System 1 Playground

UCSC AI Hackathon · Local AI

Fast decisions,
on your own hardware.

A System 1 model does not write text. It reads a situation and a list of choices, and returns a probability for each choice in one pass. Jev made the idea popular, but it is closed. We train our own small open models, run them on an ASUS GX10, and race them against Jev in real games.

One pass
State + options in, probabilities out. No tokens to generate.
Same seed
Every lane plays the exact same pipes and cactuses.
Local
Our models answer from a GX10 on the desk, not a data center.