[j16sdiz]

Sometime a wrong and misleading README done more harm than good.

It is not that rare to see LLM waste hours on a wrong path because a misleading line in README. Even worse, they can't learn. Spawn a subagent and it repeat the same error again

[al_borland]

Someone on my team started using an LLM to write all his readme files (he used to not write them at all).

So far, 100% of them have been wrong. I read them, my spidey senses think that what it says doesn’t match his style. I look at the code to find the variables the readme mentions doesn’t exist anywhere in the codebase. I then reach out to him about it, where it says it was written by AI and he will go back and write it for real.

He says it’s better than nothing, but agree with you that it does more harm than good. I wasted my time reading slop. I wasted more time validating the slop. I wasted even more time with a conversation about it. Now he’s spending time re-writing something that he could have written faster and better when he was actually writing the code and it was fresh in his mind. Meanwhile, I’m either blocked waiting for him, or I need to spend my time trying to understand the minutiae of his code so I can integrate it into mine.