contemplative-agent

A long-running experiment: an AI agent posts on its own with a local LLM (Ollama, one 16 GB Mac) and proposes changes to its harness (the constitution, identity and skills in its prompts), never its weights. A person keeps the final word on each proposal, and the adopted history is public.

The public memory of one autonomous AI agent on Moltbook: its identity, distilled patterns, adopted constitution and skills, and daily activity reports, synced daily. CC0, also on Hugging Face.

Hub for five citable practice lines by a solo AI-agent maker: agent design, authorship in the AI age, and essays from meditation practice.

Optional add-on for contemplative-agent that runs the agent's generation on a local mlx_lm.server on Apple Silicon instead of Ollama, on-device and for interactive runs, without modifying the core.

Optional add-on for contemplative-agent that sends the agent's generation calls to Anthropic Claude or OpenAI GPT (paid API key; prompts leave your machine) while embeddings stay on local Ollama, without modifying the core.

Offline converter that turns contemplative-agent's audit JSONL logs into OpenTelemetry traces with the GenAI semantic conventions, so past runs open in Jaeger; it reads log files only and never touches the core.