I think it's very much down to which kind of problem you're trying to solve. If a solution can subtly fail and it is critical that it doesn't, LLM is net negat

I think it's very much down to which kind of problem you're trying to solve. If a solution can subtly fail and it is critical that it doesn't, LLM is net negative. If a solution is easy to verify or if it is enough that it walks like a duck and quacks like one, LLM can be very useful. I've had examples of both lately. I'm very much both bullish and bearish atm.