I was fascinated by the possibilities that a System One Model like Jev can open, so I decided to recreate it from scratch within a fully embeddable runtime.

Performance is extremely good (less than 200 microseconds per decision), and memory is never allocated while scoring.

I created a self-running demo of Snake, Doom, and Autopilot that runs fully in a browser thanks to the WASM version.

I plan to add more experiments in the upcoming days, especially how a Blink-like decision system can fit into SQLite or Postgres.

Any feedback or participation to the project is more than welcome.

Repo: https://github.com/sqliteai/blink