There are a few emotional trigger points that LLMs seem to cause in programmers and this is a common one -- the need for deep, first-principles understanding that LLMs make obsolete.
One thing that gets me in a lot of pieces like this is they kind of assume people have no agency, that now that these tools exist we won't be able to help ourselves but use them despite our better judgement.
The broader topic which I don't see discussed so much are values. If you value deep understanding, well you should continue programming and learning in such a way. In some cases you may just want to use a language model to spin up a quick tool or PoC. And there is an entire grey area in between. It's a value judgement to decide what you use LLMs for, as much as what you don't.
> we won't be able to help ourselves but use them despite our better judgement.
Who is we, exactly. Programmers are very rarely the people that are making money in businesses that develop software. In fact they typically represent a massive expense. So in a huge portion of the cases 'we the programmer' will be told what to do in the sense they have to use LLMs to increase their productivity.
When looking at what we tell LLMs to do, you realize there are a lot of cases were humans have less agency than they think.
> There are a few emotional trigger points that LLMs seem to cause in programmers and this is a common one -- the need for deep, first-principles understanding that LLMs make obsolete.
Is it also an "emotional trigger point" that causes people to treat their hunches as facts?
> deep, first-principles understanding that LLMs make obsolete.
I don’t think that’s the case. I agree with the rest of what you wrote. But it’s not a value out of thin air. You need understanding, unless all you ever do is “spin up a quick tool or PoC”. And even then it depends on what you want to quickly use the tool for, or what concept you want to prove.
Idk, as someone who has done LLM driven development of fairly complex things (type systems, memory allocation gymnastics etc) I don't think the need to understand what's going on from first principles has really gone away. If I just want some isolated thing to work I can vibe code with no understanding, but there's no way to get coherence between behaviour, performance characteristics, purity etc without fully understanding the problem space. The LLM just saves (a shitload) of time on grunt work.
Of course if you're building some crud app it's all already tread ground, and you probably can just throw a prompt at an LLM and get something acceptable out.
>Of course if you're building some crud app it's all already tread ground, and you probably can just throw a prompt at an LLM and get something acceptable out.
This is what I think most people who haven't had boring CRUD jobs just don't get - the impact of having some deep technical knowledge goes to waste if all you're struggling with is dumb stuff like bad database design and basic security vulnerabilities everywhere. This was all done by people who are no longer there and were just in it for the paycheck. But also no one who is good is doing these jobs because the pay is too low compared to what they can get.
I'm sure all of this is true if you are teaching at MIT or are working anywhere near people who have gone there though.
Feelings aside[1], a large part of it is about having management above you. Take that plus the ever-present online nagging about productivity. If the latter is true then, well, it’s not like there is a choice.
For people who’s main computing devices are phones, this isn’t hard to believe at all.
Interacting outside of the tech bubble is eye opening. Conversely, the hair stylist might have mentioned the brand of a super popular scissor supplier/other equipment you’d have never heard of.
reply