New world — choose your assumptions
How does this run? (no, it isn’t asking an AI)
Nothing intelligent runs while you play. No Claude, no Codex, no model of any kind is queried — there are no network calls at all. Each run is one Monte-Carlo rollout of a 43-technology dependency graph, executed as a few hundred lines of ordinary JavaScript in your browser (view-source to see all of it).
The algorithm, in one breath: every technology has prerequisite groups (all groups must be met; any member satisfies a group — that is how substitutes like “steam or waterpower” work). Once its prerequisites exist, a technology arrives after a lag calibrated so that the no-assumptions world reproduces real history, jittered by a log-normal draw (your “turbulence” dial is its σ), and shortened by any programs your assumptions create. Randomness comes from a seeded generator (mulberry32 + the Box–Muller transform), so the seed on the end card reproduces a world exactly. The “replay history exactly” setting bypasses the dice entirely and uses the raw historical lags, so — with every other assumption at our world’s setting — every technology lands on its real date and the satellite flies in 1957 — the other settings answer “how else could it have gone?”, that one confirms the machine knows how it did go. The chronicle sentences are pre-written templates keyed to each technology — not generated text.
Where AI does enter: the model itself — the node list, the edges, the calibration procedure, this page — was written by Claude (Fable 5) working with the prompter, and the full model with equations, calibration details, and honest caveats is documented in chapter 4, Replaying the Tape (the graph itself: timelines_data.js). The genre is cliodynamics — history treated as a dynamical system — at deliberately toy scale. A theory primer — what cliodynamics is, and how this family of sims uses it — lives in the Turning World’s footer.