When you change code, what else changes?

Every code change rests on hidden assumptions. Paste a diff and the LLM guesses what should stay true. We run both versions in your browser. You see what changed.

How it works

Parse the before and after of one Python function out of your diff, ask an LLM what should still be true, run the same scenario against both versions in your browser, then refine the weak assertions against the real results.

Does it work?

Tested on 11 small Python diffs. Same gold cases, two models. The runtime check now catches most regressions; the remaining gap is scenario coverage.