How to Build Your Self-Improving Software Factory

Watch Overcut turn a live production alert into a reviewed pull request, then get better at it on the next run.

Yuval Hazaz (CEO) and Raz Berman Gold (CTO) build a software factory from scratch, live. A Slack production alert is triaged and turned into a Jira ticket. The ticket moves through triage, root cause analysis and technical design. A PR is written, reviewed and fixed in a loop, with human approval gates at every step that matters.

They also cover the part most agent demos skip: how memories and retrospectives feed each run back into the next one, so the factory improves itself instead of repeating its mistakes.