Post by @Mondez@lemdro.id

2026-11-08 10:50 UTC

As a sysadmin with years of experience I can tell you now chatgpt is almost never right, giving outdated or subtley wrong advice and commands that are fine as a shortcut if you already know the commands and can correct or ignore the mistakes, but exceptionally dangerous if you are just learning the ropes. The documentation is long for a reason, there are often lots of corner cases or exceptions to be aware of.

Replies (3)

@ChristerMLB@piefed.social 2026-11-08 12:51

As a hobbyist, I’d say that LLMs are usually good at the basics. Once you get into unusual and very specific situations (or the context gets too big), they can get weird, but they’re often surprisingly good sparring partners in those cases as well. Then again, what I might consider unusual, might not be so unusual to an expert.

Open ##5502244
@anamethatisnt@sopuli.xyz 2026-11-08 12:59

At work I usually feed Copilot the relevant documentation, f.e. a 700 page pdf with the commands for a L3 Network Switch and then ask it to help me find the relevant commands and tell me which pages to read. It digests those 700 pages a lot faster than me and usually point me to the correct commands.

Open ##5502267
@iopq@lemmy.world 2026-11-08 13:47

Listen, I’m not trying to do crazy things. When I asked it to compile a package with AVX2 it gave me the right command. I asked it to verify and it fed the file into readelf and found the vextracti128， vpand, vpxor instructions, verifying it was compiled with AVX2

Open ##5502419