Gone. Just gone. Two weeks of work. Gone to waste? Two minutes earlier I had been perfectly happy. The new feature worked on the first try. Exactly the way I’d described it. I clicked through the application one more time, just out of habit, and suddenly stopped. Two features that had been working flawlessly for weeks were gone. Not broken. Just gone.
Until recently, companies at least had to make the same mistakes independently. One organization might overvalue prestigious universities. Another might mistake confidence for competence. A third might quietly screen out unconventional careers. Their judgments were often flawed. But they were flawed in different ways.
We all knew AI would eventually generate fake citations. That was almost the boring part. The more interesting question is why so many of them passed through systems designed to evaluate knowledge in the first place.
The first time I spent an entire afternoon working with AI, I closed my laptop with that strangely satisfying feeling of having done hard intellectual work. My brain felt cooked. I had compared models, refined prompts, rewritten outputs, tested workflows, chased better phrasing, discarded entire approaches. It felt intense. Dense. Productive.
A few years ago, I contributed an essay to a book about running. It was called Die Philosophie des Laufens (The Philosophie of Running) and published by Mairisch, a small independent press with excellent taste and the kind of literary courage that larger houses often like to claim for themselves.