I think it’s really context dependent. I haven’t found LLMs to increase my productivity in coding in my field because the quality of the output matters much more than the quantity. I don’t think it’s the same across the board though, and there are plenty of domains where code generation is a force multiplier. Sometimes you need a chainsaw and sometimes you need a scalpel and in my own experience I have found that using coding agents as scalpels is not a very efficient use of my time. shrug