logoalt Hacker News

hansmayerlast Saturday at 10:17 PM1 replyview on HN

Is it really just a personal anecdote ? Please do read some other comments on this post. The snark comes from everyone and their mother recommending "just write CLAUDE.md", when it is clear that this technology does not have intrinsic capability to perform reliable outputs based on human language input.


Replies

blitz_skulllast Saturday at 10:21 PM

Yeah… that’s the point of LLMs: variable output. If you’re using them for 100% consistent output, you’re using the wrong tool.

show 1 reply