BLT-Leaf: A Small, Human Speed Bump for the AI Era

#ai #webdev #programming #code

Pull requests feel different lately. They’re bigger. Faster. Polished in that slightly suspicious way. The description is immaculate, the code compiles, CI is green — and yet something feels unfinished. A comment from three days ago is still hanging there. A reviewer asked a real question and never got a real answer. But it has approvals, so… merge?

That’s the environment BLT-Leaf was built for. Created under the umbrella of OWASP and maintained through OWASP-BLT, it’s not trying to be another flashy dashboard or productivity hack. It’s a quiet, slightly stubborn tool that asks a simple question: is this actually ready to merge? Not “does it pass CI?” Not “did someone click approve?” But is it done in the way humans mean when they say done?

What makes it refreshing is what it pays attention to. Unresolved conversations. Stale feedback. Whether the author responded. Whether the discussion loop actually closed. It turns those signals into a readiness score, and sometimes that score is lower than your ego would prefer. Good. That tiny sting is the point. In a world where code can be generated in seconds and reviews can drift into autopilot, BLT-Leaf reintroduces a little friction — the healthy kind.

This isn’t an anti-AI manifesto. AI isn’t the villain. The problem is merging things that look finished but aren’t. The real risk is speed without reflection. BLT-Leaf doesn’t lecture or block you dramatically; it just makes it harder to ignore the loose ends. It’s a small speed bump in front of the merge button, and sometimes that’s all you need.

We don’t need more hype around our pull requests. We need more honesty. And sometimes honesty looks like a tool quietly saying, “You still have three open threads. Maybe fix those first.”

Note: This post was edited with AI — which feels fitting, honestly.