AI safety nonprofit · San Francisco
Open-source, agentic tools that accelerate the work of AI safety researchers.
The long-term goal is to fully automate safety research and defensive systems — so safety can keep pace with, and eventually outpace, AI capabilities.
What we build
We build open-source, agentic tooling that gives AI safety researchers more leverage — taking on the repetitive, brittle parts of the work, and keeping results trustworthy as models grow more capable.
More on specific tools soon.
Who's building this
Neolithic was founded by Leo McKee-Reid. His path ran from engineering drones and orbital rockets at Rocket Lab, to ML research in computational neuroscience, to AI safety — working on scheming and deception evaluations at LASR and interpretability at Leap Labs — before starting Neolithic. The team is small and growing.