I'm kinda tired of obsession for one-shot tests in coding, there are good tests for multi-step debugging with analyzing output/images/videos?

Personally, i think good coding model shouldn't be focused on one-shot "everything in one html-file" tests, but should be really good on debugging, fixing and modifying its own output. Anyone know such simple tests that i would able to run with local models?

May be some kind of synthetic stuff that forces LLM to build something that is broken by design and then asking the model to do a multi-step changes, fix issues, analyze program's output (preferably with getting screenshots/videos)?