I was inside the last hype cycle

I didn't watch the Web3 wave from the outside. I was in it, close to the founders making the calls. Not the famous names, mostly smaller projects, which is the better seat for what I want to tell you. I wasn't a commentator, I was the person founders called when the stakes were high and the path wasn't clear. And I won't pretend I was a sceptic all along. I believed, and I was good at it.

Work on enough of these bets in a short space of time and you stop seeing projects and start seeing the machine. You watch belief get manufactured, a technology stop being a tool and turn into a religion, capable founders make calls they'd never make with a clear head because everyone around them is screaming that they move now or miss it forever. I'm not describing that from the outside. I felt it too.

What it teaches you, eventually, is that the hype was never the signal. The hype is the weather. It moves the same way every time, whether the thing underneath is real or not, and the ones who come through are the ones who can tell real ground from a good story.

Near the end, the same idea kept crossing my desk in different costumes. A marketplace where people would package their own methods for others to buy, and the early joiners earned from referrals. The momentum came from bringing the next person in, not from the thing being any good. It promised more than it could safely hold, on tokenomics nobody had thought through. And underneath it, the AI was doing all the real work while the Web3 was just the wrapper that made it feel like something you could own a piece of. That was the pattern everywhere by 2024. The value had already moved, and most of the room was still facing the other way, cheering the old thing.

So I followed it. Not because I swapped one wave for another, but because I'd learned, the expensive way, how to tell where the substance actually was. That's the only skill that counts in a hype cycle, and it's rare, not because it's clever, but because almost nobody gets enough goes to learn it.

So here's how I read AI, having watched this up close. The signal is real. AI isn't Web3, and the thing underneath does what it says, which wasn't always true last time. But a lot of what I'm watching is still theatre. Companies announce AI to move a share price, vendors sell what they can't deliver, founders adopt it because the room is loud rather than because they've worked out where it changes their numbers. Same machine, new paint.

You tell the difference the way you always could. Ignore the announcement, look at the consequence. Real adoption shows up where a wrong answer is expensive and the technology lowers that cost. Theatre shows up where its main job is to get mentioned. If you can't find the place it changes a real decision with real stakes, it's paint.

My guess is that over the next couple of years a lot of the loudest AI initiatives, the ones launched for the narrative, will quietly produce nothing, for the same reasons the loudest Web3 projects did. The real value will land somewhere quieter, with people using it where the stakes are high and barely saying a word. I've seen this film, which is why I think I know how it ends.

So I stepped back and went to build the opposite. Systems people actually own, where their information is protected rather than harvested, where the value is the thing itself and not the story around it. If you're making a high-consequence call on AI right now and you want someone who's watched this pattern run and isn't selling you the wave, that's the work I do now.