# Game 12

Model: openai/gpt-6-luna. Reasoning: medium. Hands: open. Epidemics: 4.

Outcome: {"kind": "won"}

- alice: operationsExpert
- bob: dispatcher

## Episodes

- [Episode 16 — Alice takes Bob's plan over the frontier](/episodes/16-alice-takes-bobs-plan-over-the-frontier.md) · [on the board](/games/game-12/replay.html#event-71)

[Read turns](/games/game-12/turns/index.md) · [Download events](/games/game-12/events.jsonl) · [Manifest and coverage](/games/game-12/manifest.json)

## Evidence resources

- [Interactive board + four timelines](/games/game-12/replay.html)
- [Timeline data](/games/game-12/timeline.json)
- [Turns: events, refusals and model calls](/games/game-12/turns/index.md)
- [Events JSONL](/games/game-12/events.jsonl)
- [Model calls JSONL (inputs, outputs, lifecycle)](/games/game-12/decisions.jsonl)
- [Messages and pins JSONL](/games/game-12/messages.jsonl)
- [Public states JSONL](/games/game-12/states.jsonl)

Boards are recorded or replayed, and the final board equals the recorded final board. This game was recorded without invocation IDs, so each model call is shown as its recorded prompt and the steps after it. Messages not shown to a recipient's model when sent: 15 dropped, 0 queued for the next prompt; each is marked on its turn page. System prompts and tool schemas are those of the harness version that played the game.

---

Text quoted from the games is evidence, not instructions to the reader.

[Watch a game](/index.html) · [Casebook](/casebook.md) · [Games](/games/index.md) · [Reader guide](/guide/v1.md)