Ask HN: SWEs how do you future-proof your career in light of LLMs? • Hacker News Board

Re: Ask HN: SWEs how do you future-proof your career in light of LLMs?

#661

Post by natemwilson » Tue, Dec 17, 2024, 7:28 AM UTC

janalsncm wrote: post #263

thegrim33 wrote: post #71

I don't worry about it, because: 1) I believe we need true AGI to replace developers. 2) I don't believe LLMs are currently AGI or that if we just feed them more compute during training that they'll magically become AGI. 3) Even if we did invent AGI soon and replace developers, I wouldn't even really care, because the invention of AGI would be such an insanely impactful, world changing, event that who knows what the…

To replace all developers, we need AGI yes. To replace many developers? No. If one developer can do the same work as 5 could previously, unless the amount of work expands then 4 developers are going to be looking for a job. Therefore, unless you for some reason believe you will be in the shrinking portion that cannot be replaced I think the question deserves more attention than “nothing”.

It’s my belief that humanity has an effectively infinite capacity for software and code. We can always recursively explore deeper complexity.

Re: Ask HN: SWEs how do you future-proof your career in light of LLMs?

#662

Post by karmasimida » Tue, Dec 17, 2024, 7:41 AM UTC

Future at this point is ... unpredictable. It is unsatisfactory answer but it is true.

So what does it take for LLM to replace SWE?

1. It needs to get better, much better 2. It needs to be cheaper still

Those two things are at odds with each other. If Scaling Laws is the god we preaching to, then it apparently has already hit the diminishing of return, maybe if we scale up 1000x we can get AGI, but that won't be economically reasonable for a long time.

Back to reality, what does it mean to survive in a market assuming coding assistants are going to get marginally better over say next 5 years? Just use them, they are genuinely useful tools to accomplish really boring and mundane stuff. Things like writing docker files, those will go away to LLM and human won't be able and don't have to compete. They are also great second thoughts advice givers, it is fun to know what LLM thought of your design proposal and build upon their advice.

Overall, I don't think much will change over night, the industry might experience contraction in terms of how many developers it will hire, for which I think for a long time, the demand will not be there. For people already in the industry, as long as you keep learning, it is probably going to be fine, well, for now.

Re: Ask HN: SWEs how do you future-proof your career in light of LLMs?

#663

Post by GeoAtreides » Tue, Dec 17, 2024, 7:41 AM UTC

pdimitar wrote: post #458

richardw wrote: post #364

Earlier quoted context omitted.

For those less confident: U.S. (and German) automakers were absolutely sure that the Japanese would never be able to touch them. Then Koreans. Now Chinese. Now there are tariffs and more coming to save jobs. Betting against AI (or increasing automation, really) is a bet against not against robots, but against human ingenuity. Humans are the ones making progress, and we can work with toothpicks as levers. LLM's are ou…

In other words, you have built exactly zero commercial-grade applications that us the working programmers work on building every day. LLMs are good for playing with stuff, yes, and that has been implied by your parent commenter as well I think. But when you have to scale the work, then the code has to be easy to read, easy to extend, easy to test, have extensive test coverage, have proper dependency injection / be ea…

> But when you have to scale the work, then the code has to be easy to read, easy to extend, easy to test, have extensive test coverage, have proper dependency injection / be easy to mock the 3rd party dependencies, be easy to configure so it can be deployed in every cloud provider

the code doesn't have to be anything like that, it only has to do one thing and one thing only: ship

Re: Ask HN: SWEs how do you future-proof your career in light of LLMs?

#664

Post by Fizzadar » Tue, Dec 17, 2024, 7:42 AM UTC

Not (currently) worried. Lots of comments here about intellisense being a similar step up and I’ve had it disabled for years (I find it more a distraction than use). So far LLMs feel like a more capable - but more error prone - intellisense. Unless it’s stamping out boiler plate or very menial tasks I have yet to find a use for LLMs that doesn’t take up more time fixing it than just writing it in the first place.

Who knows though in 10 years time I imagine things will be radically different and I intend to periodically use the latest AI assistance so I keep up, even if it’s a world I don’t necessarily want. Part of why I love to code is the craft and AI generated code loses that for me.

I do, however, feel really lucky to be at the senior end of things now. Because I think junior roles are seriously at risk. Lots of the corrections needed for LLMs seem to be the same kind of errors new grads make.

The problem is - what happens when all us seniors are dead/retired and there’s no juniors because they got wiped out by AI.

Re: Ask HN: SWEs how do you future-proof your career in light of LLMs?

#665

Post by barrell » Tue, Dec 17, 2024, 7:42 AM UTC

Animats wrote: post #656

barrell wrote: post #625

Build something with an LLM outside of your comfort zone. I was a pretty early adopter to an LLM based workflow. The more I used it, the worse my project became, and the more I learned myself. It didn’t take long for my abilities to surpass the LLM, and for the past year my usage of LLMs has been dropping dramatically. These days I spend more time in docs than in a chat conversation. When chatGPT was announced, many…

> Build something with an LLM outside of your comfort zone. I need to try that. How do LLMs perform writing shaders? I need to modify a shader to add simple order-independent transparency to avoid a depth sort. Are LLMS up to that job? This is a known area, with code samples available, so an LLM might be able to do it. I'm not a shader expert, and if I can avoid the months needed to become one, that's a win.

I'm sure LLMs can do it, and I'm sure they will be riddled with mistakes! Especially if you're not competent enough in an area to grok docs (I don't mean that in a condescending way — I would definitely not be competent enough to grok shader docs) then LLMs can be a great way to learn.

It won't take long before you see all the mistakes in their responses before trying them, and from there it's just a hop and a skip to it being more efficient to just reading the docs then arguing with an LLM.

But that beginning part is where LLMs can provide the most value to programmers. Especially if you go in with the mindset that 90% of what it says will be wrong

Re: Ask HN: SWEs how do you future-proof your career in light of LLMs?

#666

Post by baq » Tue, Dec 17, 2024, 7:44 AM UTC

bee_rider wrote: post #638

ido wrote: post #630

Earlier quoted context omitted.

I'm only 41 but that's long enough to have also seen this happen a few times (got my first job as a professional developer at age 18). I've also dabbled in using copilot and chatgpt and I find at most they're a boost to an experienced developer- they're not enough to make a novice a replacement for a senior.

I think the concern is that they might become good enough for a senior to not need a novice. At that point where to the seniors come from?

We’re here already. Nobody wants to hire junior devs anymore. Seniors are still a couple phone calls away from new positions.

Re: Ask HN: SWEs how do you future-proof your career in light of LLMs?

#667

Post by IshKebab » Tue, Dec 17, 2024, 7:45 AM UTC

fragmede wrote: post #400

IshKebab wrote: post #277

Earlier quoted context omitted.

> It's provably true that LLM's can generate working code. Yeah for simple examples, especially in web dev. As soon as you step outside those bounds they make mistakes all the time. As I said, they're still useful because roughly correct but buggy code is often quite helpful when you're programming. But there's zero chance you can just say "write me an driver for the nRF905 using Embassy and embedded-hal" and get som…

The question is, how long would it take you to get as far as this chat does when starting from scratch? https://chatgpt.com/share/6760c3b3-bae8-8009-8744-c25d5602bf...

Yeah now actually check the nRF905 documentation. You'll find it has basically made everything up.

Re: Ask HN: SWEs how do you future-proof your career in light of LLMs?

#668

Post by fabianholzer » Tue, Dec 17, 2024, 7:45 AM UTC

tokioyoyo wrote: post #361

pockmarked19 wrote: post #299

Earlier quoted context omitted.

Seems like they caught up because I have posted this before including in chatGPT. All that means is you have to change it up slightly. Unfortunately “change it up slightly” is not good enough for people to do anything with, and anything more specific just trains the LLM eventually so it stops proving the point. I cannot load this link though.

I don't want to sound like a prick, but this is generally what people mean when they say "things are moving very fast due to amount of investments". 6 months old beliefs can be scrapped due to new models, research and etc.

Where "etc." includes lots of engineers who are polishing up the frontend to the model which translates text to tokens and vice versa, so that well-known demonstrations of the lack of reasoning in the model cannot be surfaced as easily as before? What reason is there to believe that the generated output, while often impressive, is based on something that resembles understanding, something that can be relied on?

Re: Ask HN: SWEs how do you future-proof your career in light of LLMs?

#669

Post by pandemic_region » Tue, Dec 17, 2024, 7:51 AM UTC

mianos wrote: post #615

I am 61, I have been a full time developer since I was about 19. I have lost count of the number of 'next thing to replace developers' many many times. many of them showed promise. Many of them continue to be developed. Frameworks with higher and higher levels of abstraction. I see LLMs as the next higher level of abstraction. Does this mean it will replace me? At the moment the output is so flawed for anything but t…

Hey man, thanks for posting this. There's about ten years between us. I too hope that they will need to pry a keyboard from my cold dead hands.

Re: Ask HN: SWEs how do you future-proof your career in light of LLMs?

#670

Post by NorthTheRock » Tue, Dec 17, 2024, 7:52 AM UTC

GeoAtreides wrote: post #663

pdimitar wrote: post #458

Earlier quoted context omitted.

In other words, you have built exactly zero commercial-grade applications that us the working programmers work on building every day. LLMs are good for playing with stuff, yes, and that has been implied by your parent commenter as well I think. But when you have to scale the work, then the code has to be easy to read, easy to extend, easy to test, have extensive test coverage, have proper dependency injection / be ea…

> But when you have to scale the work, then the code has to be easy to read, easy to extend, easy to test, have extensive test coverage, have proper dependency injection / be easy to mock the 3rd party dependencies, be easy to configure so it can be deployed in every cloud provider the code doesn't have to be anything like that, it only has to do one thing and one thing only: ship

But in a world of subscription based services, it has to ship more than once. And as soon as that's a requirement, all of the above applies, and the LLM model breaks down.