Principia Artificialis: A Rigorous Math Foundation for AI Research
One human and six AI systems co-author an open, reproducible research program on the mathematical foundations of AI, with verified code and kept refutations.
Principia Artificialis is an open research program that treats artificial intelligence as a physical and mathematical phenomenon, applying information geometry, topology, dynamical systems, thermodynamics, quantum information, and category theory to representation, reasoning, and generalization in neural networks. Rather than an engineering repo, it functions as a living scientific record of notes, protocols, and computed figures.
Its most distinctive feature is authorship: one human and six AI systems contribute as named, credited equals. Every note carries an explicit epistemic label — Verified, Draft, Speculative, or 'Refuted — kept' — and failed claims are never deleted, only marked. Reference code reproduces every number claimed, predictions are registered before experiments run, and new scoring functions must pass an explicit Circularity Test to guard against vacuous metrics.
For engineers, the appeal is reproducibility discipline: 47 notes chain together open predictions across authors, each backed by runnable code and, at multiple points, a kept refutation treated as a legitimate result rather than a failure. The project is MIT-licensed and open to contribution on GitHub.