Reviewing an agent’s work inside the session it came from, then handing the comments straight back to the agent.

Start a session from anywhere, pick your repos, and let Polygraph provision each one into its own worktree.

Why teams keep reaching for a layer above their agent, and what that layer actually has to do.

Picking up somebody else’s Codex session in a different agent, reviewing it, and pushing back to the same PR.

Cheaper paper gave Renaissance artists room to experiment. AI agents could do the same for software, if developers keep enough sympathy to judge what agents produce.

The full walkthrough: meta-harness framing, repository graph, session graph, resume, and code review.

Agentic memory means many different things. A few core attributes help classify what a product actually does, and show why code is not like other domains.

Agents hit two walls: they are stuck in one repo, and they start every session blank. Polygraph removes both.

Launch day conversation about what Polygraph is, why it exists, and where it goes next.

The launch listing, with the discussion thread from launch day.

Coverage of the launch and the synthetic monorepo idea behind it.

A short introduction to the product and the problem it solves.

One minute on what Polygraph does.

The framework wars taught us that the winning layer often does less. The same pattern is emerging for AI agents.

Breaking down what a software factory actually requires, and which parts are real today.

Connecting separate repositories into one dependency graph without moving any code.