IT/Software career thread: Invert binary trees for dollars.

TJT

Mr. Poopybutthole
<Gold Donor>
47,580
131,247
It remains interesting in backend engineering. AI is forcing organizations to build clean backend. Something that backend engineers like myself or architects have demanded for decades but were completely ignored. Developing the backend has now become a task in being a ranch hand for a herd of AI agents. I have always liked the backend and data platform because of its variety. When compared to focused app dev or something.

It is nice to finally be listened to for the first time in over a decade that things need to have strict naming conventions. They always did but they very much need to now. Because if you cant identify some object in 5 seconds the AI can't either.

I couldn't find the original meme of this so here ya go:

1788527520135.png
 
  • 1Like
Reactions: 1 user

Deathwing

<Bronze Donator>
17,824
8,763
It's all the AI stuff.

We are no longer writing code but rather doing spec driven development. And people doing code reviews are just copy and pasting the AI slop from their agent into the code review. And people are just feeding that to their agent to fix whatever. Then I open a bug ticket and the first comment is 10 paragraphs from someone's AI agent where it thinks it knows what the problem is.

Add onto that the company BS, which I have tried not to bring up too often here but anyone know knows the internals of the company knows, it's super tiring. But this at least isn't new.

I actually like moving up and being in the position im in. I'm just feeling a bit disillusioned by what's going on here and in the industry. There's no way this is sustainable for development.
This was the last straw with my subordinate. I've posted about him from time to time, but after many tiring years, he's finally gone. I was encouraging him to use AI more. The stupid questions we've rightfully shamed him for, he can ask the AI instead. Now, he won't spend all afternoon resizing his WSL2. Not choosing which stuff can be delete, just how to.

I tried to take your suggestion to heart, so on an MR with a handful of short issues that I brought up, each one is responded to with paragraphs of first person expertise that I know he doesn't have.

A lot of the initial fearmongering with AI was about it enabling some unstoppable entity. It seems, a tale as old as time, that it enables some bad tendencies in humans. I'm trying to be very conscious about where and when I use AI. If only, but certainly not just only, because it seems to upsetting a fair amount of people to the point of quitting the industry.

Initially, I wanted the bubble to pop because I don't like what it's doing to my vidya games. Still don't like that. But now I'm hoping it does because it seems obscene token costs is the only way to force a real conversation about usage.
 

moonarchia

The Scientific Shitlord
<Bronze Donator>
32,324
63,857
it seems obscene token costs is the only way to force a real conversation about usage.
It's actually good that nvidia and those other fucks got too greedy too soon, and now people are having to look at the actual costs while there are still workers out there that can dig them out of the hole they created.
 

Deathwing

<Bronze Donator>
17,824
8,763
You got the guy fired?
In a manner of speaking, yes. I put him on a PIP for the second time in the 5+ years that I've managed him. He failed it again despite clearly documented goals and weekly coaching. Only this time, the PIP wasn't toothless because I was promised a replacement(I'm still pessimistic on that). My boss wouldn't let me action the first PIP because management at the time wouldn't replace him. And my boss is a packrat and thinks shitty labor is better than no labor.

Trust me, it's not just poorly interacting with a brand new technology. That alone is understandable. I've made some embarrassing blunders integrating AI into my workflow, and I'm pretty sure I'm not done.

It took 3 rounds of review before we figured out why his local AI review of MR was finding drastically different findings than mine. Well beyond the nondeterminism inherent to AI. He was implementing fixes by hand into one repo and then running the review in another repo. In both repos, the target branch was well out of date but by differing amounts. o_O

He's far better off in a blue collar job where he doesn't have to use his brain. AI is making that type of job vanishingly small in tech. That's another reason I was able to action the second PIP. Previously, he was doing hours worth of work. Shitty work full of repeated mistakes, but there was some undeniable value there. After we were bought out and merged with another company, his workload basically vanished because we are generating far less test results daily. And we were able automate a fair amount of it. Thus, he needed to prove himself adaptable to a more traditional developer role and he failed at that.
 

Fucker

Log Wizard
17,125
42,605
It took 3 rounds of review before we figured out why his local AI review of MR was finding drastically different findings than mine. Well beyond the nondeterminism inherent to AI. He was implementing fixes by hand into one repo and then running the review in another repo. In both repos, the target branch was well out of date but by differing amounts. o_O
Was he drinking gallons of vodka at work every day?
 

Noodleface

A Mod Real Quick
40,032
19,917
In a manner of speaking, yes. I put him on a PIP for the second time in the 5+ years that I've managed him. He failed it again despite clearly documented goals and weekly coaching. Only this time, the PIP wasn't toothless because I was promised a replacement(I'm still pessimistic on that). My boss wouldn't let me action the first PIP because management at the time wouldn't replace him. And my boss is a packrat and thinks shitty labor is better than no labor.

Trust me, it's not just poorly interacting with a brand new technology. That alone is understandable. I've made some embarrassing blunders integrating AI into my workflow, and I'm pretty sure I'm not done.

It took 3 rounds of review before we figured out why his local AI review of MR was finding drastically different findings than mine. Well beyond the nondeterminism inherent to AI. He was implementing fixes by hand into one repo and then running the review in another repo. In both repos, the target branch was well out of date but by differing amounts. o_O

He's far better off in a blue collar job where he doesn't have to use his brain. AI is making that type of job vanishingly small in tech. That's another reason I was able to action the second PIP. Previously, he was doing hours worth of work. Shitty work full of repeated mistakes, but there was some undeniable value there. After we were bought out and merged with another company, his workload basically vanished because we are generating far less test results daily. And we were able automate a fair amount of it. Thus, he needed to prove himself adaptable to a more traditional developer role and he failed at that.
Is he dumb or just extremely lazy?
 

Sheriff Cad

scientia potentia est
<Nazi Janitors>
35,170
85,995
It's all the AI stuff.

We are no longer writing code but rather doing spec driven development. And people doing code reviews are just copy and pasting the AI slop from their agent into the code review. And people are just feeding that to their agent to fix whatever. Then I open a bug ticket and the first comment is 10 paragraphs from someone's AI agent where it thinks it knows what the problem is.

Add onto that the company BS, which I have tried not to bring up too often here but anyone know knows the internals of the company knows, it's super tiring. But this at least isn't new.

I actually like moving up and being in the position im in. I'm just feeling a bit disillusioned by what's going on here and in the industry. There's no way this is sustainable for development.
Got bad news for you dog, I switched from SWE to law in ~2006... it's different, not really better or worse, but we're having AI training classes and all sorts of emails and programs "encouraging us to use AI" and so on all over the place too. It's less mature than SWE I'm sure, but its going the same direction. I'm really not sure what industry you'd switch to if you wanted to get away from AI usage.
 

Noodleface

A Mod Real Quick
40,032
19,917
Got bad news for you dog, I switched from SWE to law in ~2006... it's different, not really better or worse, but we're having AI training classes and all sorts of emails and programs "encouraging us to use AI" and so on all over the place too. It's less mature than SWE I'm sure, but its going the same direction. I'm really not sure what industry you'd switch to if you wanted to get away from AI usage.
I visited scotland recently and I want to live on a farm there. Only slightly joking.
 
  • 4Like
Reactions: 3 users

Deathwing

<Bronze Donator>
17,824
8,763
Was he drinking gallons of vodka at work every day?
I don't think so, but it would have been preferable. There's an established methodology to fixing and/or dealing with alcoholics.
Is he dumb or just extremely lazy?
Both. He struggles with fundamental concepts like git, but I would argue git can be a bit challenging. A more basic example would be struggling with the concept of virtualization. He would often think that because he authenticated on WSL2, it should also apply to Windows.

We've been 100% WFH since COVID. Since that time he's never bought, procured, stolen, whatever a monitor or two, working off his tiny laptop screen. He claims he works from his dining room table and has no room for a proper work space.

A wind storm came through recently and collapsed his bike shed, damaging his thousands of dollars in bikes. Why weren't they in the garage? His adult, jobless son-in-law does woodworking in there.

He's the quintessential apathetic cuck. He would frequently say "I'm doing my best" with the unsaid implication of "and get off my case". I explicitly and frequently brought up changing his mindset so that the implication was instead "and how can I do better?".
 

Deathwing

<Bronze Donator>
17,824
8,763
It's actually good that nvidia and those other fucks got too greedy too soon, and now people are having to look at the actual costs while there are still workers out there that can dig them out of the hole they created.
Maybe. The counter narrative is that open models(like qwen) are right behind frontier models. To paraphrase the AI product manager at my company, why else would Anthropic give us more and more tokens?

I actually don't know if we're getting more tokens. My subscription reports in percent usage and I've never quite dug into what my plan specifically represents. However, Claude Code does report cost per session, even if they aren't charging you that cost. On the $20/month plan, I can easily spend over $50/day. I'm hesitant to adopt more and more AI practices into our workflow if it means a true cost of $1500/month.
 

TJT

Mr. Poopybutthole
<Gold Donor>
47,580
131,247
I don't think so, but it would have been preferable. There's an established methodology to fixing and/or dealing with alcoholics.

Both. He struggles with fundamental concepts like git, but I would argue git can be a bit challenging. A more basic example would be struggling with the concept of virtualization. He would often think that because he authenticated on WSL2, it should also apply to Windows.

We've been 100% WFH since COVID. Since that time he's never bought, procured, stolen, whatever a monitor or two, working off his tiny laptop screen. He claims he works from his dining room table and has no room for a proper work space.

A wind storm came through recently and collapsed his bike shed, damaging his thousands of dollars in bikes. Why weren't they in the garage? His adult, jobless son-in-law does woodworking in there.

He's the quintessential apathetic cuck. He would frequently say "I'm doing my best" with the unsaid implication of "and get off my case". I explicitly and frequently brought up changing his mindset so that the implication was instead "and how can I do better?".
Brother I can't even get a Math PHD from Purdue to use git. But I have to hear him complain about rebuilding things as they design things in raw files and have to rebuild them constantly. Which would be solved with git.

WSL2 is gay anyway, just us a Mac and be done with it lol.
 

TJT

Mr. Poopybutthole
<Gold Donor>
47,580
131,247
Claude charges you the most by far. The norm is going to be using 3-5 AI tools at the same time with multiple agents on them each. That is just going to be the game we have to play.
 

Deathwing

<Bronze Donator>
17,824
8,763
Brother I can't even get a Math PHD from Purdue to use git. But I have to hear him complain about rebuilding things as they design things in raw files and have to rebuild them constantly. Which would be solved with git.

WSL2 is gay anyway, just us a Mac and be done with it lol.
WSL1 was indeed gay. For all of Microsoft's WELL deserved faults, WSL2 seems solid. The problem is that we need to support Windows.

Pick the least gay:
  • Native Linux and a Windows VM
  • Windows VM on AWS with barely functional persistent storage that lasts a week at most
  • Native Windows and a Linux VM
 

Noodleface

A Mod Real Quick
40,032
19,917
Brother I can't even get a Math PHD from Purdue to use git. But I have to hear him complain about rebuilding things as they design things in raw files and have to rebuild them constantly. Which would be solved with git.

WSL2 is gay anyway, just us a Mac and be done with it lol.
Have they ever tried tortoisegit? We have some old fudds here that use that because they can't understand git bash