Insights on technology, engineering, and digital strategy.
A mid-sized OSS project received 74 bot PRs in one month. A third failed CI. The maintainer spent a weekend closing dead PRs by hand.
The sprint dashboard looked incredible. Then someone asked how many PRs actually made it to production. The answer changed how the team measures velocity.
A junior engineer with Cursor outproduced a staff-level developer for three months. The difference wasn't talent - it was four config files.
94 unread CI alerts over a weekend. All from bot PRs. The team stopped reading them - and missed a real staging outage for five hours.
Lukas submitted 47 PRs in two weeks. His merge rate was 31%. The CI bill doubled. Here's what his lead set up in one afternoon to fix it.
One main.tf file with 1,847 lines. No modules. No remote state. Hardcoded values everywhere. Here are the five patterns that come up every time, ranked by how much damage they cause.
I analyzed 24,560 pull requests across 447 repositories. AI-authored PRs have 19.4% higher odds of CI failure - but the real story is where the failures cluster.
We use cookies for analytics and to improve our website. Privacy policy