If an LLM can give you feedback on a way to proceed it sounds more like you might be the junior? :P

LLMs seems to be ok'ish at solving trivial boilerplate stuff. 20 attempts deep I have not yet seen it able to even remotely solve anything I have been stuck enough on to have to sit down and think hard.

> If an LLM can give you feedback on a way to proceed it sounds more like you might be the junior? :P

More like it has no grey matter to get in the way of thinking of alternatives. It doesn’t get fixated, or at least, not on the same things as humans. It also doesn’t get tired, which is great when doing morning or late night work and you need a reality check.

Deciding which option is the best one is still a human thing, though I find that those often align too.

Curious: what kind of problem domain to you work on? I use LLMs every day on pretty hard problems and they are always net positive. But I imagine they’re not well trained on material related to your work if you don’t find them useful.

What kind of problem domain do you work on?

> I spent a few hours yesterday pulling it’s massive blobby component apart by hand. But on the plus side, I didn’t have to write the whole thing.

The question, really, is: are you confident that this was better than actually writing the whole thing yourself? Not only in terms of how long it took this one time, but also in terms of how much you learned while doing it.

You accumulate experience when you write code, which is an investment. It makes you better for later. Now if the LLM makes you slightly faster in the short term, but prevents you from acquiring experience, then I would say that it's not a good deal, is it?

Those are the right questions indeed, thank you for being one of the commenters who looks past the niche need of "I want to generate this and move along".

I tried LLMs several times, even started using my phone's timers, and found out that just writing the code I need by hand is quicker and easier on my brain. Proof-reading and looking for correctness in something already written is more brain-intensive.

Asking questions and thinking critically about the answers is how I learn. The information is better structured with LLMs.

Everyone isn't "generating and moving on". There's still a review and iteration part of the process. That's where the learning happens.

I would not mind having a junior type out the code, or so I thought for a while. But in the case of those of us who do deep work it simply turned out that proof-reading and correcting the generated code is more difficult than writing it out in the first place.

What in my comment did you find defensive, btw? I am curious on how does it look from the outside for people that are not exactly of my mind. Not making promises I'll change, but still curious.