qualm 0.1.1 is out.
It fails a pull request when an experienced reviewer would ask for the change to be simplified, and it tells your coding agent what to fix.
It has no parser, so it judges code in any language: Go, TypeScript, Python, Ruby, whatever your agents write.
Agents write more code than anyone reads. When green merges itself, something has to decide which changes still get human eyes. qualm asks that of every changed file.
The judge is Jev, TypeSafe's decision model. qualm sends it each file's diff with ten typed questions and gets probabilities back. One question gates: would a reviewer push back on this? The rest say why, like a class that grew too big or logic copied from nearby.