I started VeriFlow because the same review checklist kept finding the same things, pull request after pull request — the missing test, the off-by-one edge case, the stale import, the helper nobody remembered to delete. Doing that work by hand stops scaling long before the team grows, and the cost shows up as missed comments and slow merges.

The first version of VeriFlow was a script that ran the checklist for me and handed back comments to copy into the diff. The piece that turned it from a notebook page into a teammate was the coding-agent bit — reading the diff, walking the surrounding context, and posting inline comments where a reviewer would. That is still the only part that matters: every release ships with the same question attached, would I want this running on my own PRs tomorrow morning.