MISSION — pdoomcount.org

What this is at maturity

A public website that maintains a live, clinical estimate of the probability that AI causes a terminal catastrophe for humanity — an honest number with its reasoning trail, updated on a schedule and moved only by documented events.

The register is actuarial, not apocalyptic: an instrument that reads like an insurance risk table or a lab readout, not a doom blog. No countdown theatrics, no black backgrounds, no monetized fear.

The founding thesis (operator, 2026-09-15)

The most plausible route to AI-caused catastrophe is probably not a superintelligence erupting in a lab — it is the mundane deployment path: capitalism pushing not-ready AI into sensitive systems (grid, military, health, finance, logistics) to save money, faster than verification, liability, or regulation can follow. The methodology must track that route as a first-class gauge, not a footnote.

Non-negotiables

1. "An index of judgments, not a measurement" — stated on the front page. The number aggregates published forecasts and conditions them through a documented rubric; it is not observed data.
2. Every number is attributable. Full source ledger, per-version archived calculations, every movement tied to logged events with rationale.
3. Uncertainty is always displayed. A point value with no band is dishonest on this subject.
4. Falsifiability is symmetric. "What would move this number down" sits beside the number itself.
5. No bombast. Clinical language, plain numbers, sourced claims.
6. The mundane-deployment path is a named gauge (see methodology).

Audience

Curious, serious general readers; people in tech and policy who want one honest number with receipts instead of vibes. Not a doomer subculture product.

Out of scope

Doom-mongering, clickbait, ads, advocacy for any single ideology or lab, prediction-market gambling products.