Comment Re: What we made ourselves .... (Score 1) 58
It's also a useful skill to know if the LLM is over-engineering a solution, smuggling hardcoded values as const, not using idiomatic expressions and possibly introducing regressions in the solution. The model doesn't generally have an understanding of the whole codebase and may duplicate existing code, or even hallucinate solutions, so developers still need to be careful with those things unless you work at some YOLO startup.