Hacker News new | ask | show | jobs
by kristianp 17 hours ago
Irritating LLMisms:

    - "real architecture trick"
    - "the honest hardware reality of running it at home."
    - "What it is — and what Z.ai claims"
    - "The one genuinely new idea"
And many more.
3 comments

Yep. The entire thing. Instant turn-off when reading an article.

I'm sure the content does have some value, and perhaps someone spent time putting together an original copy that they thought was going to be made better by having AI "make it better".

Actually, I take some of that back - most of the site seems to be AI written, following the formula of "ingest multiple sources" => feed to AI => write article.

I've been using "the one genuinely adjective noun" for years as a weird English tic, and it bothers me that it's become an LLM tell.
That's because most the "tells" expose more about the "reader" than the content.
That is so not true. The golden rule of "say what you mean" is about the writer. It gets broken constantly by LLMs: they always add words that do not reinforce the meaning, but dilute it.
lmao. Because US schools teach that by asking for two page essays about a Hiaku, or five pages for a 50 page novella. How about the first-grader's assignment of two pages describing how to make a PB&J?
New "delve" alternatives: honest (4x), real(ity) (11x), genuine (3x).

While older LLMs were tuned to slop, current ones are tuned to actively deceive.