Post by @hen@social.lol

2026-02-27 23:52 UTC

Honestly, I&#39;ve found LLMs to be *fantastic* at troubleshooting technical issues I&#39;m dealing with. But I have been thinking to myself, when I fix a new problem with an LLM, it never ends up on the internet to help others. Obviously an issue for other people who may have the same issue, but ironically it&#39;s an issue for the LLMs too. Future problems won&#39;t be solvable with nobody to report solutions to the public. Has this problem been discussed at all? Proposed solutions? Only thing I can think of is try to make it our responsibility to post about solutions we find to issues online.

Replies (3)

@michaelc@scholar.social 2026-02-28 00:07

@hen@social.lol I don't think it's a problem. To the extent that you got a useful response to a prompt is *also* the extent to which the answer was already in the training data. If a statistical text-predictor predicts the most likely word to come next and that *happens* to be the correct answer, that means the answer is probably already out there at a statistically significant frequency. 1/2

Open ##5113086
@NBCactus@mastodon.social 2026-02-28 03:20

@hen@social.lol I find it’s hit or miss. Some things it’s good at, and others it leads me astray. Just today I was trying to fix something and provided log files to it; it suggests it was X, and I asked could it be Y? It said most likely not. I fixed Y and the problem was fixed xD

Open ##5113091
@jnthnkl@mastodon.social 2026-02-28 06:19

@hen@social.lol Reversed StackOverflow ? (:

Open ##5113093