Maith: A Disciplined Framework for AI-Assisted Math Research

Maith is an open-source research workspace that treats AI as an idea generator, not an oracle, for tackling unsolved math problems like the Riemann Hypothesis, P vs NP, and the Jacobian Conjecture. It organizes each open problem into a structured "dossier" with files for precise statements, exploratory ideas, executable experiments, and lemma-by-lemma proofs, explicitly separating unproven conjectures from independently verified theorems.

The project's core discipline is reproducibility: every computational claim must ship with exact source code, input data, version information, and output hashes so results can be independently rerun. Numerical evidence over a large range is never accepted as proof, and any claim overturning a famous open problem is expected to survive adversarial review from a second AI, a computer algebra system, or a human expert.

For engineers, Maith offers a concrete template for using LLMs and CAS tools in high-stakes technical work: use AI to search for counterexamples and map proof dependencies, but require independent verification, exact reproducibility, and formal checking (e.g., Lean or Coq) before trusting any result.