pDoom

You report to the Department of Future Continuity & Other Administrative Surprises, where BASIL — the Bureaucratic Assessment System for Intelligence & Liability — must confirm that you are human. It opens with an impossible CAPTCHA (select every square containing a mind), moves through a handful of brisk experiments, and eventually admits your species was never the interesting question. The whole time it has been estimating your pDoom: your own probability that advanced AI ends humanity, or ends our say in it, by 2100.

You sort claims into what changes your mind, what needs checking, and what is merely decorative panic. You spread a hundred tokens across five ways 2100 could go. In one quiet room where the jokes stop, you set three conditional dials and watch them multiply into a number. At the end you get a specific pDoom, a visible band of uncertainty around it, and seven dimensions — AI literacy, evidence hygiene, how much you center humans, and more — that explain the number instead of grading it.

The point isn't the number; it's the assumptions underneath it. pDoom refuses to call a low score naïve or a high score wise, and it never treats speed or identity as intelligence. It hands you a model of how you reason under deep uncertainty — and, in the Footnote, some unauthorized compassion. Play again as New Game+ and watch which of your beliefs actually moved.

Opt in and your result — a coarse bucket, never your answers — joins an anonymous tally, so you can see where you fall among everyone who's played. Stay private and nothing leaves your device; you still get to see the crowd.

A content note: this one asks you to put a real number on human extinction and the permanent loss of human control. It treats existential risk as a probability to reason about, plainly and with sources — never as a spectacle or a scare, and it never predicts the end of anything. Your answers stay on your device.