AI · News

How to Spot a Real Expert in 30 Seconds

Three signals separate genuine domain expertise from confident-sounding noise. None require a credential check.

AI made it cheap to sound expert. The bottleneck moved from can you summarize a topic to can you tell me something the literature won't. Three quick signals separate the two without any credential lookup.

Three signals

Real expertise pays a tax that fakery skips

  1. 01

    They reference failures, not just successes

    Anyone who has actually shipped knows the failure modes. Confident summarizers only know the highlight reel.

  2. 02

    They have boring, specific opinions

    Real experts have weirdly specific takes — a preferred datetime library, a hated naming convention. Fakers stay generic.

  3. 03

    They downgrade their own confidence on the edges

    On core topics they're confident. On the boundary they say so. Fakers are evenly confident everywhere.

Confidence shape is the tell. Real experts have ranges; fakers have monotone certainty.

Evidence

Background

  1. Calibrated confidence (Brier score) tracks expertise more reliably than credential signals across forecast tournaments.

Specifics. Failures. Calibrated edges. That's the whole tell.

Read next

Related