Polymath Zero

Premise

Polymath Zero is being built by Ihab McShea.

The premise: AI coding tools are fast, but unverified. Work gets marked “done” without evidence, and problems come back. Zero adds structure — cases, sessions, proof, and memory — to make AI-assisted engineering traceable and trustworthy, whether you’re shipping features, closing bugs, or building something new.

Why this exists

AI coding tools ship code fast. They don’t track whether the code actually solved the problem. “It works on my machine” is not proof. The bug comes back six weeks later and you’re starting from scratch.

Zero is structured around what happens after the AI writes code: Does the fix hold? Is there evidence? When the same bug opens again, does anyone remember the last investigation?

The answer today is usually: no, no, and no. Zero changes that by making proof mandatory and memory automatic.

What Zero is and isn’t

Zero is: a local-first engineering tool for tracked, verified AI-assisted Work — bugs, features, refactors, epics, investigations, greenfield projects.

Zero is not:

• ✗
  Not a code review toolZero doesn’t review PRs. It tracks investigations and requires you to review all changes.
• ✗
  Not a CI/CD pipelineZero is a local desktop tool. It does not run in CI.
• ✗
  Not a chat interfaceZero is structured around Work, sessions, and proof — not free-form AI conversation.

Zero is in early access. Every request is reviewed personally. If you’re working with code and find the proof-gate model interesting, reach out.