Build AI agents from first principles using a local LLM - no frameworks, no cloud APIs, no hidden reasoning.
Terminal score: 0–1000 raw, weighted across 4 dimensions. Public score: 0–10 normalized (shown in the 30-day stars chart above).