Post #1973991

I think the thing about ascendant #LLM usage that I have the hardest time connecting with is the amount of self-congratulation I see from people who... asked an LLM to do something for them. When I was working on software systems professionally, there was often a clear and sincere pride expressed—whether when building something new or fixing a tricky bug—in having developed a new understanding of the underlying problem. Yes, elegance in solutions was valued, but more importantly I would see people get this instinctive spark from having acquired a bit of enlightenment, however niche. My partner is the primary developer in the house now, so my experiences are more vicarious. However I get the sense, from overhearing her meetings and from our conversations, that the satisfaction stops at being able to prompt the machine to make something that approximates the prompter's understanding of the spec. In fact, some of her biggest frustrations are with developers who explicitly _don't_ want to understand the problem at hand, in favor of "velocity". I know that my observations are partly second-hand, and I'm only getting the high- and lowlights, so I don't assert that that impression is The Global Truth. But I'm pretty sure there's a nugget of truth to it, and to the extent there is, it really saddens me. Developing mental models of the inner workings of complex systems was the biggest thrill of software development to me. Yes, code had to ship, and yes it had to work. But for me, that started with the mental model. With experimentation. With getting it wrong and figuring out why before getting it really right. I wonder how foreign this experience is and will be to current and future developers, and I can't help but find it a bit weird when someone seems to solicit attaboys and participation ribbons for asking the robots nicely. #ai