The Great Reallocation
Where value flows when AI does the work and humans reclaim their time
Earlier this summer, Elon Musk claimed Grok had outperformed OpenAI’s best model. Not too long after, GPT-5 landed. It wasn’t a mind-blowing leap forward, but it was a meaningful step. And it underscored a more important point: even if the frontier models don’t keep leapfrogging each other every few weeks, the rate of progress—even at a slower pace—will compound into innovations that feel extraordinary just a few years from now.
The takeaway isn’t just that “AI will keep getting better.” It’s that the fabric of the economy, the structure of work, and the ways we spend time will be reshaped on a timeline human institutions are not well-equipped to manage.
For the past few months, I’ve been forcing myself to zoom out — beyond the hype cycles — and think about where the biggest, most defensible opportunities will be for startups and for venture capital. Not in the next six months, but in the world we’re clearly steering toward over the next 5, 10, 20 years. That’s what this week’s post is all about.
This isn’t a prediction of “what will be.” It’s a structured mental model:
If AI takes over the majority of repetitive, high-cognitive work…
If robots become physically capable and AI-native…
If foundational models continue to commoditize software…
…then what products, networks, and infrastructure will matter most?
premise 1 — the AI workforce will shrink companies and end “full employment” as we know it
Engineering is the clearest early signal. GPT-5 can already architect and build MVP features with minimal human guidance (not necessarily production ready, though). I think it’s generally accepted that, in the next few years, we won’t see massive engineering teams building SaaS products. We’ll see an architect and one or two engineers overseeing AI that writes, tests, and deploys production code on its own.
Extend that trend to other knowledge fields — law, finance, consulting, even creative work — and AI won’t just “augment” jobs. It will do most of them. The humans left in the loop will be there for oversight, ethics, and the occasional edge case.
That means:
Companies will be run by orders of magnitude fewer people.
The humans in those companies will have far more time on their hands.
The majority of today’s “knowledge work” will disappear or morph into AI-enabled supervision.
It’s not just white-collar. Blue-collar work will change too. Which brings us to…
premise 2 — every home will have a personal robot
Tesla is openly working on consumer humanoids. Figure is racing to market. At some point in the near future, having a robot will be as normal as having a car.
But these won’t be clumsy Roombas. Powered by frontier models, they’ll have generalist competence across thousands of tasks:
Repairs: Robot diagnoses your AC problem, orders the correct part, and installs it when the package arrives.
Cooking: Robot designs a meal plan based on your diet, does the grocery shopping, and has dinner plated by 7 p.m.
Logistics: Robot coordinates car repairs, handles trip planning, packs the luggage, and walks the dog.
The secondary impact here is as big as the robots themselves: when they take on household labor, errands disappear from human calendars. That’s hours per day returned to people’s lives — multiplied by billions of people.
Note: this vision also opens the door to a number of startup infrastructure opportunities in the physical world to support this future.
premise 3 — AI commoditizes most software
If a single prompt to GPT-8 can generate a better project-management tool than an entire funded startup can build in six months, traditional software moats collapse.
Defensibility in software will hinge on:
Proprietary networks — user bases or networks that can’t be cloned by a model.
Exclusive data — the only thing a foundation model can’t guess.
Embedded distribution — products deeply integrated into existing workflows or hardware.
Everything else? The margin is competed away. Which is why the most important plays in the AI era are not “apps,” but categories that AI cannot fully replicate or instantly commoditize.
the five biggest opportunity areas
1. Networks — The last true software moat
Networks can’t be spun up by an LLM. You can’t prompt your way to 500M active users who trust each other, share content, and build culture together.
That’s why consumer social will make a comeback. Platforms like Instagram, X, Substack, and whatever comes next will become more defensible, not less. In a world with infinite AI content, human-anchored networks become the filter, the trust layer, and the place where attention really lives.
2. Gaming & Virtual Worlds — The next economy
If robots and AI free up 6–8 hours a day for billions of people, some of that time will be spent in physical hobbies. But a massive portion will flow into virtual worlds — not as 2D screen games, but as immersive, Ready Player One-style universes.
A real-world example today is Sandbox VR. It’s the v0 of what’s coming — a location-based experience where you put on a headset, pick up a gun or a sword, and step into a fully realized game world. Now, imagine that Sandbox VR experience:
Costs a fraction of today’s price.
Takes up a fraction of the physical space.
Lives in every home.
Is paired with purpose-built games designed for massive, persistent, fully immersive worlds.
These worlds will:
Be persistent and fully explorable.
Have functional economies where you can earn and spend real-world money.
Allow people to build careers inside them.
The opportunity isn’t just in building the worlds — it’s in the infrastructure around them: payments, identity, governance, interoperability across universes. Whoever controls that stack controls the economy of the 22nd century.
3. Entertainment — Hyper-personalized and participatory
Today’s entertainment is mass-produced. Tomorrow’s will be generated for you. Entire films, games, and story arcs will be customized to your preferences, generated on demand by AI.
It’s not just passive viewing. You’ll co-create the worlds you consume. And you’ll expect those worlds to be as interactive as real life — with your favorite actors, fictional or otherwise, playable in real-time.
This isn’t decades away. We’ll see early forms of it in the next five years. I’m actively advising some companies here; more to come soon.
4. Transportation — Beyond Earth
The travel industry will change when more people have more free time. Intra-Earth travel will surge. But the truly long-term plays are in interplanetary transport and infrastructure: getting humans and goods between Earth, the Moon, Mars, and beyond.
Companies that own the supply chains, logistics, and propulsion tech here won’t just be building rockets — they’ll be building the future’s trade routes.
5. Robotics Infrastructure — The picks and shovels of the robot economy
If every home and business has a humanoid robot, the bigger opportunities are in the unseen layers:
Navigation systems for real-world environments.
Payment infrastructure so robots can transact on your behalf.
Secure internet protocols optimized for machine-to-machine browsing.
Maintenance ecosystems — parts, service networks, and upgrades.
This is the robotics equivalent of AWS — the layer that makes everyone else’s robots possible.
healthcare as a cross-cutting layer
Healthcare deserves its own mention because robots + AI can radically expand access to care. Imagine at-home diagnostics, routine testing, and ongoing monitoring — handled by your personal robot, with instant escalation to telemedicine or in-person specialists when needed.
This is more than convenience. It’s the infrastructure for cheaper, more accurate, and more preventive healthcare on a global scale.
a note to end on
These are just some of the things I’ve been thinking about — not a complete map, not a conclusive prediction. I’m sure there are gaps, and I’m sure parts of this will play out differently than I imagine.
But in the patterns I’m seeing today, these are the opportunity areas that feel both inevitable and defensible in a world where AI and robotics change the nature of work, time, and human potential.
If you’re working in any of these spaces — or see something I’ve missed — I’d love to talk.