Kiru
AI systems fail in repeatable, nameable patterns.
Document them with Kiru.
The Project
Kiru treats model failures not as random bugs, but as symptoms of structural conditions. We identify, diagnose, and categorize these behaviors to build a reproducible disorder standard for engineered minds.
The Lab
Kiru Lab builds the practitioner from the ground up — Python, machine learning, deep learning, language models, inverse Jacobian kinematics, and finally mechanistic interpretability. Bio Mirror runs alongside it, holding human brain and body function against its analogues in AI and robotics. Structure dictates function in both.