AI strategies · Resaca

One AI call writes a strategy (plan + description); every simulation after that replays it as pure data. Tweak the rules, regenerate from an edited description, and duel strategies against the built-in bots.

Game

Generate a strategy

Steering description (optional — e.g. "aggressive rush", "avoid buying points until late")

Model (blank = gpt-5.6-luna → gpt-5.6-terra)

Strategies

None yet — generate one above.

Duel

Seat 1 vs seat 2 (10 games; run twice with sides swapped for a fair read). Results land in the playtest report too.

Seat 1

Seat 2 — pick "(same)" to mirror seat 1