Post #2895013
2026-05-05 20:25 UTC
@viraptor@cyberplace.social
Sadly this is what enough people to be a problem do. Output looks great, done.
If they *know* to ask the LLM to review the changes (I've done this too, it's great but yes sometimes brutal), if they *know* they can ask the AI to check for security, accessibility or other issues - then it's a great partner!
I had Claude outright lie to me the other day while trying to troubleshoot and issue though, and that sucked.
@stephaniewalter@front-end.social
Replies (1)
-
@viraptor@cyberplace.social 2026-05-05 20:35
@sarajw@front-end.social Yup, I guess I reacted strongly to "They generate outputs, instead first trying to shape the problem (...)". When it's really "they do what you ask for", we get the opportunity to teach and encourage better questions instead. It works with both people and LLMs. If you want something to improve, you have to explicitly talk about it. @stephaniewalter@front-end.social