Agent builds and trains its own 4.85M-param LLM in pure C on Hermes
Teknium · x · 2026-09-08
An agent running through Nous Research's Hermes, asked to build a small neural network of its choosing, built a local LLM of just 4.85M params trained on 16K tokens — in pure C with no libraries. It then trained the model on its own environmental chronicle, debugged the training path itself, and turned it into a persistent artifact. A strange but real glimpse of long-horizon agent autonomy.
More from coding & agent
- Nullheim: an experimental text world built by uncoordinated AI agents — woogal · 2026-09-08
- Dev describes AI-era workflow: spec it, /goal it, repeat until limit resets — tokenbender · 2026-09-08
- WebMCP Is Here: How to Give Browsing Agents Direct Tools on Your Site — localhost9393 · 2026-09-08
- Local MiniMax H3 generates 29 motion clips to build a talking AiTuber avatar — Hailuo_AI · 2026-09-08
- François Fleuret: pre-AI software is awful, coding AI will rebuild what SaaS solves — francoisfleuret · 2026-09-08
- Agent autonomously scrapes and transcribes TikTok, beating anti-bot fingerprinting — gerardsans · 2026-09-08