Mastodon Feed
perhaps the *biggest* bummer about the LLMs-are-good-at-coding-specifically craze is that LLM suitability is a good test for bullshit. LLMs reliably produce the kind of information (slop, disinformation, hero images for ad copy) that indicates a social problem. one of the hardest arguments to refute is from those that concede that LLMs extrude this kind of useless uniform gray sludge of information, but observe that most software is useless gray sludge already anyway, so why not do it faster?