So you assume code written by the agent canno be trusted? But how comes you are trusting the tests?

At my work we still have a QA dept that runs the manual test suites before each release. We have also automated almost all of those manual tests and they gate every merge to mainline.

Things are still found but normally it is a new test case or an edge condition we don’t have an automated test for. The ever growing e2e suite has reduced the amount of regressions pretty drastically.

EDIT: I forgot to note that automated e2e tests were always the answer to improving quality. AI has just made it cheap to have several layers of test coverage.

To start, you make it a rule that every test must pass with the fix being tested, and it must fail if the fix is missing.

You of course use another AI to test your tests, have you been living under a rock? /s