Show HN: I ran an offline tiny expert runtime on an ESP32-C3
github.comI built this because I think a lot of edge deployments do not actually need a general chatbot. They need something much narrower: small, offline, deterministic, cheap to run, and debuggable when it fails.
The direction here is to take a task-specific reasoning surface and compress it into something that can live in flash and run on a very cheap MCU. Not as a replacement for cloud LLMs, but as a different endpoint entirely: an auditable offline expert for constrained environments.
For me the interesting question is not “can I squeeze a tiny chatbot onto a board?”, but “can useful benchmark-level behavior be crystallized into a deployable embedded runtime?” That is what this project is trying to explore.
No comments yet.