AI that fixes bugs and opens tested PRs.

Example run, ticket LIN-482. The film starts on its outcome, an open pull request with review requested, in the Atlantic app window on the ocean. The window shrinks into the Atlantic app icon, a pointer double clicks the icon and the window opens on the run list, where a new run arrives at the top. The pointer opens that run, and the film plays the run that led to the pull request. On a wide screen a camera moves in close on the ticket, the thesis, the code change, the checkout form and the review; on a phone the window fills the frame and the camera moves in once, at the end. Step one shows the ticket and a support thread about the same problem. In step two three finders read the ticket, the thread and the code without seeing each other's work, and the judge types out one thesis with its reason. Step three confirms the problem by reading the code again at its current head. Step four writes the fix on its own branch in the web-app repository and shows the change line by line. In step five a preview boots, a browser agent enters a VAT ID at checkout and places an order, a confirmation email is sent and the change works. In step six a second model reviews the change and finds one blocker, which is fixed in the first review cycle, and the review ends with no blockers. In step seven the pull request opens with the thesis, the evidence and the QA verdict, the camera moves in on it and pulls back, and the film ends where it began.