Proof-of-skill.

"5 years experience as a line cook" tells you almost nothing about a candidate. "Busy restaurant, fast-paced kitchen" tells you almost nothing about a job. Both sides describe in vague generalities, then act surprised when day one doesn't match expectations.

The premium on description over demonstration is an accident of medium. Resumes are text, so we describe. Job posts are text, so we describe. The whole hiring stack evolved to evaluate paper credentials because paper was the cheapest format. Phone cameras have changed that. The cost of showing the work — on both sides — is now lower than the cost of describing it well.

A 10-factor hire is built on demonstration. Candidates show the cuts, the plating, the fix, the install. Employers show the kitchen, the station, the shop floor, the team. Both sides walk into day one knowing what the other actually looks like.

There's a detail to proof-of-skill that's becoming critical as AI improves: the person doing the work has to be visible. A clip of hands chopping vegetables, a finished plate, a completed haircut from behind — these can be AI-generated, recorded by someone else, or pulled from the internet. A continuous shot of the candidate's face and the work they're producing is much harder to fake. As AI-generated content gets better, the proof bar has to get higher. Show the person doing the work, not just the work.

The pushback is that demonstration takes effort, and that requiring the face on camera is intrusive. Both are fair concerns. The cost of recording a 60-second clip is far less than the cost of a bad hire. And the face-on-camera requirement is what separates real proof from AI-generated fakes — without it, the whole signal collapses in an age where anything can be synthesized.