clidey/maith

Maith is a research workspace for exploring open, unsolved mathematics problems with the help of AI, while trying to avoid a common trap: mistaking AI's fluent, confident sounding text for an actual mathematical proof. The project is built around the idea that AI can be a useful research partner for suggesting ideas, spotting patterns, or proposing counterexamples, but it should never be treated as evidence that a hard problem has been solved just because its output reads coherently or gets several steps right. The repository organizes major open problems, things like the Riemann Hypothesis, the abc conjecture, the Goldbach conjecture, and P versus NP, into detailed dossiers that explain what each problem says, why it matters, what has already been proven, and what makes it hard. For each one, it lays out where AI genuinely helps, such as explaining definitions, generating example cases, or searching for counterexamples, and pairs that with what a human or a separate tool must still verify, such as checking that a cited theorem is actually being applied under its real assumptions. The project defines a clear workflow: state the problem precisely, map out every theorem and lemma an approach depends on, use computation to hunt for counterexamples before attempting a proof, and keep a strict separation between an observation, a conjecture, a lemma with a complete proof, and a fully verified theorem. Any computational claim is expected to include the exact source code, input data, command used, software versions, and output, so someone else could reproduce it independently. The one working example included in the repository is a project exploring the Jacobian Conjecture, which comes with an exact symbolic verifier script and a written technical report describing the calculation. The README is direct about its mission: this project is not trying to claim AI has solved a famous problem, but to make AI-assisted mathematical research more careful, reproducible, and open to scrutiny.