Ten predictions and observations from the beautiful madness of the AI era. The good and the bad.
1. The word "moat"
If I hear "moat" or "there are two kinds of X and Y is the one that matters" one more time, I'll go berserk. And more seriously: online communities, and if we're not careful open source software too, are in long-term trouble.
2. It goes way beyond software
In the six months since I started using AI as the daily driver, I've run structural calculations, designed a full exterior door in CAD and fab it on a CNC machine, fixed a few appliances, designed and fab a faucet adapter for my RO system, and started building my own small 6-wheel robot platform. I would never have attempted this before, the amount of prior research would have been overwhelming. So, it clearly goes way beyond software and programming.
3. Capability plateaus, tooling doesn't
Raw model capability is plateauing. But the integration, the outreach, the tooling around the models still has years of catching up to do. We ain't seen nothing yet. Also that's where the money is.
4. The revolution is regulatory
Still, the real revolution will be about regulation, not technology. It's only a matter of time before AI gives better diagnostics than humans, and anything built on pattern matching follows the same path. The hard part is the bureaucracy and adoption, and everything will wait on that.
5. The best teacher, not the best programmer
There may be many programmers better than Claude, but I'm fairly sure there are few, if any, better teachers. Universities are done in the form we know them. The professor who simply stands there and explains stops making sense the moment regulation allows you to just sit the exam.
6. The supervisor economy
We're moving toward a supervisor economy. The machine does the work; a human understands it well enough to check it. That's the seat worth fighting for, because the value stops being in producing the output and starts being in judging it.
7. Single source of truth means automated
Any domain with a single source of truth will be automated in your lifetime. This is the point I'm most certain about. If your job is applying a fixed set of rules mechanically over a domain, hard times are coming. Even with every regulatory brake, within 10 - 20 years any economy that wants to stay competitive will automate that work. So, pick skills that keep you in the supervisor seat. Or live in Europe. I'm almost certain we'll protect the old jobs right up until the rest of the world forces us not to.
8. Technology was never the moat
Large players build the future, small players reinvent the same things in a loop. If Claude taught me one thing, it's that technology has never been a competitive advantage, and it's less of one now than ever. My ideas are really not unique. Yours probably aren't either.
9. Whoever ships the AI-native OS wins
Once hardware catches up and we run models locally, whoever ships the first genuinely AI-driven OS becomes the next Apple. Maybe Apple becomes the next Apple. Same story for phones.
10. And if I hear "moat" one more time
I'll go berserk.