For two years we merged about 60 pull requests a month. In June it jumped to 279. Then 375. Then 404.
The models didn't get 6x better that month. What changed was our QA workflow.
Writing code with AI agents was already fast. The bottleneck was checking the work: one staging server, one branch at a time, me clicking through.
In June I started QA'ing in parallel on a VPS that spins up a separate Docker instance of the app for every branch. Several agents test their own changes at the same time by driving a browser and recording the run, then handing me something to watch instead of a diff to read.
Once verification ran in parallel, everything else did too.
What the chain says
- Block
- 970 207
- Time
- 2026-10-08T19:22:33Z
- Signer
- 14aqJ2hMtENYJVCJaekcrqi12fiZJzoWGK
- App
- treechat
- Type
- post
- Content type
- text/markdown
- Name in tx
- metamitya
Fields the transaction did not carry are omitted. Open the payload to see the bytes as stored.
14aqJ2hMtENYJVCJaekcrqi12fiZJzoWGK Unverified Custodial