@AdamGolds@Steve_Yegge you will be stuck in code review hell quite quickly though if you dont automate that. youre right that self-grading is not robust, but you can use zeroshot to define your own adversarial multi-agent graph for this.
@bcardarella Disagree, code quality is the ability to execute realistic maintainability and extension tasks with the statistically minimal amount of cognitive overhead (or nowadays, tokens)