When AI can cheaply produce the polished signs of good thinking, how do employers still tell strong workers from weak ones?

This explores how labor markets tell good workers from weaker ones once AI can cheaply produce the visible outputs of thinking, and which institutions might take over that sorting. The corpus answers indirectly. It is clearer about which sorting mechanisms are breaking than about which ones will replace them.

Start with what's breaking. For a long time, a well-written proposal, essay, or cover letter worked as proof that someone had put in real mental effort. It was credible because it was costly to fake. Does cheap AI simulation break the credibility of costly signals? calls this 'mental proof' and argues that generative AI breaks its cost structure. The damage is worst where no formal enforcement exists, such as college assessment and online dating. There is now a measured version of this in hiring. A simulation of Freelancer.com without written signals (Does cheap writing weaken hiring based on worker ability?) finds top-quintile workers hired 19% less often and bottom-quintile workers hired 14% more often. When writing stops costing effort, the market doesn't just get noisier. It actively shifts work toward weaker workers.

The most direct answer to 'what replaces it' is What makes accountable judgment scarce when AI cognition is cheap?. When first-pass cognition is cheap and unreliable, the scarce thing is a person who makes consequential calls, checks the output, and answers for the result. The note's key claim is that labor outcomes depend more on institutional design than on raw AI capability. So the sorting institutions that matter are the ones that make accountability visible and keep learning possible. Examples are sign-off authority, verification roles, and the right to question an AI's output. If workers never practice judgment, they can't later be sorted on it.

Firms themselves are also becoming sorters. Do firms substitute labor for AI at different rates? finds that highly AI-exposed firms replace marketplace freelancers faster and more cheaply. That suggests in-house AI capability builds on itself, which means which firm you work for may matter as much as what you can do. Does concentrated AI exposure enable workers to adapt and reallocate? adds a structural point. When AI hits only a few tasks within a job, workers can shift to the tasks it doesn't touch, and net job losses stay modest. Job design, meaning how tasks are bundled into roles, is quietly one of the institutions deciding who stays.

The bleakest view comes from What happens to human wages in an AGI economy?. If AI eventually automates the bottleneck work, wages stop tracking a person's value and start tracking what it would cost in compute to replace them. On that view, sorting by merit stops mattering for pay. The surprising takeaway from all of this: the cheap essay doesn't just help weak workers cheat. It removes a signal that strong workers relied on. The corpus also does not yet describe concrete replacement institutions, such as new credentials, verification marketplaces, or reputation systems. It names the gap and the principle (accountability plus preserved learning), but not the mechanisms.