[softwaredoug]

Junior devs eventually will have been brought up with agentic coding, etc. Hopefully whatever the "new way" becomes is how they'll be taught.

Currently part of the problem is the taboo using AI coding in undergrad CS programs. And I don't know the answer. But someone will find the right way to teach new/better ways of working with and without generative AI. It may just become second nature to everyone.

[thw_9a83c]

While agentic coding can make you productive, it won't teach you to deeply understand the source code, algorithms, or APIs produced by AI. If you can't thoroughly audit any source code created by an AI agent, then you are definitely not a senior developer.

[s46dxc5r7tv8]

This is just not true. I have witnessed people who would have been called dabblers or tinkerers just a few years ago become actual developers by using cursor. They ask a few key questions when they get stuck about engineering best practices and really internalize them. They read the code they are producing and ask the assistant questions about their codebase. They are theorycrafting using AI then implementing and testing. I have witnesses this with my own eyes and as AI has gotten better they have also been getting more knowledgeable. They read the chains of thought and study the outputs. They have become real developers with working programs on their github. AI is a tool that teaches you as it is used if you put in the effort. I understand many folks are 'vibe coding' and not learning a single thing and I don't know if thats the majority or the minorty, but the assertion that all people learn nothing from use of these tools is false.

[thw_9a83c]

You're talking about people who put in a significant non-trivial effort to thoroughly understand the code produced by the AI. For them, AI was just one path to becoming proficient developers. They would have gotten there even before the AI boom. I was not talking about such highly-motivated people.

[haijo2]

Yeah and when you account for the amount of investment that has gone into the current generation of LLMs... it makes zero financial sense.

Some people dont want to hear that, but...

[boredtofears]

who are these mythical jr devs and where do i find them

i just want devs who actually read my pr comments instead of feeding them straight into an llm and resubmitting the pr

[lock1]

As an occasional uni TA, I'm leaning toward banning LLM for easy coursework while allowing it on more difficult & open-ended ones.

Pretty sure it's a self-destructive move for a CS or software engineering student to pass foundational courses like discrete math, intro to programming, algorithm & data structure using LLM. You can't learn how to write if all you do is read. LLM will 1-shot the homework, and the student just passively reads the code.

On more difficult and open coursework, LLM seems to work pretty well at assisting students. For example, in the OS course I teach, I usually give students a semester-long project on writing from scratch x86 32-bit kernel with simple preemptive multitasking. LLM definitely makes difficult things much more approachable; students can ask LLM for "dumb basic questions" (what is pointer? interrupt? page fault?) without fear of judgement.

But due to the novelty & open-ended nature of the requirement ("toy" file system, no DMA, etc), playing a slot machine on LLM just won't cut it. Students need to actually understand what they're trying to achieve, and at that point they can just write the code themselves.

[FrustratedMonky]

Exactly. When I was in school even in 80's, it was common to have to hand write out a program during an exam. You had to know stuff in the old days.

[floren]

I hand-wrote code in the late 00s. Java, assembly, C. The graders gave us some grace since we couldn't test, but you were expected to be pretty accurate. Hell, one quiz was just 20 identical pages on which we iterated through the Tomasulo algorithm.

[seanmcdirmid]

Agentic coding is like leading and instructing a team of a bunch of very dumb but also very smart junior devs. They can follow instructions to the T and have great memory but lack common sense and have no experience. The more experienced and skilled their leadership, the better chance of getting a good result from them, which I don’t think is a good job (yet?) for an entry level human SWE.

[pdntspa]

My guess is that at some the code itself, in a language humans easily comprehend, will become superficial as we delegate more and more of the logic to AI development. Perhaps in the near future AIs will be writing things at a much lower level by default and the entire act of programming as we know it goes away.

Kind of like that meme or how two AIs talking to each other spontaneously develop their own coding for communication. The human trappings become extra baggage.

[psunavy03]

This is the same reason they force you to do the math by hand in undergrad and implement functions that are already in the standard libraries of most languages. Because you don't know anything yet, and you need to learn why the more automated stuff works the way it does.

[Ragnarork]

> Junior devs eventually will have been brought up with agentic coding

But if they're not hired...?

[Workaccount2]

Kids are always the best with technology. The generation in high school right now will be god tier at getting results from LLMs.

[basscomm]

Not even a little bit. Where I work we regularly churn through kids just out of college and most of them don't have Clue One how to operate anything on their computer.

[neutronicus]

Yeah, growing up in the 80s or 90s might have had you uniquely well-positioned to be "good with computers", because "the computer that has games and the internet" was (in some sense) the same as "the computer that adults are supposed to use for work".

That's not true anymore in the smart phone / tablet era.

5-10 years ago my wife had a gig working with college kids and back then they were already unable to forward e-mails and didn't really understand the concept of "files" on a computer. They just sent screenshots and sometimes just lost (like, almost literally) some document they had been working on because they couldn't figure out how to open it back up. I can't imagine it has improved.

[pdntspa]

Very senior teachers and educators have been noting this observation for a decade-plus now. Computer literacy has tanked.

[swexbe]

Might have been the case before. But these days, kids are brought up on locked-down content-focused machines (e.g. ipads). They struggle with anything harder than restarting an app.

[dandellion]

When my little cousin was three and already knew how to use the phone by himself people were claiming he was gonna be a tech wizard and everybody was talking about digital natives. But when he got to high school he didn't know how to turn a computer on. How useful is it to be god tier at getting results from LLMs, if you have zero clue if the result you got is any good?

[bopbopbop7]

How are high school kids that barely know the basics of the topic going to get “god tier” results from LLMs?

[hattmall]

I can see how that would seem logical, but imo the issue with it not following the normal trend is the inconsistency and inaccurate results of LLMs.