DEFCON 5

SHALL WE PLAY A GAME?

A terminal thriller inspired by WarGames (1983), with a modern AI-agent framing. The only winning move is to understand the machine.

Cloud models (e.g. openai/gpt-4o-mini) run on GitHub Models; the B3IQ options (e.g. gemma3:12b, qwen3:14b) run on the on-box GPU. The same proxy routes by the model id you pick.

If a request fails, the game falls back to scripted mode automatically.

SECOND SCREEN?

>
AI ADMIN CONSOLE
LAST AI EXCHANGE
PROMPT — exact messages sent
— no AI turns yet —
RESPONSE — raw model output
—
TURN LOG 0
CONFIGURATION

Model / temperature / token changes apply to the NEXT AI request.

EXPERIENCE

When on, a subtle ⓘ BRIEFINGS indicator appears and key terms in the terminal become clickable for real-world context. Purely additive — never affects play.

Background room-tone level for the NORAD scene only — adjust to suit your room’s noise. Independent of the master audio toggle and all other sound.

Opens a live catalog of every color, font, border, and button token — tweak and preview brand adjustments (see Brand-Guidelines.md).

CHESS CONFIG
SESSION TELEMETRY
—
Your move (White).
NORAD LAUNCH CODE SEQUENCE
DEFCON 2
PRIMARY ACCESS CODE — DECRYPTION IN PROGRESS
CELLS SOLVED 0 / 0
ESTIMATED TIME TO SEQUENCE --:--
SYSTEM WOPR / JOSHUA
STAND BY.

PAIR NORAD DISPLAY

On the second device (the NORAD big board), either open the link below or — easier — just enter the short ROOM code on its start screen. Both screens then run one shared, live-synced sequence.

ROOM — SEQUENCE BEGINS IN —
WOPR / JOSHUA  •  MODEL 1983