p(doom), the number everyone throws around

there is a number floating through every conversation about AI. p(doom). the probability that this all ends badly. people say it like a joke, or a shrug, or a confession. "my p(doom) is 10%." "mine is 50." nobody ever says where the number comes from.

so i built a calculator for it.

the p-doom quiz asks you about the things the number actually depends on: how fast capabilities arrive, how hard alignment is, whether regulation happens, who builds it first. your answers feed a small bayesian network, and it outputs your p(doom) for 2035, 2040, 2050, 2060, and 2100. next.js, typescript, chart.js, no ML libraries. the probability math is hand-written, which is exactly why i trust it enough to show it.

it even shows confidence intervals, because a single number about the end of the world is a lie with good formatting.

then it does the part i care about most. it puts your number next to the real ones: what AI researchers and industry people have actually said, with their reasoning attached. you can click on any expert and read why they believe what they believe.

the point was never to predict anything. it was to notice what i actually believe. when you have to answer "how hard is alignment, really" with a slider instead of a feeling, the dread decomposes into positions. positions you can argue with. that is most of what thinking about AI risk should feel like, and almost none of what it feels like...

the theme got a little silly, i admit. flame effects. ember particles. a doom meter that burns in real time while you answer. (when you build a tool about the end of the world, you are allowed some theater.)

it has 1 star. you can try it at p-doom-quiz.vercel.app, and it embeds in any page with an iframe, which felt important. a number like this should be easy to pass around.

building it changed how i hear the number. when someone says "my p(doom) is 30%" now, i want to ask: 30% of what, exactly? which timeline? which assumption is doing the work? usually the number is a mood wearing math clothes.

sometimes the mood deserves the clothes. but at least now you can check the fit.