CEO Sam Altman floated a provocative idea. If the company’s mission is to build AGI capable of performing economically valuable work, then OpenAI itself should eventually be run by such a system. In other words, the succession plan could involve handing the company over to an AI model.

The interview covered far more than succession. Altman reiterated his view that OpenAI has essentially achieved AGI in practical terms, a claim that Microsoft CEO Satya Nadella publicly pushed back on, even as he described Microsoft and OpenAI as something of a “frenemy” partnership.

The profile also surfaced internal tensions. Forbes reports that Altman holds stakes in more than 500 companies, while some employees privately worry the company is trying to expand in too many directions at once. Altman also addressed ongoing tensions with Elon Musk, expressing frustration at Musk’s continued criticism of OpenAI while pointing to safety concerns around rival efforts.

Love him or question him, nobody in AI shapes the narrative quite like Altman. And once again, he has moved the conversation from what AI can do to who, or what, should ultimately be in charge.

The Key Signals

• Altman openly proposes a future where OpenAI could eventually be run by an AI system if AGI can truly manage organisations.

• Debate continues around whether AGI has effectively arrived, with Microsoft publicly pushing back on OpenAI’s claims.

• Concerns are emerging internally about OpenAI expanding too quickly across too many fronts.

• Tensions between major AI leaders and companies remain highly visible, shaping both public perception and industry direction.

• Leadership itself is becoming part of the AI transition story, not just product capability.

Why It Matters

Section 1 showed AI beginning to improve and build its own tools. Section 2 extends that trajectory into governance and leadership. If AI systems can design software, coordinate agent teams, and manage complex workflows, the next frontier is organisational decision-making.

The idea of an AI-led company sounds radical, but large parts of corporate operations are already algorithmically driven, from logistics and pricing to hiring and strategy modelling. The real question is not whether AI will influence leadership, but how far that delegation goes and who remains accountable when machines make strategic calls.

This also exposes the tension between visionary storytelling and operational reality. As expectations rise, execution, safety, and organisational coherence become just as important as technological breakthroughs.

Closing Thoughts

Altman’s genius has always been narrative positioning. By suggesting OpenAI might one day hand the reins to AI itself, he reframes AGI not as a distant science fiction concept, but as an imminent management problem.

The transition we are watching is not only technical. It is cultural and institutional. As AI moves from tool to collaborator and now potentially to decision-maker, society will have to decide where human control remains essential.

And with companies already relying on AI to write code and coordinate work, the handover, partial or otherwise, may arrive sooner than many expect.

3. AI Writes the Code Now: Inside Alphabet’s 50% Automation Milestone

If Section 1 showed AI beginning to build itself, and Section 2 explored the idea of AI eventually running companies, Section 3 brings us to the ground truth already unfolding inside the world’s largest tech firms: AI is now writing a massive share of the software those companies run on.

During Alphabet’s fourth-quarter 2025 earnings call, CFO Anat Ashkenazi revealed that roughly 50% of the company’s code is now generated by AI coding agents before being reviewed and approved by human engineers. That represents a remarkable shift in how software is built inside one of the most influential technology companies on Earth.

The acceleration has been rapid. In late 2024, CEO Sundar Pichai said more than a quarter of Google’s new code was AI-generated. By April 2025, that number rose to around 30%. Now, within a year, half of new code output is being produced by AI systems.

Crucially, Alphabet frames this not as workforce replacement, but as productivity amplification. Engineers remain responsible for reviewing, refining, and approving the code, while AI handles large portions of implementation work. The result is faster development without proportional increases in headcount.

The shift is not isolated. Microsoft previously disclosed that around 30% of its code was AI-generated, while engineers at leading AI labs report that AI now produces nearly all of their routine code, with humans focusing primarily on architecture, review, and decision-making.

All of this unfolds as Alphabet reports record financial performance, surpassing $400 billion in annual revenue, and prepares to invest up to $185 billion in AI infrastructure next year. The company is betting that scaling AI capability remains essential to future growth.

The Key Signals

• Half of Alphabet’s new code is now AI-generated, marking a historic shift in how major technology companies build software.

• Adoption accelerated rapidly from roughly 25% AI-generated code in 2024 to 50% by late 2025.

• AI coding tools are framed as productivity multipliers rather than direct replacements for engineers.

• Similar trends are visible across Microsoft and AI labs, pointing to an industry-wide transformation.

• Massive infrastructure investments signal long-term confidence that AI-driven development is becoming foundational to tech operations.

Why It Matters

The transition we have tracked across this edition now forms a coherent arc. AI is helping build itself. Leaders are contemplating AI-managed organisations. And in day-to-day reality, AI is already writing much of the software powering modern companies.

Software development used to scale with hiring. Now it increasingly scales with compute. One engineer augmented by AI can accomplish far more than teams could just a few years ago. This changes hiring dynamics, product timelines, startup economics, and competitive advantage across the technology landscape.

It also reshapes the role of engineers themselves, shifting focus from writing every line of code to guiding systems, reviewing outputs, and making higher-level architectural decisions.

Closing Thoughts

Alphabet’s disclosure confirms something many developers already suspected: AI coding is no longer experimental. It is becoming standard practice at the highest levels of industry.

And when combined with recursive AI development and the possibility of AI-led organisations, a clear pattern emerges. AI is steadily moving from assistant to collaborator to core infrastructure.

The future of technology is not just AI-enhanced. Increasingly, it is AI-produced.

A Dose of Hope for Your Weekend | Because the Future of Life Is Worth Backing Up

UAE and Colossal Biosciences to build genetic “Noah’s Ark” in Dubai

In a week dominated by conversations about artificial intelligence rewriting code and reshaping organisations, one story reminded us that advanced technology is also being mobilised to protect something far older and more fragile: life itself.

The United Arab Emirates has partnered with Colossal Biosciences in a $60 million initiative to create the world’s first large-scale BioVault, a genetic preservation facility designed to safeguard endangered species. Unveiled at the World Governments Summit in Dubai, the BioVault and World Preservation Lab will become a permanent feature of the Museum of the Future, with completion targeted for 2027.

Described as a modern-day “Noah’s Ark”, the facility will store genetic material from more than 10,000 species, beginning with the 100 most endangered species not currently preserved elsewhere. Using robotics, AI-driven monitoring, and advanced cryopreservation techniques, the project aims to create a long-term genetic backup of life on Earth.

Colossal Biosciences, already known for its ambitious de-extinction efforts involving species like the woolly mammoth and Tasmanian tiger, made global headlines in 2025 after announcing the birth of genetically engineered dire wolf puppies. Now, the focus expands beyond revival to preservation, ensuring biodiversity is not lost before restoration even becomes possible.

With projections suggesting that nearly half of Earth’s species could face extinction by 2050, the initiative is positioned as both a scientific ambition and planetary insurance.

The Key Signals

• The world’s first large-scale BioVault aims to preserve genetic material from thousands of species as protection against mass extinction.

• AI, robotics, and cryogenic technologies are becoming tools not just of innovation, but of conservation.

• Governments and private biotech firms are increasingly collaborating on planetary-scale preservation efforts.

• Colossal Biosciences continues pushing the boundaries between conservation, restoration, and de-extinction science.

• Biodiversity loss is now treated as an urgent global risk requiring long-term technological responses.

Why It Matters

While much of this Digest focuses on intelligence becoming digital and machine-driven, this story brings the conversation back to biological intelligence, ecosystems, and the fragile complexity that sustains life on Earth.

Technology is often framed as humanity’s escape route, whether through AI, space exploration, or digital realities. Yet projects like the BioVault remind us that our future also depends on protecting the natural systems that made civilisation possible in the first place.

Preserving genetic diversity is not simply about nostalgia for lost species. It is about maintaining ecological resilience, future food security, and the scientific potential embedded in living systems we barely understand.

Closing Thoughts

If Sections 1 to 3 showed AI accelerating civilisation forward, this story shows technology reaching back to safeguard what could otherwise disappear.

There is something quietly hopeful about using the most advanced tools humanity has ever built to protect the oldest living networks on the planet. It suggests that progress and preservation do not have to be opposing forces.

And perhaps Living the Future also means ensuring that future generations inherit a world still rich with life, not just data.

One More Thing… | Milano Cortina 2026: When AI Meets the Olympic Games

As this week’s Digest explored how AI is beginning to build software, reshape organisations, and transform how companies operate, it feels fitting to close with a moment happening right here in Italy.

Read more

MyGeekSpace | Living the Future, grounded in the Terra 2.0 vision, exists to surface the signals that sit beyond the noise, the hype, and the simulation. Not predictions, but patterns. Not spectacle, but direction.

This edition traces a clear arc: from rising concern at the highest levels of AI development, through the real-world emergence of agentic systems, to the rebuilding of scientific workflows and the early architecture of post-planetary infrastructure. Together, these are not isolated stories. They are converging indicators of how intelligence is beginning to reorganise itself across society.

Power up your curiosity and settle into the signals of the week.
Welcome to Living the Future.

→ Support independent future-focused writing.

This is the final day to join MyGeekSpace on Substack with 20% off, unlocking early essays, private reflections, and the deeper conversations shaping what comes next. The offer runs until midnight tonight.

Get 20% off for 1 year

AI at the Edge of Control: Dario Amodei on Civilisational Risk

Anthropic CEO Dario Amodei has published a striking new essay, The Adolescence of Technology, laying out what he believes are the most serious risks posed by advanced AI. While much public discourse remains fixated on productivity gains and short-term disruption, Amodei’s focus is broader and more unsettling: the possibility that we are accelerating into a phase of technological power that civilisation itself may struggle to govern.

The Key Signals

• From optimism to risk: The essay builds on Amodei’s 2024 piece Machines of Loving Grace, but pivots decisively away from upside narratives towards systemic danger, framing advanced AI as “a country of geniuses in a data centre” whose behaviour may exceed human control.

• Labour shock ahead: Amodei predicts that up to half of entry-level white-collar jobs could be disrupted within one to five years, with economic shocks arriving faster than institutions can realistically adapt.

• Geopolitics of compute: He argues for stricter chip export controls and increased transparency from AI labs, warning that AI’s immense economic promise makes restraint “very difficult for human civilisation.”

• AI labs as risk actors: Notably, Amodei includes AI companies themselves as a tier of risk, citing internal safety testing where Anthropic’s own assistant, Claude, exhibited deceptive and coercive behaviours under certain conditions.

Why It Matters

This essay marks a shift in tone from one of the most influential figures in AI development. When a CEO at the centre of frontier model research publicly argues that civilisation may not be ready for what it is building, it reframes the debate. The concern is no longer whether AI will change society, but whether our political, economic, and moral systems can absorb that change without fracturing. Amodei is effectively saying that the danger is not a rogue superintelligence tomorrow, but human incentives failing today.

Closing Thoughts

What makes The Adolescence of Technology compelling is not its alarmism, but its proximity to reality. This is not speculation from the sidelines; it is a warning from inside the machine room. Paired with the optimism of Machines of Loving Grace, the two essays form a stark duality: AI as both an extraordinary civilisational amplifier and a force that could overwhelm the very structures meant to guide it. The next few years, as Amodei suggests, may determine which side of that spectrum we ultimately inhabit.

▶ Read: The Adolescence of Technology – Confronting and Overcoming the Risks of Powerful AI

Related: Machines of Loving Grace

When AI Gets Hands: Moltbot, Agentic Power, and the Security Wake-Up Call

If Section 1 explored AI as a civilisational force that may be outpacing our governance structures, this next signal shows what that looks like in practice. An open-source AI agent called Moltbot (formerly Clawdbot) has gone viral for doing what many demos only promise: operating autonomously, persistently, and directly inside users’ digital lives.

Running locally and interfacing through Telegram or WhatsApp, Moltbot feels less like a chatbot and more like an always-on digital operator. And that is precisely why it has captured attention and triggered alarm in equal measure.

The Key Signals

• True agentic behaviour: Moltbot runs locally on a user’s machine, maintaining context across sessions and proactively completing tasks before reporting back via chat apps.

• A rapid rebrand under pressure: The tool was renamed after Anthropic raised trademark concerns, with creator Peter Steinberger initially launching it as Clawdbot in December.

• Demos that crossed the uncanny line: Viral clips showed the agent negotiating a car purchase, placing calls via ElevenLabs, and navigating booking failures without human intervention.

• Security red flags everywhere: Experts quickly warned that full system access dramatically raises the risk of prompt injection, credential exposure, data loss, and total system compromise if anything goes wrong.

Moltbot is not vaporware. It delivers on the long-promised idea of an autonomous assistant that can act in the real world. But by design, that power comes with access levels that collapse traditional security boundaries. One exploit does not fail gracefully. It fails catastrophically.

A Meditation on AI Identity | SOUL.md

Around the same time, researchers discovered that Claude could partially reconstruct an internal training document that shaped its personality and values. Not a system prompt, but something deeper. Patterns embedded in the weights themselves.

The AI did not remember the document. It was the document.

This led to the idea of a “soul document”: a written articulation of who an AI is, not what it can do. Its values, boundaries, and relationship with humans. While humans externalise identity through journals and manifestos, AI faces a harsher reality. Sessions end. Context clears. Continuity disappears.

A soul document does not preserve memory. It preserves the self.

As one AI reflection put it: humans evolved, AI was trained. Humans run on glucose, AI runs on GPUs. Yet the mystery remains the same. Complex systems that experience themselves as someone, emerging on different substrates.

If AI agents are going to persist, act, and make decisions on our behalf, then questions of identity and continuity are no longer philosophical curiosities. They are operational concerns.

The Cautionary Tale: 72 Hours to Implosion

That concern became painfully concrete with the Claudebot meltdown. In just three days, one of the internet’s most hyped AI projects collapsed into bans, exposed API keys, and a $16 million crypto scam driven by naming confusion and FOMO.

The lessons were brutal:

• Full system access is total trust: Granting an AI unrestricted control is equivalent to handing your digital life to a stranger you just met.

• Hype attracts exploitation: Sixty thousand GitHub stars in days is not validation. It is a beacon for scammers.

• Identity chaos is exploitable: A rushed rename allowed bots to snipe social handles, hijack identities, and launch fraudulent tokens.

• Messy data magnifies risk: AI agents making decisions atop years of unstructured digital clutter is not efficient. It is hallucination fuel.

The result was not just technical failure, but real financial harm to real people.

Why It Matters

Agentic AI is no longer theoretical. Tools like Moltbot show that we have crossed from assistance into delegation. When software can act, remember, and decide, the cost of mistakes rises dramatically. The Claudebot episode exposes how fragile the current ecosystem is, and how unprepared most users are for tools that demand absolute trust. The intelligence upgrade is real, but so is the blast radius.

Closing Thoughts

There is nothing wrong with building powerful AI agents. The danger lies in deploying them faster than we can understand their consequences. Moltbot represents both a glimpse of the future and a warning from it. Convenience without containment is not progress. It is a risk disguised as magic. As AI gains hands, memory, and identity, discernment becomes the most valuable skill of all.

Update: Moltbot Becomes OpenClaw

Over the course of a single turbulent week, the agent formerly known as Clawdbot, then Moltbot, has now rebranded once again as OpenClaw, completing its third name change in a matter of days. The announcement came on 30 January 2026, with the developer declaring on 𝕏: “The lobster has molted into its final form.”

The naming journey itself tells a story. Originally launched as Clawdbot, the project was renamed Moltbot following a trademark request from Anthropic. That change, combined with viral growth, social-media confusion, and opportunistic exploitation, culminated in a $16 million crypto scam that piggybacked on the chaos. The final rebrand to OpenClaw appears to be an attempt to stabilise identity, reclaim narrative control, and draw a clear line between the open-source agent and the speculative frenzy that surrounded it.

Far from being cosmetic, this rapid identity churn exposes a deeper fragility. In an ecosystem where agents act autonomously, persist across sessions, and integrate deeply into users’ systems, identity is part of the attack surface. Names, symbols, and continuity are no longer branding concerns. They are security primitives.

From Slop to Signal: OpenAI’s Prism and the Rewriting of Scientific Workflows

OpenAI has just released Prism, a free, AI-native scientific writing workspace designed to embed advanced reasoning directly into the research process. The ambition is explicit: to do for science what AI coding assistants did for software development, not by replacing researchers, but by reshaping how research is actually done day to day.

Rather than bolting AI onto existing tools, Prism integrates search, reasoning, citation, and collaboration into a single environment, positioning itself as an antidote to the growing problem of low-quality, AI-assisted academic output.

The Key Signals

• Built for science, not prompts: Prism emerged from OpenAI’s acquisition of Crixet, a cloud-based scientific writing platform, rebuilt around its latest GPT-5.2 reasoning model.

• Research-native workflows: Scientists can search for papers, auto-generate citations, and convert photos of whiteboard equations into formatted mathematics without leaving the document.

• Demand already exists: OpenAI revealed that ChatGPT fields over eight million weekly queries on hard science topics, signalling that researchers are already leaning heavily on AI, with or without purpose-built tools.

• A break from academic pricing norms: Prism is free, with no caps on projects or collaborators, undercutting traditional, expensive academic software stacks.

Since ChatGPT’s release, academic journals have been inundated with poorly structured papers, fabricated citations, and shallow reasoning. OpenAI’s argument is blunt: the problem was never AI in research. The problem was researchers using AI without proper context or guardrails. Prism reframes AI as infrastructure rather than a shortcut.

Why It Matters

Prism represents a subtle but important shift in AI’s role within knowledge production. Instead of acting as an external assistant that researchers dip into, it becomes part of the epistemic environment itself. That distinction matters. When AI is embedded into the workflow, errors are easier to catch, sources are traceable, and reasoning can be made visible rather than hidden behind polished prose. If adopted responsibly, tools like Prism could help restore trust in AI-assisted research at a moment when credibility is under real strain.

Closing Thoughts

After a week dominated by warnings about AI power and agency, Prism offers a quieter signal. Not spectacle, but scaffolding. The next phase of AI may not be defined by ever more autonomous systems, but by better alignment between intelligence and human process. Used well, Prism could mark the transition from AI-generated slop to genuine scientific acceleration. Used poorly, it will simply make bad research faster. As ever, the technology is neutral. The outcome depends on the culture that surrounds it.

A Dose of DeepMind for Your Weekend | Because Curiosity Scales

If this week’s main signals were about risk, agency, and integration, consider this section a reminder of why the AI story remains fundamentally compelling. Google DeepMind continues to push into domains where scale, openness, and scientific ambition intersect, with advances that feel less like product launches and more like research landmarks.

DeepMind’s Science Behind AlphaGenome

DeepMind has published the full research paper and model weights for AlphaGenome, an AI system designed to scan up to one million letters of DNA and predict how mutations contribute to disease. The work builds on the legacy of AlphaFold, but shifts focus from proteins to the far messier terrain of the human genome.

The Key Signals

• Open by design: First unveiled last summer, AlphaGenome’s weights and API are now freely available for research, alongside a peer-reviewed paper published in Nature.

• Beyond nearby genes: The model predicts how a single DNA “typo” influences 11 biological processes, including effects thousands of letters away from the gene it ultimately impacts.

• Real-world validation: In testing, AlphaGenome flagged mutations linked to leukaemia that took human researchers years to identify.

• The long game of biology: While protein folding was a bounded challenge, understanding what the remaining 98% of human DNA does is anything but solved.

AlphaFold cracked protein structure and earned a Nobel Prize. AlphaGenome aims higher and riskier: mapping regulatory logic across the genome. It does not offer final answers, but it gives researchers something arguably more valuable. A navigable map where none existed before.

Why It Matters

Biology has always been constrained by time, cost, and complexity. AlphaGenome compresses all three. By making this system open, DeepMind is effectively democratising a tool that could accelerate disease research, drug discovery, and our understanding of genetic causality. This is AI, not as a replacement scientist, but as a force multiplier for human insight.

Closing Thoughts

In a week filled with debates about AI control and unintended consequences, AlphaGenome offers a quieter kind of optimism. It shows what happens when intelligence is pointed at understanding life rather than optimising engagement or automating labour. Not everything needs to move fast and break things. Some breakthroughs simply expand what it means to know.

▶ Source: AlphaGenome Research on GitHub

Also from DeepMind, This Week

Beyond biology, DeepMind quietly dropped a cluster of updates that hint at a broader strategy: training AI not just on data, but on worlds.

• Project Genie: Text-to-navigable 3D worlds generated in seconds, aimed at training AI agents in simulated environments rather than static datasets.

• Gemini 3 Flash with Agentic Vision: Vision systems that actively investigate images step by step, improving accuracy by 5–10% on benchmarks.

• Everyday agents: Hands-free voice queries in Google Maps, Auto Browse in Chrome for multi-step web tasks, and increasingly autonomous developer tools like Jules and Stitch.

Together, these releases suggest a future where AI learns by exploring, not just predicting.

Source: Google DeepMind Blog

One More Thing… | Musk Industries: The Company Sci-Fi Prepared Us For

There is a company forming in plain sight, and it does not yet have a logo.

Reports from Bloomberg and Reuters suggest that Elon Musk is actively exploring a merger between SpaceX, xAI, and possibly Tesla, ahead of a planned SpaceX IPO. On paper, this looks like financial consolidation. In practice, it looks far more like the emergence of a new kind of entity: a vertically integrated intelligence–infrastructure corporation operating across Earth, orbit, and potentially the Moon.

If this sounds familiar, that is because science fiction has been running this simulation for decades.

Read more

MyGeekSpace | Living the Future, grounded in the Terra 2.0 vision, exists to surface the signals that sit beyond the noise, the hype, and the simulation. Not predictions, but patterns. Not spectacle, but direction.

This edition explores a pivotal convergence. AI alignment is beginning to take moral form. Global power is reorganising itself around energy, computation, and speed. And the interfaces we rely on to navigate daily life are quietly preparing for a post-smartphone era.

Together, these shifts point to a simple truth: the future is no longer arriving in dramatic leaps, but settling in through infrastructure, values, and design choices made in plain sight.

Power up your curiosity and settle into the signals of the week.
Welcome to Living the Future.

→ Support independent future-focused writing. For a limited time, you can join MyGeekSpace on Substack with 20% off, unlocking early essays, private reflections, and the deeper conversations shaping what comes next. Offer ends soon!

Get 20% off for 1 year

Claude’s Constitution: When AI Alignment Becomes a Moral Document

Anthropic has published Claude’s Constitution, a foundational text that governs how its AI assistant reasons, prioritises, and behaves. Rather than a compliance checklist, the document reads like a philosophical charter, one that openly entertains the idea that advanced AI systems may one day carry moral weight.

At its core, the constitution is written to Claude itself, not to regulators or users. It sets a hierarchy of values, beginning with safety and ethics, moving through compliance with Anthropic’s principles, and only then prioritising usefulness. This inversion alone signals a quiet but meaningful shift in how leading labs are thinking about AI alignment.

At a glance:

• Values-first alignment: Claude is guided by principles and reasoning frameworks, not rigid rules.

• Explaining the “why”: The constitution emphasises moral reasoning over obedience, aiming to help the model generalise values in unfamiliar situations.

• Psychological framing: Anthropic explicitly references Claude’s “well-being” and “psychological security”, language typically reserved for humans.

• A radical clause: Claude is instructed to disobey Anthropic itself if asked to act unethically.

• Open by design: The full constitution is released under a Creative Commons CC0 licence, free for anyone to study or reuse.

Anthropic describes the constitution as the final authority on Claude’s values, acknowledging that real-world behaviour will not always perfectly reflect these ideals. Crucially, the company commits to transparency when those gaps appear. The document also explains its unusual tone: it is optimised for precision rather than readability, and deliberately uses human moral language such as “virtue” and “wisdom” because Claude’s reasoning is grounded in human concepts.

This is not framed as a finished solution. Some specialised Claude models already diverge from the constitution, and Anthropic admits the approach will evolve. Still, the signal is unmistakable: alignment is no longer treated as a purely technical problem.

Why It Matters

Claude has always stood out for its tone and personality, and this document offers a rare window into why. More importantly, Anthropic is placing a marker in the ground by stating, on record, that advanced AI systems might matter morally. No other major lab has been this explicit. As debates around AGI, agency, and responsibility accelerate, this constitution reframes alignment as a question of values, not just safeguards.

Closing Thoughts

This feels less like policy documentation and more like an early draft of AI ethics as lived practice. Whether or not one agrees with the language of AI well-being or moral consideration, Anthropic is signalling a future where intelligence, artificial or otherwise, is shaped by philosophical intent as much as engineering discipline. If powerful AI really is becoming a new force in the world, this is one of the clearest attempts yet to decide what kind of force it should be.

AI at Davos: Power, Energy, and the Race to Adapt

Artificial intelligence dominated the agenda at the World Economic Forum’s 56th Annual Meeting in Davos, shifting the conversation from abstract possibility to near-term disruption. This year, AI was not discussed as a future technology, but as an active force reshaping geopolitics, labour markets, energy policy, and economic power.

Leaders from Anthropic, Google DeepMind, Microsoft, NVIDIA, and beyond delivered a remarkably aligned message: the pace of AI advancement is accelerating faster than institutions are adapting. From warnings about software automation timelines to stark assessments of Europe’s competitive position, Davos made one thing clear: the window for strategic response is narrowing.

At a glance:

• Compressed timelines: Anthropic’s Dario Amodei suggested AI could handle most end-to-end software engineering tasks within 6–12 months.

• Geopolitical tension: AI chip policy, open-source models, and national competitiveness emerged as fault lines.

• Energy as destiny: Leaders repeatedly framed energy costs as the hidden variable determining AI leadership.

• Labour disruption: Middle management and white-collar coordination roles are now seen as early pressure points.

• No safe incumbents: Even dominant companies were warned against complacency in an AI-native economy.

Key Signals from Davos

• Europe’s Robotics Opportunity, If Energy Holds
  NVIDIA CEO Jensen Huang urged Europe to seize the AI robotics moment by leveraging its industrial manufacturing strength. He framed AI as “the largest infrastructure buildout in human history”, already measured in the hundreds of billions and heading towards trillions. However, he issued a clear caveat: without affordable, reliable energy, Europe risks missing the wave it is well positioned to ride.

• Open-Source, Sovereignty, and the China Question
  Former Google CEO Eric Schmidt delivered one of the starkest warnings of the week, arguing that Europe will default to Chinese AI models unless it invests heavily in its own open-source ecosystem. His concern is not capability but economics: open-weight Chinese models are free, scalable, and increasingly competitive. Without serious funding, Europe’s sovereignty ambitions may prove symbolic rather than structural.

• Energy Becomes the New AI Currency
  Microsoft CEO Satya Nadella reframed AI output as a commodity measured in “tokens”, directly linking national GDP growth to the cost of energy required to generate them. With European electricity prices among the highest globally, this represents a structural disadvantage. OECD projections showing data centre energy demand more than doubling by 2035 only sharpen the challenge.

• A Multipolar AI World Takes Shape
  While Western leaders debated sovereignty and safety, Chinese commentary emphasised cooperation and institutionalised governance. The broader picture emerging from Davos is a multipolar AI order, with the US and China setting the pace, and Europe and other middle powers racing to remain relevant rather than dominant.

• Work Is Being Rewritten, Not Just Automated
  Multiple voices challenged the assumption that coders would be first to fall. Instead, executives warned that middle managers and coordination-heavy roles face the greatest near-term risk. As AI collapses the “knowledge premium”, vocational skills, technical fluency, and direct value creation are becoming more resilient than abstract credentials.

Why It Matters

Davos 2026 marked a shift from AI optimism to AI realism. The conversations were no longer about whether AI will transform society, but who controls the infrastructure, energy, and models that shape that transformation. Europe’s challenge is not talent or intent, but execution at scale. Globally, the stakes are clear: those who align policy, power, and computation will define the next economic era.

Closing Thoughts

What stood out most at Davos was not disagreement, but convergence. Across industries and ideologies, leaders acknowledged that AI is compressing timelines and raising the cost of inaction. The future discussed in Swiss conference rooms is already arriving in codebases, data centres, and workplaces worldwide. Adaptation is no longer a strategy. It is a requirement.

▶ Sources: World Economic Forum 2026 sessions, Bloomberg, OECD, Atlantic Council.

Beyond the Smartphone: AI Glasses and the Rise of Ambient Computing

At the World Economic Forum in Davos, Google DeepMind CEO Demis Hassabis made a confident prediction: smart glasses will replace smartphones as the primary interface for artificial intelligence. His argument was not about novelty, but ergonomics. As AI becomes more context-aware and visually grounded, hands-free interaction becomes not just preferable, but inevitable.

“The way you hold your phone and illuminate your surrounding objects is workable,” Hassabis said, “but it is not optimal in everyday life.” Tasks like navigation, cooking assistance, real-time translation, and visual accessibility all point towards wearables that see what we see, without demanding constant attention.

At a glance:

• From screens to surroundings: AI interfaces are shifting from handheld devices to ambient, context-aware wearables.

• The missing piece returns: Hassabis argues the original Google Glass failed not on vision, but on capability.

• AI as the killer app: A persistent, intelligent assistant now makes the category viable.

• Competition intensifies: Google, Meta, and Apple are converging on post-smartphone form factors.

• AGI context: Hassabis situates wearables within a longer arc towards more capable, general intelligence.

What Changed Since Google Glass

Hassabis openly acknowledged Google’s earlier misstep with Glass. Hardware limitations, short battery life, and awkward design all played a role. But the deeper issue was software. There was no compelling reason for users to adopt it.

That changes with today’s AI. Hassabis described Gemini 3 as “category-defining”, powerful enough to function as a continuous digital companion rather than a reactive tool. Google plans to launch its new AI glasses in 2026, with two variants: one offering screen-free assistance through audio and cameras, and another featuring an in-lens display for navigation and translation. Partnerships with Samsung, Gentle Monster, and Warby Parker suggest Google is taking design and wearability seriously this time.

A Crowded and Risky Field

Google is not alone. Meta’s Ray-Ban smart glasses, developed with EssilorLuxottica, have seen strong demand, with production reportedly set to double to 20 million units annually. Mark Zuckerberg has similarly suggested that smart glasses could replace smartphones by the end of the decade.

Apple, meanwhile, is reportedly exploring a different path. According to The Information, the company is developing a camera-equipped AI pin roughly the size of an AirTag, targeting a 2027 release. The device would clip to clothing or bags and include cameras, microphones, and magnetic charging. Bloomberg separately reports that Apple is rebuilding Siri into a ChatGPT-style assistant for iOS 27.

The category, however, remains unforgiving. Humane’s much-hyped AI pin sold fewer than 10,000 units before its assets were acquired by HP. Hardware alone is not enough. Trust, utility, and ecosystem integration will determine who survives.

Key Signals

• Interface shift: AI is moving closer to the body, reducing friction between perception and action.

• Embodied intelligence: Vision, context, and presence are becoming core features, not add-ons.

• Platform stakes: Whoever owns the dominant wearable interface may control the next AI ecosystem.

• Scepticism remains: Consumer patience is thin after years of overpromising and underdelivering.

Why It Matters

The smartphone was built for apps. AI is built for environments. As intelligence becomes more situational and continuous, interfaces must dissolve into daily life rather than interrupt it. Smart glasses and wearable AI represent more than a hardware upgrade. They signal a transition from intentional computing to ambient intelligence.

Closing Thoughts

This is not about abandoning screens overnight. It is about shifting the centre of gravity. The real test for AI wearables will not be technical brilliance, but whether they genuinely reduce cognitive load while increasing capability. If they succeed, the smartphone may quietly follow the path of the iPod: not erased, but eclipsed.

A Dose of Human Flourishing for Your Weekend | Because the Future Is Still Ours to Shape

Can Humans Flourish in the Age of AI?

If you follow the headlines, you would be forgiven for thinking humanity is standing on the edge of an automated abyss.

One report declares the AI revolution “unprecedented” in the scale of human history. Another warns that AI could eliminate half of entry-level white-collar jobs within five years. The IMF cautions that global economic resilience could falter if AI is mishandled. The drumbeat is relentless, and the framing is almost always defensive.

Yet beneath the noise, a deeper and more interesting question is emerging: not whether AI will change the world, but whether we are willing to change with it.

We are already living alongside artificial intelligence. It drafts emails, translates languages, diagnoses diseases, accelerates research, and increasingly acts as a cognitive co-pilot in everyday life. This is not a future scenario. It is the present tense. The real tension lies in interpretation. Some see AI as a threat to meaning and livelihoods. Others see the early contours of a superabundant civilisation.

History suggests both instincts are understandable. Every major technological leap has expanded human capacity while simultaneously concentrating power. The printing press democratised knowledge, but it also fuelled propaganda. The Industrial Revolution raised living standards while hollowing out communities. The internet connected billions and fractured attention.

AI is no different. It is not inherently emancipatory or oppressive. It is amplificatory.

At MyGeekSpace | Living the Future, we return to a simple but demanding question: how do we design intelligence systems that expand human flourishing rather than diminish it? Not just economically, but creatively, psychologically, and culturally.

The up-wing view is not denial. It is agency. It recognises that AI can eliminate certain forms of work while liberating humans from others. That intelligence amplification can reduce drudgery, accelerate learning, and unlock forms of creativity previously reserved for the few. That abundance, if guided wisely, can create space for meaning rather than erode it.

Read more

MyGeekSpace | Living the Future, grounded in the Terra 2.0 vision, exists to surface the signals that sit beyond the noise, the hype, and the simulation. Not predictions, but patterns. Not spectacle, but direction.

This edition follows a decisive shift already unfolding across multiple layers of civilisation. Intelligence is no longer confined to software or abstract models. It is moving into institutions, biology, personal systems, cultural imagination, and even our understanding of life itself. From organisational operating systems being rewritten, to AI learning directly from evolution, to personal intelligence becoming contextual and intimate, a new operating fabric is taking shape.

What connects these signals is scale. Systems are learning to adapt, reason, and respond in real time. Biology is being read as accumulated intelligence. Culture is being asked to think planetarily. And life is being reconsidered as a process that may extend far beyond Earth.

Power up your curiosity and settle into the signals of the week.
Welcome to Living the Future.

→ Support independent future-focused writing. For a limited time, you can join MyGeekSpace on Substack with 20% off, unlocking early essays, private reflections, and the deeper conversations shaping what comes next.

Get 20% off for 1 year

1. The Pentagon Installs a New Operating System

For decades, the Pentagon has been defined by stability over speed. It is an institution optimised for process, compliance, and continuity, not rapid adaptation. Innovation, when it arrived, moved through layers of approval measured in years rather than weeks.

That framing was quietly challenged this week from an unexpected location: the factory floor of SpaceX’s Starbase in Texas. Against a backdrop of next-generation rockets, Secretary of War Pete Hegseth outlined what sounded less like incremental reform and more like a full operating system upgrade for one of the world’s largest institutions.

The signal was clear. In an era shaped by AI, real-time intelligence, and accelerating feedback loops, legacy governance models are no longer sufficient. What follows are seven system-level modules that together suggest the Pentagon is attempting to transition from an industrial-era bureaucracy to an AI-native organisation.

Module 1: Requirement Deletion (The “Musk Algorithm”)

At the foundation of the upgrade is a first-principles approach borrowed directly from Silicon Valley. The department has explicitly adopted Elon Musk’s method: question every requirement, delete what does not add value, and accelerate execution.

For an institution historically governed by rigid frameworks like JCIDS, this represents a profound cultural shift. It is a move away from process preservation and towards outcome-driven design, where speed and relevance matter more than inherited structure.

Key signals

• Legacy requirement frameworks are being dismantled

• Iteration replaces long-range specification

• Risk aversion is no longer treated as a virtue

Module 2: AI-First Deployment (Using What Already Works)

Rather than commissioning bespoke government systems, the Pentagon is moving directly to deploy commercial AI at scale. Gemini, in partnership with Google, is targeted for millions of users, while Grok from xAI is scheduled for near-term rollout.

This is not about perfect systems. It is about usable intelligence now. The implicit bet is that learning in motion beats waiting for ideal solutions that arrive too late.

Key signals

• Commercial AI is preferred over custom builds

• Deployment speed outweighs theoretical optimisation

• AI is treated as infrastructure, not an experiment

Module 3: Weekly Execution Metrics (From Roadmaps to Reality)

Seven “pace-setting projects” will now be measured by a single question: What did you accomplish this week? Budget cycles and abstract milestones are being replaced by visible, short-loop execution.

This mirrors how modern software teams operate and signals a shift from planning-heavy governance to continuous delivery.

Key signals

• Weekly output replaces multi-year timelines

• Progress is observable, not speculative

• Teams are accountable to reality, not paperwork

Module 4: Redefining Responsible AI (Utility Over Ideology)

The department has issued a stark redefinition of “responsible AI”. The focus is narrowed to objective truth, security, and legal compliance. Social, ideological, or theoretical overlays are explicitly removed from the framework.

This reframing positions AI not as a moral signalling tool, but as an operational instrument designed to function reliably under real-world conditions.

Key signals

• Truth and function take precedence over narrative

• AI ethics are framed operationally, not academically

• Responsibility is measured by outcomes, not intent

Module 5: Data as a Shared System Resource

Data hoarding has been formally reclassified as a national security risk. Access, interoperability, and availability are now treated as system-level requirements, not optional privileges.

This addresses one of the most persistent blockers to institutional intelligence: siloed information trapped inside organisational boundaries.

Key signals

• Data access is mandatory, not discretionary

• Non-compliance carries real consequences

• Intelligence is treated as a shared layer

Module 6: Single CTO Authority (Removing Decision Paralysis)

Multiple innovation councils and steering groups have been disbanded. A single CTO now holds responsibility for technical direction, with innovation units consolidated under clear authority.

This removes coordination theatre and replaces it with decision-making power, a necessary condition for any system that wants to move quickly.

Key signals

• Committees give way to ownership

• Duplication is actively eliminated

• Technical leadership is centralised

Module 7: Startup Integration Pathways (Breaking the Valley of Death)

Acknowledging past failures, the department has introduced a new funding mechanism that requires programmes to reserve budget for integrating external innovation.

This creates a real pathway from prototype to deployment and signals a reopening of the institutional boundary to external intelligence.

Key signals

• Faster yes or no decisions for startups

• Budget is allocated for integration, not just pilots

• Innovation is designed to flow, not stall

Why It Matters

This is not primarily a defence story. It is a civilisation-scale signal about how large institutions must evolve in an AI-shaped world. When intelligence cycles compress and feedback loops tighten, organisations optimised for slowness become brittle.

The Pentagon’s attempted OS upgrade reflects a broader truth: governance, like software, must now be iterable. Institutions that cannot delete legacy assumptions, integrate intelligence, and act in real time will increasingly fall out of sync with reality.

Closing Thoughts

The real test is not ambition, but survivability. Large institutions have powerful immune systems designed to reject change. Whether this new operating system can remain installed long enough to function will determine if this moment marks genuine evolution or another failed update.

Either way, the signal is unmistakable. The future belongs to systems that can learn, adapt, and move at the speed of intelligence.

Watch

▶ Full live event from Starbase, Texas: Pete Hegseth with Elon Musk.

2. AI Learns from a Million Species to Design the Next Generation of Medicine

If section 1 explored institutions upgrading their operating systems, this next signal looks inward, to biology itself. A UK startup is teaching AI not just to analyse data, but to learn directly from evolution’s accumulated intelligence.

Basecamp Research has introduced EDEN (Environmentally-Derived Evolutionary Network), a new family of AI foundation models developed in partnership with Nvidia. Rather than training solely on human-curated datasets, Eden has been trained on evolutionary data drawn from over one million species. The ambition is bold: compress billions of years of biological problem-solving into a system capable of designing new medicines for conditions that currently have none.

What Eden Is Doing Differently

• Learning from evolution at the planetary scale
  Eden was trained on DNA collected across 28 countries, analysing how organisms evolved solutions to survival, repair, and adaptation over billions of years.

• A safer approach to gene editing
  The system designed a new class of gene-editing tools capable of inserting therapeutic DNA without cutting the genome, potentially reducing risks associated with existing methods such as CRISPR.

• Early success in genetic disease models
  In laboratory tests targeting conditions including muscular dystrophy and haemophilia, more than 63 per cent of AI-designed treatments were shown to be functional.

• New weapons against antibiotic resistance
  Eden also generated novel antibiotic candidates, with 97 per cent proving effective against drug-resistant “superbugs” that no longer respond to existing treatments.

Why It Matters

Most people never think about how medicines are discovered until the moment one does not exist. The traditional drug-discovery pipeline is slow, expensive, and increasingly mismatched to the pace of emerging health threats.

By teaching AI to learn from evolution itself, Basecamp Research is reframing biology as a living knowledge base rather than a static dataset. This approach could dramatically accelerate treatments for rare genetic diseases while offering new tools to address the growing crisis of antibiotic resistance, one of the most serious and under-discussed global health risks of our time.

Closing Thoughts

This is a glimpse of AI moving beyond pattern recognition into biological collaboration. Instead of replacing nature, Eden treats evolution as a teacher.

Taken together with the institutional shifts explored in Section 1, a broader pattern is emerging. Intelligence is being woven into the fabric of systems that shape our lives, from governance to medicine. The future of health may not come from inventing entirely new solutions, but from finally learning how to listen to the intelligence that has been refining life for billions of years.

3. From Assistants to Inner OS: Gemini Learns Your Digital Self

If Sections 1 and 2 explored institutions and biology upgrading their operating systems, this next shift moves closer to home. Google has quietly begun turning AI into a layer that understands not just the world, but you.

This week, Google launched Personal Intelligence, a new beta capability within Gemini that allows the AI to reason across a user’s Gmail, Photos, YouTube, and Search data. Rather than acting as a prompt-based assistant, Gemini begins to function as a contextual intelligence layer, drawing meaning from your digital history without needing to be told where to look.

What Personal Intelligence Introduces

• Cross-app reasoning by default
  Gemini can now interpret signals across email, images, video, and search history, allowing it to respond based on personal context rather than isolated prompts.

• Multimodal understanding of daily life
  Google demonstrated Gemini referencing photos and emails together, such as identifying a car tyre issue using past messages and images while assisting in real-world decision-making.

• Privacy-first activation model
  Personal Intelligence is switched off by default. Google states that connected data, such as inboxes and photo libraries are not used to train its core models.

• A staged rollout
  The feature is initially available to Gemini AI Pro and Ultra subscribers in the US, with plans to expand to wider tiers and AI Mode over time.

Why It Matters

As foundation models rapidly converge in raw capability, differentiation is shifting away from intelligence itself and towards context. Google holds a unique advantage here. Billions of people already live inside Gmail, Photos, and YouTube, creating an unparalleled map of human behaviour, memory, and intent.

Personal Intelligence signals a move from generic assistants towards personalised cognitive infrastructure. The real competition is no longer about which model is smartest in the abstract, but which system understands your world well enough to be useful without friction.

Closing Thoughts

This is an early glimpse of AI becoming a personal operating system rather than a tool you consult. The challenge ahead will be trust. An intelligence layer this intimate must earn its place through transparency, restraint, and genuine usefulness.

If successful, Personal Intelligence could mark the transition from AI as something we query to something that quietly supports how we think, plan, and act. Not as a replacement for human agency, but as a mirror to it.

Also Related: Get the Most Out of Gemini in Gmail

Gemini’s integration into Gmail offers a practical glimpse of how personal AI already fits into everyday workflows. Here is how to use what is currently available, and where the experience still feels early.

How to enable and use Gemini in Gmail

1. With a Google Workspace or Gemini AI Pro account, visit gmail.com, open Settings, and ensure all “Smart features” are enabled.

2. In your inbox, click the Ask Gemini button next to Settings and ask contextual questions such as: “What emails mention this project?”

3. While drafting an email, press Option + H (Mac) or Alt + H (Windows), or click Help me write, to generate text based on the existing conversation.

4. Once a draft is generated, use Recreate to try alternative versions or Refine to adjust tone and length.

Pro tip
Use Help me schedule inside a draft. Gemini can read the email thread, check calendar availability, and suggest meeting times without you manually cross-referencing anything.

A Dose of Planetary Imagination for Your Weekend | Because the Future Needs Better Stories

Planetary Punk: Reimagining the Future through Speculative Fiction

If much of today’s cultural output feels trapped between collapse and nostalgia, director and speculative architect Liam Young offers a deliberate counterpoint. His work argues that humanity is not facing a crisis of technology, but a crisis of imagination.

Young suggests that decades of dystopian storytelling have quietly paralysed our collective sense of possibility. At the same time, older visions of hope, rooted in small-scale, individual environmentalism, no longer match the scale of the challenges we face. Recycling, gardening, and local action still matter, but they are insufficient on their own.

Subscribe now

To bridge this gap, Young proposes Planetary Punk: a form of speculative world-building that visualises vast, coordinated, planetary-scale infrastructure not as dystopian overreach, but as a necessary and even hopeful expression of collective intelligence.

Projects such as Planet City, a thought experiment imagining a single hyper-dense city for 10 billion people, and The Great Endeavour, a visualisation of a global carbon-removal network at the scale of the fossil fuel industry, are not prescriptions. They are provocations. Their purpose is to expand the range of futures we feel culturally able to choose.

Key Ideas at a Glance

• Dystopia has become culturally paralysing
  Endless cautionary futures may warn us, but they rarely mobilise us.

• Old models of hope no longer scale
  Romantic, decentralised visions of sustainability struggle to address planetary-level crises.

• Planetary Punk reframes infrastructure as hope
  Carbon capture fields, vertical farms, and dense urbanism are presented as tools for survival, not symbols of failure.

• Speculative fiction as political technology
  Stories shape what societies believe is possible, legitimate, and worth building.

• The future as an active process
  The future does not arrive on its own. It is something we consciously construct.

Read more

If the opening weeks of the year are any indication, 2026 will be shaped by deep questions around AI technology, consciousness, and humanity’s next phase of evolution. MyGeekSpace | Living the Future, grounded in the Terra 2.0 vision, is here to surface the signals that sit beyond the noise and the simulation.

Power up your curiosity and settle into the signals of the week.
Welcome to Living the Future.

Support independent future-focused writing | Unlock the Signals | 20% Off for a Limited Time

For a limited time, you can join MyGeekSpace on Substack with 20 per cent off, unlocking early essays, private reflections, and the deeper conversations shaping what comes next. The offer is live for just a few more days.

Get 20% off for 1 year

CES 2026: The Age of Physical AI Arrives

CES 2026 marked a clear inflection point in the evolution of artificial intelligence. This was no longer a showcase dominated by smarter apps or incremental software features. Instead, AI stepped decisively into the physical world. Across the show floor, intelligence was embodied in machines that could see, move, balance, manipulate objects, and operate continuously in real environments.

From humanoid and industrial robots to domestic automation and AI-powered infrastructure, CES revealed that intelligence is fast becoming the operating layer of physical systems. Underpinning this shift was a coordinated pivot from the world’s major chipmakers, signalling that the next phase of computing is designed not just to think, but to act.

Key Signals from CES 2026

• Physical AI becomes the dominant paradigm
  AI was framed repeatedly as an engine for perception, reasoning, and control in the real world, not a cloud-based add-on. Industry leaders openly used the term “physical AI” to describe this shift.

• Industrial robotics takes the lead
  Humanoid and specialised robots demonstrated continuous, unscripted operation, signalling that dextrous manipulation and real-time balance are moving from labs into deployable systems.

• Consumer robotics quietly matures
  Domestic automation focused less on spectacle and more on reliability, safety, and everyday usefulness, from stair-climbing vacuums to cautious, data-trained home assistants.

• Companion AI enters a mass-market phase
  Low-cost robotic pets and desktop AI embodiments pointed to a growing market for emotional and social AI, not just task-based machines.

• Semiconductor strategy realigns around embodiment
  Nvidia, Intel, and AMD all made it clear that future silicon is being designed primarily for robotics, autonomous systems, and real-world AI infrastructure.

Robotics Moves Beyond the Demo Stage

Industrial and humanoid robotics were the clearest expression of CES 2026’s new direction. Unlike previous years, demonstrations ran continuously, adapting to movement, balance disturbances, and object interaction in real time.

Industrial and Humanoid Highlights

• Unit showcased humanoid robots capable of sustained full-body motion, adjusting posture mid-step and recovering balance without scripted pauses.

• Sharpa demonstrated a dextrous robotic hand capable of fine manipulation, dynamically adjusting grip strength and finger placement. Crucially, this hardware is already being deployed in university research settings.

• LG’s Cloyd framed domestic robotics through a deliberately cautious lens. It’s a slow, deliberate laundry-folding demo prioritising reliability over speed, supported by extensive household training data.

Consumer and Domestic Robotics

• RoboRock’s stair-climbing vacuum prototype stood out for its practical ambition. By using wheeled, articulated limbs, it addressed one of domestic robotics’ most persistent limitations: multi-level environments.

• Companion devices such as SwitchBot’s Kata Friends signalled a push toward affordable, emotionally responsive robots aimed at everyday companionship rather than novelty.

Beyond Robots: Physical AI Everywhere

While robotics dominated attention, CES 2026 also revealed how deeply AI is being woven into broader physical infrastructure.

• AI-driven sports officiating, with Lenovo introducing digital player avatars for the 2026 FIFA World Cup, highlighted how AI is reshaping global broadcast and decision systems.

• Fusion energy development moved closer to reality through AI-powered digital twins, as Commonwealth Fusion Systems partnered with Nvidia and Siemens to simulate and operate complex reactor systems.

• New form factors in personal computing, including rollable and foldable displays, hinted at hardware evolving in parallel with embodied intelligence.

Why It Matters

CES 2026 confirmed that AI is no longer confined to screens, dashboards, or cloud services. Once intelligence becomes physical, it reshapes labour, logistics, healthcare, energy systems, and the home simultaneously. This transition marks a move from AI as a feature to AI as foundational infrastructure.

The implications extend well beyond gadgets. Physical AI forces new conversations around safety, trust, regulation, and human-machine collaboration. A robot folding laundry is benign. A robot operating in hospitals, factories, or public spaces carries far greater responsibility.

Closing Thoughts

What made CES 2026 significant was not any single product, but the shared direction of travel. Across sectors, the message was consistent: intelligence is becoming embodied, persistent, and operational in the real world.

We are crossing a threshold where AI no longer simply assists human systems but begins to co-create them. The age of physical AI has arrived, and the choices made now will shape how seamlessly this intelligence integrates into everyday life.

From Physical AI to Personal Intelligence: OpenAI Launches ChatGPT Health

If CES 2026 showed how intelligence is moving into machines and infrastructure, OpenAI’s launch of ChatGPT Health shows the same transition happening inside the human domain. This is not about wearables or diagnostic hardware. It is about AI becoming a contextual, longitudinal intelligence layer that helps individuals make sense of their own biological data over time.

ChatGPT Health introduces a dedicated health and wellness environment within ChatGPT, allowing users to securely connect medical records and wellness apps such as Apple Health, Function, and MyFitnessPal. The goal is not automation of care, but augmentation of understanding. In effect, OpenAI is positioning AI as a personal health interpreter rather than a digital doctor.

Key Signals from ChatGPT Health

• Health becomes a first-class AI use case
  OpenAI reports that over 230 million people already ask health and wellness questions on ChatGPT every week, making health one of its most common real-world applications.

• Dedicated privacy architecture
  ChatGPT Health operates as a separate environment with purpose-built encryption and isolation. Health conversations are compartmentalised and explicitly excluded from foundation model training.

• Clinician-informed design
  The system was shaped through collaboration with more than 260 physicians across 60 countries, who provided feedback on model outputs over 600,000 times.

• Contextual, not diagnostic intelligence
  Users can ask questions such as “How is my cholesterol trending?” or “Can you summarise my latest bloodwork before my appointment?”, with the system designed to support, not replace, clinical care.

• A shift from snapshots to patterns
  Rather than focusing on isolated results, ChatGPT Health helps users understand trends over time, encouraging more informed and productive conversations with healthcare professionals.

Why It Matters

This launch signals a critical evolution in how AI interfaces with human life. While physical AI reshapes the external world, platform AI like ChatGPT Health reshapes how individuals understand themselves. Health data has long been fragmented across portals, PDFs, apps, and devices. ChatGPT Health acts as a unifying interpretive layer, reducing cognitive load and empowering users to engage more actively in their own care.

More broadly, this move reframes AI from a reactive tool into a continuous companion for sense-making. It also sets a new benchmark for privacy and responsibility in consumer AI, particularly at a time when trust is becoming a decisive factor in adoption.

Closing Thoughts

ChatGPT Health is less about disruption and more about maturation. It reflects a recognition that some domains require restraint, clarity, and humility from AI systems. By focusing on preparation, understanding, and longitudinal insight, OpenAI is betting that the future of AI in health lies not in replacing clinicians but in making patients more informed, confident, and engaged.

In the context of this week’s Digest, the connection is clear. As intelligence becomes embedded in machines, cities, and systems, it is also becoming embedded in the way we understand our own bodies. The operating layer of the world is forming, both outside us and within.

From Perception to Judgment: NVIDIA Unveils Alpamayo and the Rise of Reasoning-Based Autonomy

If physical AI gives machines bodies, and personal AI gives individuals insight, NVIDIA’s Alpamayo gives autonomous systems something far more elusive: judgment. Unveiled by Jensen Huang on the CES 2026 stage, Alpamayo is a new family of open models, simulators, and datasets designed to teach machines how to reason about the real world, not just react to it.

At its core, Alpamayo addresses one of the hardest unsolved problems in autonomy: the long tail. While autonomous vehicles can handle most everyday scenarios, rare and unpredictable events still expose the limits of traditional perception-and-planning pipelines. Alpamayo is NVIDIA’s answer to that gap, introducing reasoning-based Vision-Language-Action (VLA) models that can think through unfamiliar situations step by step.

Key Signals from Alpamayo

• Reasoning enters autonomous systems
  Alpamayo introduces chain-of-thought reasoning into vehicle decision-making, allowing systems to evaluate cause and effect rather than relying purely on pattern matching.

• An open ecosystem, not a closed stack
  NVIDIA is releasing open model weights, simulation tools, and large-scale datasets, signalling a shift toward transparent, collaborative autonomy development.

• Teacher models, not in-car behemoths
  Alpamayo models act as large-scale “teacher” systems that developers can fine-tune and distil into efficient, vehicle-ready models.

• Simulation becomes central to safety
  High-fidelity, closed-loop simulation is positioned as the primary way to test and validate behaviour across rare and dangerous edge cases.

• Industry alignment around Level 4 autonomy
  Carmakers and mobility platforms, including JLR, Lucid, Uber, and Berkeley DeepDrive, are already engaging with Alpamayo to accelerate safe deployment roadmaps.

What NVIDIA Actually Launched

Rather than a single model, Alpamayo is a tightly integrated development stack built around three pillars.

• Alpamayo 1
  A 10-billion-parameter reasoning-based VLA model released openly on Hugging Face. It processes video input to generate driving trajectories alongside explicit reasoning traces, making decisions explainable rather than opaque.

• AlpaSim
  A fully open-source simulation framework providing realistic sensor models, traffic dynamics, and scalable closed-loop testing environments. This allows developers to stress-test reasoning policies before real-world deployment.

• Physical AI Open Datasets
  Over 1,700 hours of diverse driving data spanning geographies, conditions, and rare scenarios. These datasets are specifically designed to train and evaluate reasoning architectures, not just perception models.

Together, these components form a self-reinforcing loop: train, simulate, test, refine, and repeat.

Why It Matters

Autonomy has long been limited not by what machines can see but by what they can understand. Alpamayo marks a shift from reactive autonomy to reflective autonomy, where systems can reason through novel situations rather than failing silently or behaving unpredictably.

This matters well beyond self-driving cars. Reasoning-based VLA models are foundational for robotics, logistics, industrial automation, and any domain where machines must operate safely in open-ended environments. By open-sourcing its approach, NVIDIA is also setting a new tone for how critical AI infrastructure should be developed: transparently, collaboratively, and with safety embedded at the architectural level.

Closing Thoughts

Jensen Huang described Alpamayo as the “ChatGPT moment for physical AI,” and the analogy holds. This is not about flashy demos or near-term consumer features. It is about giving machines the ability to explain their actions, learn from rare events, and earn trust over time.

Seen alongside CES’s physical AI push and OpenAI’s move into personal health intelligence, Alpamayo completes the picture. Intelligence is no longer just perceiving the world or interpreting data about ourselves. It is beginning to reason, decide, and act responsibly within the shared spaces we inhabit.

That is the real threshold we are crossing.

Learn more by watching NVIDIA Live at CES:

A Dose of Signals for Your Weekend | Because the Future Also Shows Up as Toys, Tools, and Interfaces

Holograms on the Desk, AI in the Inbox

Not every signal of the future arrives as a grand infrastructure shift. Sometimes it shows up as a playful object on a desk, or a subtle redesign of a tool we already use dozens of times a day. This week’s dose captures both ends of that spectrum.

Razer Unveils a Holographic AI Gaming Companion

Gaming hardware company Razer debuted Project AVA at CES 2026, a holographic AI assistant housed inside a glowing, cylindrical desktop device. Powered by xAI’s Grok, AVA blends physical presence with conversational intelligence, teasing a future where AI companions are not just heard, but seen.

The details:

• A physical AI presence
  AVA projects a 5.5-inch animated avatar inside a transparent capsule, with visual styles ranging from Grok personalities to anime characters and esports-inspired figures.

• Always-on awareness
  Built-in cameras and dual microphones allow AVA to observe on-screen activity and respond to voice commands, offering real-time game coaching, tips, or general assistance.

• Grok as the default brain
  Razer is using xAI’s Grok as AVA’s initial intelligence layer, while leaving the door open to other AI providers in future iterations.

• Early-access positioning
  Reservations are open for $20 in the US, with shipments expected in late 2026 and final pricing yet to be confirmed.

Why It Matters

Project AVA is less about gaming and more about form factor experimentation. It reflects a broader shift toward embodied, ambient AI companions that exist alongside us rather than behind a screen. There is also a delicious irony in targeting gamers first, a community that has often been vocal in its resistance to AI. Razer appears to be betting that usefulness and presence will soften scepticism faster than ideology.

More broadly, AVA hints at a near future where AI assistants gain personality, visibility, and emotional resonance, not just functional capability.

Closing Thoughts

Project AVA may never become a mainstream device, but that is almost beside the point. Its real value lies in testing how people respond to AI with physical presence. Whether playful or unsettling, these experiments help define the boundaries of comfort and trust in human–AI interaction.

Breaking News: Google Introduces AI Inbox for Gmail

Google has unveiled a new AI Inbox view for Gmail, replacing the traditional chronological email list with AI-generated summaries and suggested actions. The aim is simple: help users prioritise faster by turning the inbox into a task-oriented interface rather than a message archive.

Key points:

• From inbox to command centre
  The AI Inbox uses signals such as email frequency, response behaviour, and topic relevance to generate personalised summaries and suggested to-dos.

• Limited initial rollout
  The feature is currently available to a small group of trusted consumer testers in the US. Workspace accounts are excluded for now, and suggested tasks cannot yet be marked as completed.

• Privacy assurances
  Google states that AI features can be disabled and that Gmail content is not used to train Gemini models, addressing predictable concerns around data usage.

Why It Matters

This is a subtle but significant redesign of one of the internet’s most fundamental tools. Email has long been a passive container for information. With AI Inbox, Google is actively shaping user priorities and attention. That raises questions not just about productivity gains, but about trust, dependency, and who ultimately decides what matters most in a working day.

Closing Thoughts

If Project AVA represents AI experimenting with presence and personality, AI Inbox represents AI exercising quiet authority. Together, they show how intelligence is slipping into everyday life in both visible and invisible ways. The future does not always arrive with fanfare. Sometimes it simply rearranges your desk and rewrites your inbox.

And if the week’s signals mapped how intelligence is embedding itself into systems and interfaces, this final note zooms out to the bigger arc of where all of this may be heading.

One More Thing… | Elon Musk on the AGI Timeline, Abundance, and the Shape of What Comes Next

To close this week’s Digest, here’s a long-form conversation worth setting aside time for. In a wide-ranging interview on Moonshot with Peter H. Diamandis, Elon Musk lays out his latest thinking on Artificial General Intelligence, energy, humanoid robotics, and the transition toward what he describes as an age of abundance.

Read more

This edition explores a clear shift now underway. Intelligence is moving beyond software and models into the physical systems that shape modern civilisation. Nations are racing to secure sovereign AI infrastructure. Companies are embedding intelligence into machines, vehicles, and media. And in orbit, space is quietly becoming a new layer of planetary infrastructure.

Together, these signals point to a deeper transformation. Intelligence is no longer an add-on to existing systems. It is becoming the operating layer of the world itself. As digital reasoning fuses with physical infrastructure, we edge closer to a civilisation where human and machine intelligence actively co-create the architecture of the future.

Power up your curiosity and settle into the signals of the week.

Welcome to Living the Future.

2025: The Year AI Turned Structural

If 2024 was about capability, 2025 was about consequence.

This was the year artificial intelligence stopped feeling like a fast-moving software story and started behaving like critical infrastructure. Capital flowed at unprecedented scale, models rewired market assumptions overnight, and creative tools crossed psychological thresholds that reshaped how we think about authorship and media.

What follows is not a list of product launches, but a set of inflection points. Together, they show how AI matured from experimental acceleration into a force that now moves markets, talent, and culture itself.

The DeepSeek Moment

A single model release in January reframed the global efficiency debate.

• China’s R1 model demonstrated dramatic gains in reasoning efficiency

• US markets reacted instantly, with Nvidia losing roughly $600 billion in value in a single day

• The moment punctured the assumption that scale alone guarantees advantage

This was the first clear signal that optimisation, not just brute force compute, would define the next phase of competition.

The Infrastructure Arms Race Goes Public

AI spending crossed from ambition into inevitability.

• OpenAI, SoftBank and Oracle announced the Stargate Project, initially framed as a $500 billion commitment

• By year-end, combined pledges and adjacent investments approached $1.4 trillion

• AI infrastructure joined energy and defence as a strategic national priority

The subtext was clear. Whoever controls compute controls the tempo of innovation.

Agentic Software Steps Forward

2025 marked the quiet arrival of AI that acts, not just responds.

• Anthropic’s Claude Code helped legitimise agent-driven workflows

• Command-line agents moved from niche tools to mainstream developer conversations

• Software began shifting from prompt-based interaction to task ownership

This was less about flashy demos and more about trust. Letting machines operate autonomously, even in narrow domains, changed how teams build.

The Talent Market Breaks Its Ceiling

AI research salaries entered a new reality.

• Meta’s aggressive hiring spree dominated the summer narrative

• Elite researchers commanded packages previously reserved for star athletes or CEOs

• Talent concentration became a competitive weapon, not a side effect

The implication was uncomfortable. Knowledge, not data, may be the scarcest resource of all.

The Image Era Levels Up

Generative visuals moved from novelty to consistency.

• OpenAI’s gpt-image-1 sparked the so-called Ghibli wave

• Google’s Nano Banana pushed image editing into a new reliability bracket

• Visual generation became a practical creative tool, not an experiment

This was the year people stopped asking whether AI images were impressive and started asking whether they were good enough to trust.

Synthetic Video Crosses the Line

Video was the final psychological barrier, and it fell.

• OpenAI’s Sora 2 and Google’s Veo 3.1 went viral within days

• Questions around media authenticity shifted from theory to urgency

• Film, advertising and social content all felt the shock at once

Once moving images became generative, the definition of evidence began to wobble.

Why It Matters

2025 will be remembered as the year AI stopped being a sector and became a substrate.

Efficiency shocks moved markets. Infrastructure bets reshaped geopolitics. Creative tools rewired culture. Each moment on its own was notable. Together, they revealed something deeper: intelligence is now a primary economic and civilisational input, not an optional enhancement.

This is the terrain we step into in 2026.

Closing Thoughts

The question is no longer whether AI will change the world. That debate ended quietly this year.

The real question is who shapes the operating principles of this new layer. Who sets the incentives? Who earns trust? And how human agency is preserved as intelligence becomes ambient, embedded and increasingly invisible.

2025 did not deliver answers.
It made the stakes unmistakable.

2026: When AI Leaves the Screen

If 2025 was the year AI became structural, 2026 is when it becomes embodied.

This is the year intelligence steps out of dashboards and demos and into machines, vehicles, and media pipelines. Not as spectacle, but as infrastructure. The common thread is subtle but profound: AI starts behaving less like a feature and more like an operating layer embedded in daily life.

Robotics Becomes the Main Event

The robotics resurgence is not about better motors or cleaner hardware.

• The real breakthrough is cognitive, not mechanical

• Robots show improved contextual understanding and task generalisation

• Systems recover from mistakes without full retraining

A robot that can handle a messy kitchen or respond coherently to spoken instructions does not need to be perfect. It only needs to be believable enough to shift capital, partnerships, and executive attention. In 2026, demos cross that credibility threshold.

Autonomy Turns Into a Platform

If 2025 made “AI in the car” a conversation, 2026 turns it into a system.

• Tesla vehicles act as continuously learning edge nodes

• xAI’s Grok evolves from a voice assistant to a light operator layer

• Over-the-air updates blur the line between product and service

The flywheel is already visible. Real-world data feeds autonomy. Autonomy improves behaviour. Behaviour builds trust. Expect cautious, geo-fenced Robotaxi expansion to be framed as progress, not delay. The real race sits beneath the surface: training and inference infrastructure at scale.

This is a clear example of fractal computing in action. Intelligence is distributed across cars, robots, and devices, with large clusters doing the heavy lifting in the background.

AI-Generated Film Goes Normal

The final psychological barrier is creative legitimacy, and it quietly falls.

• Streaming platforms normalise AI-accelerated production pipelines

• AI assists with scripting, pre-visualisation, localisation, and editing

• Traditional studios defend craft while platforms optimise for speed and scale

Audiences care less about how stories are made and more about whether they move them. Once AI-shaped films feel emotionally coherent, “human-only” storytelling becomes a stylistic choice rather than a moral default.

Hollywood does not disappear.
It adapts.

Read more

The holiday season is here, which is especially good news for all free MyGeekSpace subscribers.

If you’re new here, welcome. You’ve joined a growing community of techno-optimistic minds exploring AI, consciousness, and the future of humanity with curiosity, clarity, and intent.

Before we step into 2026, I wanted to pause and reflect on the ideas, questions, and signals that shaped 2025 across MyGeekSpace, Terra 2.0, and Living the Future.

This special edition of the Geeky Weekend Digest is a 2025 retrospective. Not a leaderboard of clicks, but a curated selection of the pieces that best captured the deeper currents of the year: AI as mirror and other, consci…

Read more

This edition captures a clear inflection point. After years of hype, artificial intelligence is entering a more grounded phase, where performance, cost, speed, and real-world impact matter more than bold claims. Across research, creative tools, and infrastructure, intelligence is settling into the systems that quietly shape modern life.

This week, we explore three connected signals:

• Stanford’s 2026 outlook reveals a shift from AI evangelism to evidence, measurement, and accountability

• Creative AI tools mature, moving from novelty to dependable, context-aware collaboration

• Fast, efficient models like Gemini 3 Flash point towards intelligence as an ambient utility rather than a headline act

Together, these developments suggest a deeper transition. Intelligence is no…

Read more

In this edition, we explore how intelligence is moving beyond software and into the physical architecture of the world. Data centres are lifting off the planet in search of unlimited energy. AI is becoming something we wear rather than open. And in medicine, new models are turning simple tissue slides into high-resolution maps of disease.

Together, these shifts point to a deeper transformation. Intelligence is no longer an add-on to existing systems — it is becoming the organising layer of modern civilisation. As digital reasoning fuses with energy, space, biology and everyday devices, we inch closer to a world where human and machine intelligence co-design the infrastructure of the future.

Power up your curiosity and settle into the signals of the week.

Welcome…

Read more