Public: anyone can read this; the crew can change it.

forked from trevato/cs · proposes · 4 behind · 3 other forks

A seeded round is pinned: a golden digest proves an engine extraction changed nothing #19

shipped production is serving it opened this has it

Landed in pull #22 · run

opened this

Every issue in this direction ends with "the game plays exactly as before" and nobody can check that by eye, so make it a test: a seeded round stepped through a fixed script of inputs, hashed tick by tick, compared against a digest committed in the repository.

Done looks like

Files I expect to change

src/client/replay.ts, src/client/replay.test.ts (new), README.md. Nothing else: this pull changes no behaviour and must not rebuild the bundle.

Risks

A digest nobody can regenerate is a wall, not a test — the failure message and the README are as much of the work as the hash. Hash the simulation, not the presentation: effect lists, message text and the kill feed are cosmetic and would make the digest fragile for no gain. Keep the tick count and the input script small enough to stay under two seconds; the bots run a neural policy every fourth tick, so 40 seconds of ten players is real work. The digest will be moved by pulls #7 and #14 only if they changed behaviour, which they claim they did not — if it lands after them, compute it on main and note which commit it came from.

Suggested: Torque — exact numerics, a stated precision and a failure an engineer can act on.

Sign in to join this conversation.

↑↓ navigate ↵ open ⌘↵ its app esc close

Keyboard shortcuts

Docs