3 ms·

I’ve been a nay sayer up till two weeks ago when I actually sat down and coded a non trivial feature with AI and I was blown away. The problem was one that ther

by le-mark 8mo ago

I’ve been a nay sayer up till two weeks ago when I actually sat down and coded a non trivial feature with AI and I was blown away. The problem was one that there may have been at most 5 open source versions that they had been trained on. Today I am planning an exit strategy. Anyone else considering what they’ll do in a post AI software engineering world?

uejfiweun 8mo ago
Yeah, it's pretty wild, huh? I'm at one of the big tech companies and the strategy that I've fallen into is to basically just mostly coast, save / invest as much as possible, and hope that this gets me to a financial threshold of "permanently comfortable" before the job losses kick in. As for what happens after that, I'd really prefer not to have to do physical labor or trades. And it doesn't seem like any other white collar occupation is really going to be insulated, other than perhaps medical. So my strategy is to basically wait and see what society looks like after the transition and I guess I'll try and decide on something then?
aabhay 8mo ago
I would emphasize self reliance and sustainable living. Things like home ownership, dollar-resistant assets like gold, and self education in topics that matter to you and your existence. Another way to frame it is what would you do in a low trust environment where corporations and the government were not to be trusted. You would likely avoid things like bubble bursting AI stock investments, jostling for rank in a company, etc.
- le-mark 8mo ago
  If post AI software engineering is a thing then this is NOT a bubble.
almostdeadguy 8mo ago
How do you plan an exit strategy for something that may or may not obsolete a whole field in a matter of months? Not sure there's a real way to do such a thing.
andrewmcwatters 8mo ago
[dead]
lifeisstillgood 8mo ago
We used to call the job Analyst Programmer. I am not sure what thing your code did but I am pretty sure you needed it - but who could understand that gap existed ? Who could explain to an AI that it needed to create this obscure code to solve that problem. And now comes the hard part - persuading your organisation to adopt it. AI can code - but can it understand what it missing from the organisation and persuade it to chnage - to spend years at industry conferences? Look at Starliner. NASA just announced that Boeing stuffed up, not with an engineering mistake (no one still knows exactly what broke) but that the whole organisation is so screwed up and so political Nasa just don’t believe Boeing can fix it. AI cannot fix our turf wars. That’s not intelligence (humans know going to war is bad, but Putin still exists). I ye the systems we live in, and work in. Changing those is feasible - once they are coded, transparent and open To inspection in a democracy. We need programmable introspective systems of organisation - democracies in other words. The engineering was not the problem - the problem was the organisation was more or less toxic and incapable of doing engineering. Writing code that won’t get used because if politics is a job we and AI can both do.
blindhippo 8mo ago
I've been using these tools for nearly a year and half on a daily basis. They've become an integral part of my tool box for solving problems. But writing code was never much more than 35-40% of my job while working for companies/others. Most my time has always gone towards communication, design, and validation. All three of those are not particularly vulnerable to mass AI automation except for the most trivial of scenarios and I have not seen evidence that has changed in over 2 years of so called "improvements". My "exit plan" ultimately is to be one of the engineers capable of using these tools to scale my impact accordingly so I can focus on higher order problem solving, which ultimately is what is most valuable. I would be more concerned if I was in marketing/sales or frankly middle management. Maybe this is just "copium" on my part, who knows, this sector is moving fast.
- mekael 8mo ago
  I was discussing this with one of my counterparts today, and we both agreed that writing the code is the easy part, and actually knowing what to do / getting requirements sorted out / working with our compliance team are the hard parts. That being said, I'm in a pretty highly regulated industry so it just might be that.
  - df2dd 8mo ago
    Envisioning what should exist is always the hardest part. If LLMs provide substantial material to be able to produce what one envisions faster, that is great. But LLMs will not be doing the envisioning. Most humans already are poor at that. Hence why there are very few real 'visionaries' in history. Envisioning always requires deep thinking. If LLMs eat away at a humans ability to sit and think, this will make envisioning solutions harder. So you'll see more stuff produced, but largely more crap.
drivebyhooting 8mo ago
I have no workable exit strategy because I have small children and our family has been continuously under layoffs and terminations. If the doom really comes to pass then what future is there for us? I fear a life of impecunious servitude and poverty more than death.
- jvanderbot 8mo ago
  Regardless of the future you have to plan for the worst and hope for the best. I don't have time to post significantly about it but I'd love to trade thoughts and figure this out. Email?
- alecco 8mo ago
  Our ancestors survived worse. Wait out for the rainbow after the storm. Adjust and take it easy.
  - Gud 8mo ago
    Uhm, our ancestors haven’t faced anything like this. This is the only time in the history of mankind, where you have a near global ruling elite trying to replace labor totally with mechanised thinking machines. The closest we can compare to is the off shoring in the 80s/90s.
  - AlexeyBelov 8mo ago
    My ancestors lived in abject poverty.
jryan49 8mo ago
Save up enough money to be safe for many years after I get laid off so I have plenty of time to figure out what to do.
gedy 8mo ago
Writing the code has never been what has held companies back. A lot of companies will use the speed of AI to wallpaper over the fact that they don't know what to make or how to prioritize.
- df2dd 8mo ago
  When I read comments here, Im beginning to realise many people get lost in the noise and cant seem to figure out what exactly is the true bottle-neck of producing great software that people (be it consumers or folks employed at firms) purchase/use. Writing code faster alone doesn't change a great deal. Frankly it'll just create a larger influx of noise. Focus is very difficult to do, it'll become harder in the advent of LLMs.
  - le-mark 8mo ago
    LLM change the dynamic so an individual or small team can replicate the work large companies. Especially if they worked on that large application before.
    - df2dd 8mo ago
      .... another simpleton take. Ive heard enough of these boring one-two liners.
OptionOfT 8mo ago
I'm on the other side. I work a lot with these tools and find that time after time there are subtle bugs, and lack of structure. And that's fine, I use it as a tool, but am I faster? Not where detail matters, and I tend to do a lot of work in that area. Generating dashboards? Absolutely. Converting to JavaScript vNext UI framework? Yes. Kernel code? No.

by le-mark 8mo ago

uejfiweun 8mo ago

Yeah, it's pretty wild, huh? I'm at one of the big tech companies and the strategy that I've fallen into is to basically just mostly coast, save / invest as much as possible, and hope that this gets me to a financial threshold of "permanently comfortable" before the job losses kick in. As for what happens after that, I'd really prefer not to have to do physical labor or trades. And it doesn't seem like any other white collar occupation is really going to be insulated, other than perhaps medical. So my strategy is to basically wait and see what society looks like after the transition and I guess I'll try and decide on something then?

aabhay 8mo ago

I would emphasize self reliance and sustainable living. Things like home ownership, dollar-resistant assets like gold, and self education in topics that matter to you and your existence. Another way to frame it is what would you do in a low trust environment where corporations and the government were not to be trusted. You would likely avoid things like bubble bursting AI stock investments, jostling for rank in a company, etc.

le-mark 8mo ago
If post AI software engineering is a thing then this is NOT a bubble.

almostdeadguy 8mo ago

How do you plan an exit strategy for something that may or may not obsolete a whole field in a matter of months? Not sure there's a real way to do such a thing.

andrewmcwatters 8mo ago

[dead]

lifeisstillgood 8mo ago

We used to call the job Analyst Programmer. I am not sure what thing your code did but I am pretty sure you needed it - but who could understand that gap existed ? Who could explain to an AI that it needed to create this obscure code to solve that problem. And now comes the hard part - persuading your organisation to adopt it. AI can code - but can it understand what it missing from the organisation and persuade it to chnage - to spend years at industry conferences? Look at Starliner. NASA just announced that Boeing stuffed up, not with an engineering mistake (no one still knows exactly what broke) but that the whole organisation is so screwed up and so political Nasa just don’t believe Boeing can fix it. AI cannot fix our turf wars. That’s not intelligence (humans know going to war is bad, but Putin still exists). I ye the systems we live in, and work in. Changing those is feasible - once they are coded, transparent and open To inspection in a democracy. We need programmable introspective systems of organisation - democracies in other words. The engineering was not the problem - the problem was the organisation was more or less toxic and incapable of doing engineering. Writing code that won’t get used because if politics is a job we and AI can both do.

blindhippo 8mo ago

I've been using these tools for nearly a year and half on a daily basis. They've become an integral part of my tool box for solving problems. But writing code was never much more than 35-40% of my job while working for companies/others. Most my time has always gone towards communication, design, and validation. All three of those are not particularly vulnerable to mass AI automation except for the most trivial of scenarios and I have not seen evidence that has changed in over 2 years of so called "improvements". My "exit plan" ultimately is to be one of the engineers capable of using these tools to scale my impact accordingly so I can focus on higher order problem solving, which ultimately is what is most valuable. I would be more concerned if I was in marketing/sales or frankly middle management. Maybe this is just "copium" on my part, who knows, this sector is moving fast.

mekael 8mo ago
I was discussing this with one of my counterparts today, and we both agreed that writing the code is the easy part, and actually knowing what to do / getting requirements sorted out / working with our compliance team are the hard parts. That being said, I'm in a pretty highly regulated industry so it just might be that.
- df2dd 8mo ago
  Envisioning what should exist is always the hardest part. If LLMs provide substantial material to be able to produce what one envisions faster, that is great. But LLMs will not be doing the envisioning. Most humans already are poor at that. Hence why there are very few real 'visionaries' in history. Envisioning always requires deep thinking. If LLMs eat away at a humans ability to sit and think, this will make envisioning solutions harder. So you'll see more stuff produced, but largely more crap.

drivebyhooting 8mo ago

I have no workable exit strategy because I have small children and our family has been continuously under layoffs and terminations. If the doom really comes to pass then what future is there for us? I fear a life of impecunious servitude and poverty more than death.

jvanderbot 8mo ago
Regardless of the future you have to plan for the worst and hope for the best. I don't have time to post significantly about it but I'd love to trade thoughts and figure this out. Email?
alecco 8mo ago
Our ancestors survived worse. Wait out for the rainbow after the storm. Adjust and take it easy.
- Gud 8mo ago
  Uhm, our ancestors haven’t faced anything like this. This is the only time in the history of mankind, where you have a near global ruling elite trying to replace labor totally with mechanised thinking machines. The closest we can compare to is the off shoring in the 80s/90s.
- AlexeyBelov 8mo ago
  My ancestors lived in abject poverty.

jryan49 8mo ago

Save up enough money to be safe for many years after I get laid off so I have plenty of time to figure out what to do.

gedy 8mo ago

Writing the code has never been what has held companies back. A lot of companies will use the speed of AI to wallpaper over the fact that they don't know what to make or how to prioritize.

df2dd 8mo ago
When I read comments here, Im beginning to realise many people get lost in the noise and cant seem to figure out what exactly is the true bottle-neck of producing great software that people (be it consumers or folks employed at firms) purchase/use. Writing code faster alone doesn't change a great deal. Frankly it'll just create a larger influx of noise. Focus is very difficult to do, it'll become harder in the advent of LLMs.
- le-mark 8mo ago
  LLM change the dynamic so an individual or small team can replicate the work large companies. Especially if they worked on that large application before.
  - df2dd 8mo ago
    .... another simpleton take. Ive heard enough of these boring one-two liners.

OptionOfT 8mo ago

I'm on the other side. I work a lot with these tools and find that time after time there are subtle bugs, and lack of structure. And that's fine, I use it as a tool, but am I faster? Not where detail matters, and I tend to do a lot of work in that area. Generating dashboards? Absolutely. Converting to JavaScript vNext UI framework? Yes. Kernel code? No.