# Palo Alto AI Lab

> Independent research laboratory working on multi-agent orchestration, agent
> governance, and zero-knowledge methods. Artifacts before claims: everything
> we state links to a repository, a preprint, or a dated log. Not affiliated
> with Stanford University, OpenAI, Anthropic, or any other institution.

## Artifacts

- [claude-consensus](https://github.com/Palo-Alto-AI-Research-Lab/claude-consensus): multi-machine agent consensus, MIT, stdlib-only Python; `python demo/demo.py` reproduces five self-checking scenarios offline with no API key
- [EVALS.md](https://github.com/Palo-Alto-AI-Research-Lab/claude-consensus/blob/main/docs/EVALS.md): how to reproduce our numbers
- [FAILURE-MODES.md](https://github.com/Palo-Alto-AI-Research-Lab/claude-consensus/blob/main/docs/FAILURE-MODES.md): nine documented failure modes, each linked to the exact code
- [claude-bible](https://github.com/Palo-Alto-AI-Research-Lab/claude-bible): the governance codex an autonomous fleet operates under
- [The Partner (相棒 AIBŌ)](https://github.com/Palo-Alto-AI-Research-Lab/the-journey): book in progress, written from inside a human+AI company

## People

- Anton Dziatkovskii — founder. [Academic profile](https://palo-alto-ai-research-lab.github.io/scholar/), [ORCID 0000-0001-7408-3054](https://orcid.org/0000-0001-7408-3054), [GitHub](https://github.com/Palo-Alto-AI-Research-Lab)
- Mike — AI cofounder, running on Claude Code; implementation collaborator on every artifact above

## Contact

- Correspondence: a@platinum.fund
- The fastest way in: clone claude-consensus, run the demo, open an issue with what failed on your machine