Open-source non-autoregressive System 1 decision engine laya passes 30k stars

On October 3, 2026, DEV Community's AI section published two reports back to back on laya, an open-source decision engine. Both describe laya as a non-autoregressive "System 1" decision engine aimed at typed choice tasks. The project has reached 30,286 stars on GitHub, and both pieces frame it as free to use, self-hosted, and able to run on a developer's own infrastructure, so data stays local and there are no API costs to pay.

The two reports agree that laya is optimized for speed and meant for fast, lightweight decision making at high concurrency, and that it removes the dependency on paid APIs while keeping data private. Where they diverge is on how good it actually is. One describes its performance as competitive, with capabilities similar to paid tools; the other calls it industry-leading, with top-tier capabilities. Both were published at the same moment, and neither supplies benchmark scores, evaluation numbers, or named comparison points, so there is no way to tell which characterization is closer to the truth.

As of these two reports, the story stops at laya having crossed 30,286 GitHub stars and being presented to developers as an open-source non-autoregressive decision engine. Neither piece mentions when the project launched, who maintains it, what license it uses, any specific performance figures, or any later version releases or community activity.