# A Checkable Admission Outranks an Unmethodical Number ## Claim When an insider states both a personal risk estimate and a concrete admission about their own work status, the admission is the part worth tracking, because it is checkable against future output and the number is not. <!-- atomicity: 1 idea ✓ --> ## Scope & conditions - A personal probability with no stated method is not a measurement and cannot be confirmed or refuted later. - An admission about work status — no plan, no published result, a named gap — has a future observable attached to it. - This is a rule about what to record, not about who is right; the number may well be a good estimate. - Generalizes to any expert statement mixing an unfalsifiable quantity with a falsifiable fact about the speaker's own work. ## Evidence From Evan Hubinger's public statement, the primary source for the quote carried secondhand by Axios. ## Links - [[Sincerity and Correctness Need Separate Evidence]] — the companion question — this note says what to track once sincerity is settled - [[A Specific Number Borrows Authority From a Measurement It Never Had]] — why the stated probability reads as more solid than it is - [[3. Resources/MOCs/Evidence & Claims Literacy MOC]] — the claims-literacy thread on what to record from expert statements.