System One / Arcade demoModel plays. You watch.

Less talk. More breakout.

Break Out Of Llm Slowness

A ball. A paddle. A model with reflexes.
Typed decisions, fast enough to play.

01 / Meet your player

Your model takes the controls.

A System One decision model reads the game state and chooses where the paddle should go. No generated text. Just a typed answer, and the next move.

8 zones. One choice. Every bounce.
You bring the API key. Jev brings the reflexes.

Default: Jev via Requesty.
Estimated play cost: $0.10–0.20 / hour.

02 / Player settings

API URL

API key

Show

Model API style

Your key is stored in this browser's localStorage and sent only to the URL above, never to this site. Anyone with access to this browser, or any script running on this page, could read it. Use a spending-limited key.

One test decision, then your model takes over.

Powered by typed choices · Played by a model · Watched by you

Blue = zone the model chose. The game never predicts the ball; answers are scored after it reaches the paddle row.

Last answer: zone probabilities (green = where the last ball actually crossed)

Recent decisions