GitHub · Games and demos · Agents and tools
wdobry/laya-playground
Seeing the model work before committing to a 2.3 GB download.
What it does
A local playground for Laya: a landing page where the model plays Flappy, a lane runner and Tetris at roughly 30 decisions a second, an editor where you write your own typed questions, and a benchmark that compares all three checkpoints.
The repository's own description: “A website, two games, a benchmark and an agent skill for Laya, the open-source decision model. Runs on your machine.”
How it works
A standard-library Python server serves the pages and holds the model in process, so every move in the games is a real forward pass and every answer in the editor is a live probability rather than a recording.
The repository, by the numbers
Read from the GitHub API on 2026-09-24. Stars and the last push move daily — quote them with the date, the way we do.
What we checked
- the README, read 2026-09-24
- the 2.3 GB download and roughly 90-second load time it warns about
- the note that the public site replays recordings instead of running the model
What we did not check. How well any of the games are played. No row of this repository claims a score or a win rate, which is why the note above quotes the decision rate and not a result.