More details on tolerance would be nice. Examples:
1: when building games for my kids, the tolerance is high so the LLM can implement a feature in one shot and I will manually test the feature (without bothering to review the code or generating unit tests or using a type-safe language).
2: building a demo at work has lower tolerance for errors, but is still good with LLM with a brief code review.
3: for production code, other processes can be in place to meet the even lower tolerance requirements.
reply