Post by @volpeon@icy.wyvern.rip

@volpeon@icy.wyvern.rip

2026-07-25 09:28 UTC

Looks like I'll get first-hand experience with Claude at work. I can't deny I'm curious what the fuss is about, but also, my boss inadvertently gave me a good demonstration why you need to be even more careful even if you think you are. :neofox_googly_woozy: He implemented a feature with Claude and had it write a documentation for it. I wrote (by hand) code based on this documentation, and he was confused about some of the things I had done. Turns out he let some inconsistencies slip through. Also, while it was technically correct, it presented the information not how I would've needed it and that made it harder for me to tell what I needed to do.

Replies (4)

@senil@gts.social.senil.me 2026-07-25 09:31

@volpeon@icy.wyvern.rip If anything, it really highlights the importance of proper review and ensuring you understand (and appropriately adjust) what it's creating.

Open ##4084680
@Flaky@furry.engineer 2026-07-25 13:12

@volpeon@icy.wyvern.rip it's alright i guess

Open ##5163883
@joepie91@fedi.slightly.tech 2026-07-25 12:48

@volpeon@icy.wyvern.rip Main thing I want to say is beware of the cognitohazard, honestly. I've seen concerning things happening to some people when they "try it out of curiosity", where their entire reasoning about the thing changes and becomes more 'magical' and less rational without them ever realizing/acknowledging it. The whole LLM thing really behaves like a brain exploit.

Open ##5163890
@catraxx@tech.lgbt 2026-07-25 09:44

@volpeon@icy.wyvern.rip Claude, and all AI has a tendency to write bad documentation no one ever needs, presenting a billions details but nothing important.

Open ##5163892