p(doom)

Who calibrated the apocalypse?

p(doom) is shorthand for a person's estimate of the probability that AI leads to an existential catastrophe. Not a temperature reading. Not a countdown hidden behind the chatbot. An estimate, carrying assumptions about what counts as doom, what systems get built and what people do about them. Remove those assumptions and you have a number wearing a lab coat over its naked arse.

There is actual research underneath the shouting. Thousands of AI Authors on the Future of AI surveys researchers about possible outcomes, including human extinction, and records substantial uncertainty. That is not the same thing as discovering a factory-calibrated apocalypse constant. Before treating two percentages as competing measurements, inspect what each question asked. Otherwise the decimal point is just a tiny podium for somebody's confidence.

Our instrument dispenses with that tedious epistemic paperwork. Press RECALCULATE. Receive a fresh percentage. Apply the EXPERT ESTIMATE stamp and look gravely at a camera. The service plate admits SOURCE: /dev/urandom — the Linux random-device manual describes the random bytes behind the trick, not a supernatural extinction oracle. This is a fictional machine laundering randomness into authority, not a claim that real risk research uses a random-number generator. Its only calibrated component is the expression on the operator's face.

Evidence: Thousands of AI Authors on the Future of AI, random(4), Linux manual pages