Why I think the next 5 years matter more than

I’ve spent seven years on this blog carefully hedging. “Maybe.” “Probably.” “I could be wrong.” “Someday.” The caveats were responsible. The uncertainty was honest.

I’m running out of caveats.

What’s real right now

Not predicted. Not speculated. Not “in five years.” Real, today, operational.

AI models that reason through complex problems step by step. Anthropic’s Claude can think through a multi-step architecture problem and arrive at a conclusion that accounts for constraints I mentioned three conversations ago. OpenAI’s o1 solves math problems that graduate students struggle with. DeepSeek showed you don’t need a hundred billion dollars to build a model this good.

Humanoid robots that walk, carry, and learn from watching videos. Figure AI has a contract with BMW. They’re deploying in factories. Not demoing. Deploying. The robot learns new tasks in hours, not months.

The most powerful rocket ever built landing itself and being prepared for reflight in weeks. SpaceX’s Starship program has moved from “prototype that explodes” to “vehicle on a schedule.”

Transistors that are 10 atoms wide. TSMC is manufacturing at 2nm. The physical limit of silicon is visible. The successor technologies are already in labs.

Self-driving cars carrying 100,000 passengers per week with no driver and no fatalities. Waymo is a real transportation company.

Each of these alone is significant. Together, they’re something else.

The convergence

Here’s what I think people are missing. These aren’t separate technologies on separate timelines. They’re converging.

An AI model that can reason gives a robot the ability to plan. A robot that can learn from video gives AI a body. A rocket that’s reusable gives space exploration economics. Chips at 2nm give everything more intelligence per watt. Self-driving algorithms that work in cars work in robots that navigate warehouses that fulfill orders that ship on rockets.

The threads are weaving together. And when threads weave, the fabric that emerges has properties that none of the individual threads possess.

In the next five years, I think we’ll see: humanoid robots in thousands of factories and the first ones in homes. Self-driving cars in 20+ cities with millions of weekly riders. AI models that can reason, plan, and execute multi-hour tasks autonomously. Chips approaching 1nm (or the first post-silicon architectures). Starship flying frequently enough to make space infrastructure economically viable.

Why this time is different

I know this sounds like hype. Every generation thinks it’s living through the most important transition in history. Usually that’s hubris.

But I’ve been tracking these specific technologies for seven years. I’ve watched the progress curves. And the thing that’s different about this moment isn’t the promise. It’s the delivery. These aren’t prototypes. They’re products. They’re not in labs. They’re in cities and factories and orbit.

The gap between “science fiction” and “Saturday morning” has never been this thin.

What I’m worried about

I’d be dishonest if I didn’t mention what scares me.

The energy requirements. AI data centers are consuming electricity at a rate that could strain grids. The water requirements for chip manufacturing are staggering. The environmental cost of the future isn’t zero.

The labor displacement. When robots work in factories and AI handles knowledge work and cars drive themselves, the number of humans whose skills become less valuable grows. We don’t have a plan for that.

The concentration of power. The companies building these technologies are among the largest and most powerful in history. The decisions they make affect billions of people who have no say in those decisions.

I believe the next five years will change more than the last fifty. I’m awed by it. I’m excited by it. And I’m worried about it. All three feelings are correct. All three deserve attention.

The future is arriving faster than our institutions can adapt. That’s the sentence I keep coming back to. Faster than our institutions can adapt.

I don’t know how that gap closes. But I know it needs to.

Fifty years from now, someone will look back at 2025-2030 the way we look back at 1969-1975. The years when everything shifted. The hinge. I think we’re on it.

Related thinking: