About

What this is

A tutorial in Bayesian thinking — reading beliefs as bets, weighing evidence, and updating with the arithmetic Thomas Bayes left in a drawer and Richard Price published in 1763. Six modules, each built around instruments you operate rather than figures you look at.

The method

Every core concept on this site is introduced by a panel interaction first and prose second. You count the dots before anyone says “posterior”; you multiply clues before the formula appears; the notation arrives in Module IV as a label for something your hands already know. Each panel carries a “Try this” prompt, and every number a panel displays is computed live from the site’s own simulation code — there are no canned screenshots, so you can check any claim by dragging a slider. A panel you can’t manipulate is, by this site’s own rules, a defect.

Made with AI, openly

This site was written and built by Claude, Anthropic’s AI model, working with Jörn Dinkla — the prose, the simulation code, the design system, all of it. We say this plainly because openness about AI authorship is the point, and because the site’s method makes the claim checkable in the healthiest way: the panels compute their numbers in front of you, and the worked examples use round, author-chosen teaching values rather than borrowed empirical claims. Where an idea has a named origin, the name is in the text.

Further reading

• Thomas Bayes & Richard Price, An Essay towards solving a Problem in the Doctrine of Chances (1763) — the posthumous paper that started it, and this site’s namesake.
• Gerd Gigerenzer, Calculated Risks (2002) — why natural frequencies fix what percentages break; the spirit behind Module II’s grid.
• Daniel Kahneman, Thinking, Fast and Slow (2011) — base-rate neglect and the taxicab problem, from one of its discoverers.
• Sharon Bertsch McGrayne, The Theory That Would Not Die (2011) — two centuries of the theorem’s history, exile included.
• Philip Tetlock & Dan Gardner, Superforecasting (2015) — calibration and updating as a competitive sport.
• E.T. Jaynes, Probability Theory: The Logic of Science (2003) — the deep end, for when the wading pool stops being enough.

Colophon

Design system: Doctrine of Chances — Baskervville (a revival of John Baskerville’s 1750s Birmingham type, the letterforms of Bayes’ own decade), Spectral for reading, IBM Plex Mono for everything instrument-shaped. Built with Astro and React islands; the simulation logic is plain TypeScript with unit tests.