AI didn't make companies stupid. It just stopped hiding it.

For a while, a company could look competent just by having competent people in the building. You couldn't see the strategy from the outside, so you assumed there was one. AI took that assumption away. The moment a leadership team could generate a plausible-sounding memo, deck, or product roadmap in thirty seconds, it became obvious how many of those memos, decks, and roadmaps were never that different from the thirty-second version to begin with.

That's the real story of the last two years, not the layoffs themselves but what the layoffs revealed. Most of the people let go weren't replaced by AI doing their job. They were replaced by a company deciding it could get away with not having that job done at all, and using AI as the cover story. Cutting cost was never the same thing as protecting value. It just used to be harder to tell the difference.

Cheap has a smell

If everything gets easy and cheap at the same time, none of it is actually worth much, because the thing that made it worth something was that it used to cost someone real effort to produce. Alex Smith's point about convenience applies almost perfectly here: the moment friction disappears from making a thing, so does most of its value, because the friction was doing more work than anyone gave it credit for. It was filtering out the people who didn't care enough to push through it.

AI is very good at removing friction from the parts of work that were never the point. Drafting the first version. Formatting the slide. Writing the boilerplate. None of that was ever where the value lived, it was just where the time went, and companies that were already coasting on the appearance of output are now finding out that when you remove the friction, there's nothing underneath it. Just a faster version of nothing.

Cory Doctorow's enshittification story is the corporate half of this. A platform gets good because it has to compete for people. It gets worse once it doesn't have to anymore. Companies did the same thing with talent for the last twenty years, they competed for it because they had to, and the competition kept a floor under how bad things could get. AI let a lot of leadership teams believe that floor was gone. Turns out it was the only thing holding anything up.

The antidote is doing the hard thing on purpose

Here's the part that should be good news and mostly isn't being treated as one. If cheap and easy have stopped meaning anything, the stuff that's still hard to do is now worth more than it's been in a long time, not less.

Blue oceans don't stay blue because nobody's found them. They stay blue because getting there requires doing something nobody else is willing to do the hard way. Right now that's an unusually short list: actual craft, actual judgment, actual research instead of a summary of research, a product built by people who understand the problem instead of people who prompted their way to a plausible-sounding solution. All of that got cheaper to fake and more valuable to actually do, at exactly the same time.

The companies about to get exposed are the ones betting that nobody will notice the difference between the two. Some won't. A lot of people can't tell fast from good, not yet. But the market for people who can tell the difference is growing, not shrinking, and it rewards exactly the kind of friction most companies just spent two years trying to engineer out of existence.

Make the harder thing on purpose. That's not a nostalgia move. It's the only remaining strategy that a language model can't do for your competitor overnight.