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

TomServo

<Bronze Donator>
8,658
15,381
I haven't yet heard of any company opting to run open source AI on their own datacenter or AWS to save money paying Claude Code and the like.

Today, right now, paying the freemium rate is LOTS cheaper. But that is a fuse on a bomb. One just long enough that leadership today doesn't see what's happening.
The performance goals necessitate the extreme quarterly thinking. thus AI accelerated to hell and Cloud migration to eliminate capex so he meets his goals for his bonus. it is absolutely not long term a good idea. We are not running AI on our on hardware. we are using aws bedrock to run a slew of different models.
 

TJT

Mr. Poopybutthole
<Gold Donor>
46,535
127,158
The costs of getting spitroasted by paying Claude Code (or whoever) for models and licensing and then paying AWS to operate the models is going to be absolutely insane.
 
  • 1Worf
Reactions: 1 user

Neranja

<Bronze Donator>
3,098
5,055
The costs of getting spitroasted by paying Claude Code (or whoever) for models and licensing and then paying AWS to operate the models is going to be absolutely insane.
It is the razor blade business model: Lure you in with a cheap "starter pack", only to then fleece you with the subsequent blades. And some time later offering you the next model, now with one more blade.

Worked wonders with inkjet printers and standard software licensing (e.g. Adobe).
 

Noodleface

A Mod Real Quick
39,431
17,917
I actually stopped using the windsurf IDE today because it was so bad and went back to vanilla VSCode. Told my boss the company can shove it .

From what I heard Google tried to buy windsurf, then instead just poached the founders and left the shell of a company
 

Haus

I am Big Balls!
<Gold Donor>
18,726
77,338
But those are the trends because thats currently cheaper, no? If running on their own metal was cheaper, they'd be doing that, wouldn't they?

There's a degree of cost savings, but many places are now hitting a wall of cost accelerating because of sloppy architecture trying to "scale". But I know places right now that have things they would love to bring back into their own data centers, but can't because it's getting harder and harder to find people who can run a good data center properly.

Also AI is driving a spiraling price rocket on essential hardware. Is essence AI needs all the memory, not just all the memory now, but all the memory that's going to be produced for the next 5-10 years... And all of a lot of other resources... and a lot of the electricity, in fact more than is currently being generated.

But I digress, I'm still waiting because I still think the "AI Hype Train Investment Cycle" bubble will pop in around a year. That will still leave the "AI as a technology" to keep advancing (just more organically) , which will keep the inevitable movement towards "AI breaks how we exist because we just don't need so many people anymore" factor.
 
  • 1Like
Reactions: 1 user

Khane

Got something right about marriage
21,536
15,438
"new and shiny" often overrules "cost effective" when it comes to technology and how companies decide what technology to utilize. Lots of money gets thrown around and wasted regarding tech.
 
  • 1Like
  • 1Truth!
Reactions: 1 users

ShakyJake

<Donor>
8,451
21,096
I actually stopped using the windsurf IDE today because it was so bad and went back to vanilla VSCode. Told my boss the company can shove it .

From what I heard Google tried to buy windsurf, then instead just poached the founders and left the shell of a company
Why do they even care which IDE you use? Everyone in engineering at my company uses Visual Studio or VS Code. I’m the weird snowflake who chooses JetBrains stuff but there’s no way they’d know.
 

TJT

Mr. Poopybutthole
<Gold Donor>
46,535
127,158
Because Windsurf, Cursor, etc all track your use of AI in coding. It says stuff like

Noodleface Noodleface : 7FEB-14FEB 2026
  1. AI Generated Accepted/Rejected: 100/2000
  2. Tabs Accepted/Rejected: 10/5000
  3. Commits of Accepted/Rejected Code (and to what branches): 1/900
They then look at this and get on your case about how effectively you are or are not using it. In my org its 50% of your weighted performance for 2026. I am building out these metrics solutions right now for a consolidated view of all of our official AI use. Solely so leadership can evaluate you for it.
 
  • 1Solidarity
Reactions: 1 user

ShakyJake

<Donor>
8,451
21,096
Because Windsurf, Cursor, etc all track your use of AI in coding. It says stuff like

Noodleface Noodleface : 7FEB-14FEB 2026
  1. AI Generated Accepted/Rejected: 100/2000
  2. Tabs Accepted/Rejected: 10/5000
  3. Commits of Accepted/Rejected Code (and to what branches): 1/900
They then look at this and get on your case about how effectively you are or are not using it. In my org its 50% of your weighted performance for 2026. I am building out these metrics solutions right now for a consolidated view of all of our official AI use. Solely so leadership can evaluate you for it.
Heh. Yeah, my organization is OBLIVIOUS to all of that. I'm the only dev actually using this stuff.
 

Noodleface

A Mod Real Quick
39,431
17,917
Why do they even care which IDE you use? Everyone in engineering at my company uses Visual Studio or VS Code. I’m the weird snowflake who chooses JetBrains stuff but there’s no way they’d know.
As TJT said. And the company is doing a big push to standardize development across the company. Now if you know what my company makes and the size it is, you know how laughable that is.

Also $$$. They paid an extreme amount of money for this.
 

ShakyJake

<Donor>
8,451
21,096
My company has a 200B market cap, but they aren't software focused. Our division is a teeny tiny slice, so they wouldn't even know or understand what to standardize or enforce. Oh well, good luck.
 

TJT

Mr. Poopybutthole
<Gold Donor>
46,535
127,158
We already have some devs (all of the Indians of course lol) have like >250k lines of accepted code suggestions. Which is of course completely retarded and nonsensical in every way. Except to the non dev leadership who see that this guy must be ULTRA PRODUCTIVE.

Never mind that a single feature in an enterprise application that took months of product research, requirements gathering, scoping, and so on could still end up being like 500 lines of code tops and still be a massively impactful change to the product and bring in tons of revenue.