We built a replacement for AWS DynamoDB, a key-value database for fast web content fetches. This was done with two engineers and hundreds of persistent Computer agents over two months. Migrating to our in-house database will save us up to a hundred million dollars yearly.
It iteratively kept improving in each turn when stuck and finally got the job done, one shot. No second prompt needed. Pretty impressive for a mobile browser.
It worked for 158 steps (longest I've ever seen it work) - started off by inspecting the failure logs in GitHub actions, navigated to the appropriate file in the repo, made the necessary code changes, and raised a PR for me to review