I've been thinking that a good way to figure out what documentation to write, to avoid rattling on forever about things that really are obvious and never getting to the things other people need to know (or writing so much uninteresting text that readers don't find the important information in it) would be to have an LLM try (not especially well) to explain the code and then for the human expert to document those things that the LLM got wrong. I've been finding that my reaction to inaccuracies in LLM output about my system has been largely that there's no way a human would know the right answer there, either, but new human developers are much less willing to express what they guess about the system. I think half of the benefit of being interviewed by a good tech writer is that they're willing to say the wrong things that people would think if you didn't improve your documentation.
此内容由惯性聚合(RSS阅读器)自动聚合整理，仅供阅读参考。 原文来自 — 版权归原作者所有。