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.
replies(1):