A video about Claude Fable showed up in my feed this week. Same claim as the last one, and the one before that: this is the model that finally ends design and code as jobs.
Phir wahi cycle.
I actually sat with this one instead of scrolling past. The work in the demo was fast and clean, technically right in every frame. What I've started noticing, the more of this I look at, is the absence of anything someone would've had to argue for. A decision that could've gone another way. That's the part a demo reel can't show, because there isn't one.
A good chunk of what lands in my Figma file or my PR queue at work now starts as an AI draft. Three button states, whether the component needs all three or not. A paragraph that opens with a claim, backs it with three points, and closes by restating the claim. None of it is wrong. None of it required anyone to decide anything.
My own first AI drafts look exactly like that. All of them. What's different is what happens after.
I point at what's off, it fixes it, I point at the next thing. That loop, again and again, used to run through a senior designer's time and a code review queue, over months. Now it's free, and it's instant.
The obvious move, especially early in a career, is to skip straight to the fixed version: hand AI the brief and ship whatever comes back. That skips the apprenticeship rather than shortening it. The bad first draft was always the part that was yours to write. Spotting what's wrong with it is the actual skill, and it only develops by doing it yourself, first, on your own work.
Wahi loop hai, bas ab isme AI hai.
This week, before you hand AI a brief and wait for the polished version back, write your own first. Badly, on purpose, mistakes included. Then ask it what's wrong.