The steering envelope · survival theorem · instrument v0.4

Policy on top. Dinner at the bottom.

The Survival Theorem: uncontrolled acceleration crashes, forced deceleration crashes later and harder, and what survives is controlled speed matched to grip. Pick a worldview or drag the dials. The world band races the hazard model (same equations as the Python engine); the kitchen band shows the only number that survives contact with a household: hours of work per week for food, shelter and energy, per tribe. Discover for yourself what the proof table then states. New here? Read the axiom and the theorem first.

Policy — what do you accelerate first?
World — 400 Monte Carlo races against the field
P(good) P(crash+pileup) arrival tick field wrecked anyway
WHAT THE SIX STATES MEAN (each has a historical face)
ARRIVEDyou cross the transition first, intact, and set its terms. Britain and the railway age.
CONVOYthe pack arrives together because knowledge was shared or standardized; the gains without the wreck. Rail gauge standardization, open protocols, FOSS. Once everyone knows the secret of fire, nobody has an advantage, and everybody has fire.
SHOTGUNsomeone else crosses first and you live in the world they chose. Tokugawa Japan and Perry's black ships.
HELDnobody crosses; the transition is refused, and you pay in everything it would have bought.
CRASHEDa corner taken beyond your control authority. Chernobyl.
PILEUPthe coupled crash: everyone leaning on everyone, everyone down together. The Late Bronze Age; 2008 with balance sheets instead of bronze.
Kitchen — hours per week of essentials, six decades, eight tribes (ghost lines: the other worldviews; red line: the 60h regime switch where dinner leaves the market)
Proof — which worldview minimizes each tribe's worst decade?

The honesty toggle re-runs the whole table with every worldview's coordination forced to zero. Watch what happens to the s/acc column: its wins come from the K dial, not from private virtue. force K = 0

Same constants as steering_envelope/model.py and society.py (calibration pinned by the pytest suite). Baselines are toy calibrations for ordinal realism, not measurements; the empirical claims live on the essay page with the validation suite.