Hallucinations in code are the least dangerous form of LLM mistakes

Another danger is spotted in the later paragraphs:

> I genuinely find myself picking libraries that have been around for a while partly because that way it’s much more likely that LLMs will be able to use them.

People will pick solutions that have a lot of training data, rather than the best solution.

That's exactly why I stopped using Svelte. Claude is much more sensible when generating React. Looks like a bleak future where only the most popular library survives.