What is p(doom)?

The notation borrows from probability, which is part of why it travels: a number looks like a finding. What the record can verify about any p(doom) is who said it, in what role, on what date, in what forum, over what timeframe, and on what stated basis. The Doom Number records exactly those five things for every statement and computes no average, because a decade and a century, extinction and loss of control, a survey median and a single researcher's guess are not the same object.

The 2026 discussion produced a case study in how a number moves. A figure of greater than ten percent within a decade was stated by Anthropic's alignment science lead after a colleague's resignation post that carried no number at all; within days the figure was being attributed to the colleague, and a Senate letter cited it as the view of three researchers. Settled or Not grades that attribution; the Doom Number records the statement under the name of the person who made it.

How to read one: as a fact about a person's belief, weighted by their role and their stated reasons, and never as a forecast the page endorses.