GeePawHill : "If you have so much tedious boilerplate that an LLM can reduce your workload, the problem is that your design and code require lots of tedious boilerplate, and that is not a problem you will fix without learning to program better." (via TheSocials)
Now, of course I agree whole heartedly with GeePaw, but would like to add that I think our tooling could also use some significant improving. Tooling like our ProgrammingLanguages, build tools, the way we assemble chunks of code, the interfaces to data,...) Why does all that stuff demand that we write so much boilerplate in the first place? Surely we could do better?
I think things have been... not stagnant on that front, for sure, but not half as innovative, adventurous or powerful as they might be. And the reason for that lack of real, powerful, high-leverage progress is that the industry has been drowning in cash for the past 30 years. Credit for which point also goes to GeePaw. Consequently, programmers everywhere have been kept very busy by the Money Machine guys at churning out code — much of it boilerplate or stuff that could, without too much trouble, be automatically generated — and getting very well paid for their efforts. There's been no incentive to improve the tools with which we make code; no constraints or pain that would encourage us to invent more productive ways to work. (And LLMs are not such a thing. Don't even get me started.)