Comment by kajaktum
17 hours ago
I am rebuilding numba. It is very hard for me to imagine doing it by hand. I tried it a couple of years ago but it was excruitiangly painful. It was slow and messy. So many small things that gets stacked on top of each other over years of abstraction.
I am doing it again using LLM. Legitimately, things that would have taken weeks is now done overnight. I still have to look at the code, at the generated C output, still have control over the architecture to make it easy for me and the LLM to work with in the future, etc
Is this replacing my thinking? I am not sure. I suppose I would have learnt a lot more about compilers/transpilers had I preserver through it for months with manual writes and rewrites but I would solely be working on this. Instead, I also had some time to write a custom NFS server support for a custom filesystem in Golang.
> Is this replacing my thinking?
I'm extremely confident the answer is yes.
But we have to judge how much value that particular thinking has.
As an instructor, I've implemented linked list functionality a zillion times. I'm on the long tail of skills-gain from each reimplementation. But every time I implement it, I'm gaining a little more.
Now, is it worth it? Probably not. The time spent on that marginal gain would be better spent implementing something more novel by hand. So punting to an LLM, while it costs me, might be a net gain in that case. But implementing another compiler? Hell yeah, that would be replacing my thinking. I've only ever made one PL/0 compiler plus that one yacc thing in compiler theory class, and those were a long time ago.
We should quantify the loss of thinking when we decide how much to punt the code creation to someone or something else.
I too worry about the aspects that using AI is replacing in my thought process. I've built a sophisticated enough system to where agents can go out and determine the changes that need to be made for entire features and pretty much nail it out of the box. Everything is laid out in high detail during the planning phase. The implementation phase of actually writing the code is almost always unremarkable.
I have found myself going out and actually reading code less and less over the past year. I would be lying if I said that there are not fairly regular moments where I question the comfort level I have obtained with the system that I have built. I've seen it work with such a high accuracy and success rate so many times that my instinct at this point is to not question it. I keep waiting for this to really bite me in the ass somehow, but it just keeps not happening. Sure, there have been minor issues that have slipped through the cracks that caused me to backtrack, but that is nothing new. The difference is that with the previous way, I had painstakingly written that code and had a much more personal relationship with it. The code was the problem. Now whenever that does happen, I'm going back to the system and figuring out why it didn't get the answer right on its own, or why it didn't surface the whole thing in the plan to me prior to implementation.