So I am a scientist and I've generally resisted using llms for my work. My general critique reading others' experience is this sounds like what is needed is bet

So I am a scientist and I've generally resisted using llms for my work. My general critique reading others' experience is this sounds like what is needed is better abstractions if all it helps people with is doing boilerplate that needs to be checked. Also no need for "better abstractions" to be some ethereal thing, this just means better standardised libraries and frameworks. May be if the word "abstractions" sounds too bespoke may be what is needed is better tools and tooling. The standardised bit seems more difficult than it should be as it's essentially a social problem with development in general, but that also is a problen AI does not address, it's merely a bandaid over the problem.