This is an example of a productized self-improvement loop on code review in Warp Factories. You can see here that our observer agents are scanning PR comments for patterns and suggesting changes to our factory PR skills that we then review, merge, and improve.

We're moving our agents to the cloud to improve quality and manage cost. To do this, we needed a way to configure environments, harnesses, and security permissions as code. This is configuration format we landed on:

Contrary Venture Partner applications are now open! If you're a builder on campus, interested in investing in exceptional startups, and want to be part of one of the most selective communities in tech, apply now. Contrary invests and works directly alongside exceptional founders to build enduring companies, from seed to scale. With this approach, we’ve been fortunate to back iconic companies like Ramp, Anduril, Zepto, and many others. Our Venture Partners are great builders and future founders who become the go-to resource for others on campus. As a VP, you’ll learn first-hand what it takes to build and invest in exceptional companies, from taking part in investment decisions to sitting in on fireside chats with seasoned founders and investors. Best of all, you’ll be joining and learning from the best community in tech. Most VPs go on to become venture-backed founders, early engineers/operators, or investors, and enjoy the support of Contrary’s network for life. Read more here: Apply here by Sept 13th:

We have a number of scorers running against our internal Warp factory, one of which grades the verbosity of agent conversations. I made a change earlier to make it stricter because our agents are creating overly verbose PR descriptions. This caused the scorer pass rate to drop. The scorer run details show how it detected one particularly bad PR description. This is a simple example of how we can detect our code review skill needs tuning.

What's it like to work at Warp? - Product ownership in every role - High agency, figuratively and literally - Tight collaboration, in-person or remote Join us:

This is our main internal Factory that our engineers have been using to work across 8+ repos. A couple highlights here from the last week alone: > 1000+ opened PRs (over a third have merged) > 5000+ total runs > ~6 hr cycle time The factory is still improving, but it's incredible to see how rapidly things are coming along.

Measure factory performance with cost and velocity metrics, accessible from the factory dashboard and via our API and SDK. Plus, configure self-improvement agents to grade performance and automatically improve your factory overtime.