I find this interesting for two reasons. 1. LLMs inspire and clearly strike a certain type of chord in people that other technology does not. For instance, can

I find this interesting for two reasons. 1. LLMs inspire and clearly strike a certain type of chord in people that other technology does not. For instance, can you imagine a post called "Rust Is Not Fun" at the top of HN? Or even replace "Rust" with technology that has some fans and some haters, like "PHP Is Not Fun". Can you ever imagine that finding equivalent traction? Why would you even write a post called "Skateboarding Is Not Fun"? I just did a search across HN and the only other thing I saw being called not fun that actually got traction was... Twitter. 2. The post makes two points about why LLMs are wrong, and I (as someone who gets a lot of mileage out of LLMs) pretty strongly disagree with both. a) You can't get better at using LLMs ("Nurturing the personal growth of an LLM is an obvious waste of time") This seems almost objectively false to me? I have gotten substantially better at prompting and using LLMs after about 6 months of daily use. I think at a higher level of abstraction with LLMs than I would when working directly with code, and this is a different type of thinking that requires effort and practice to develop. An LLM doesn't solve all problems you could ever have, it just allows you to think at a higher level of abstraction. Previously I might convert one 300 LoC file from JS to TS (or whatever) and call it a day. Now, in the same amount of time, I might do 10. But obviously just asking the LLM to commit that doesn't cut it because I might have broken something, so I need to think of some way to get the LLM to verify that I haven't broken anything, so maybe I'd first get it to build some unit tests to my specifications, or a couple of linter rules, or something else tailored to the problem at hand. This is the "thinking at a higher level of abstraction" and I tend to find it an interesting puzzle. It's a very different type of thinking then the "how am I going to refactor this single file", but it ends up being pretty enjoyable. b) "For me, the joy of programming is understanding a problem in full depth" I suspect that this is why a lot of people are frustrated with LLMs. I empathize here more than with part a). I mean, sure, I like digging into complex systems - it's fun and rewarding. But... and I feel this will be controversial, but I feel like this isn't really the most meaningful part of coding. Isn't the "complex problem solving" mostly a side-effect of the thing which is really the important thing (and the thing I like much more), which is delivering software that people enjoy and use? And while LLMs do a good bit of heavy lifting on the first, I don't find that they are capable at all of solving the second - which means that coding is just as fun for me as ever. In fact, probably more so, since I feel like I can iterate and build ideas faster for users.