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

TJT

Mr. Poopybutthole
<Gold Donor>
44,153
115,647
Speaking of our Engineering leaders just posted us this to read:



lol our devs just got ass blasted. why arent you agentic coding. why not use AI. why are your delivery schedules months and not hours NOW NOW NOW. why not using cline, codium cursor and zed
This is a bit stupid on our leaders part IMO. I can 100% crank out some dirty solution for something that meets minimum requirements the same day. The trouble is all the other stupid shit that prevents you from deploying it. Testing, monitoring integration, CI/CD, code reviews, etc. This is about to happen though. Everything has to be done immediately. Even more than it was before. I am not looking forward to that.

But I am more than happy to shoot from the hip.
 
  • 3Like
Reactions: 2 users

TomServo

<Bronze Donator>
7,747
12,442
yeah this shit is retarded. Claude write my code. claude debug my code you wrote for me. Claude now remove all vulnerabilities. eegads i r the developer
 

TJT

Mr. Poopybutthole
<Gold Donor>
44,153
115,647
yeah this shit is retarded. Claude write my code. claude debug my code you wrote for me. Claude now remove all vulnerabilities. eegads i r the developer

As a senior developer. Most of my time developing is not actually writing the code. It's researching, understanding, validating requirements, etc. None of which AI helps you with really. Unless you want to bullshit stuff. When I do use AI to generate code, it does make stuff. But you have to also know what you wanted to do and be able to overwrite it where you know its wrong.

The best line in that post is the cost implications. It's very clear that corporate leadership fully intends to capitalize on this by hiring foreigners or juniors and spending the cost savings partially on Cursor AI and the like. With the goal of saving money overall, while amplifying productivity exponentially (supposedly).

What do you do in 5 years when your cheap juniors with $20k+ in AI prompts a year tacked onto your costs move into senior roles that also have $20k-$30k+ a year tacked onto their salaries?
 

TomServo

<Bronze Donator>
7,747
12,442
As a senior developer. Most of my time developing is not actually writing the code. It's researching, understanding, validating requirements, etc. None of which AI helps you with really. Unless you want to bullshit stuff. When I do use AI to generate code, it does make stuff. But you have to also know what you wanted to do and be able to overwrite it where you know its wrong.

The best line in that post is the cost implications. It's very clear that corporate leadership fully intends to capitalize on this by hiring foreigners or juniors and spending the cost savings partially on Cursor AI and the like. With the goal of saving money overall, while amplifying productivity exponentially (supposedly).

What do you do in 5 years when your cheap juniors with $20k+ in AI prompts a year tacked onto your costs move into senior roles that also have $20k-$30k+ a year tacked onto their salaries?
collapse back in on itself and they rehire americans. shrug who the fuck knows.
 
  • 1Like
Reactions: 1 user

Asshat Foler

Log Wizard
<Gold Donor>
49,166
41,705
Speaking of our Engineering leaders just posted us this to read:




This is a bit stupid on our leaders part IMO. I can 100% crank out some dirty solution for something that meets minimum requirements the same day. The trouble is all the other stupid shit that prevents you from deploying it. Testing, monitoring integration, CI/CD, code reviews, etc. This is about to happen though. Everything has to be done immediately. Even more than it was before. I am not looking forward to that.

But I am more than happy to shoot from the hip.

yeah this shit is retarded. Claude write my code. claude debug my code you wrote for me. Claude now remove all vulnerabilities. eegads i r the developer
Nah, these AI agents are awesome. My company (big tech) just released an internal CLI based agent this past week. It’s insanely impressive for increasing efficiency. I can have it recursively read packages I have zero familiarity with and ask it to generate architectural diagrams, find where code is doing x/y/z and also ask it to of course generate code. The code generation is tenfold better than it has been traditionally because it’s now able to read all our internal packages and their dependencies. I find this most beneficial when I get pulled into another team that’s utilizing a language I’m not familiar with, has packages I’ve never touched and has absolutely awful documentation (I.e. none).

It’s absolutely not a complete replacement for knowing how to code or how to architect solid software but it’s pretty amazing how far it’s come in just a year. Little over a year ago we got access to a copilot like tool for our IDE and it sucked ass at generating code.
 

Phazael

Confirmed Beta Shitlord, Fat Bastard
<Gold Donor>
15,047
32,868
Its just the usual IT circle of fail. Company execs try to save a buck short term by loading up on Poos (and in this case AI they don't really understand to try to cover the skill gap). Customers and User Base ultimately get pissed off because Poos do Poo things. A couple execs get pushed out (with golden parachute of course) to make it look like their is accountability. Company has to hire local white and asian guys to unfuck everything in what is now a destroyed internal company culture, thanks to the Poo things that went on. Said dudes unfuck things and next wave of execs thing they are in the clear and repeat the same shit or try applying Poo to another part of their business. Started in tech support call centers and the aura of Poo has radiated out from there. It ends when the company goes tits up from wasting money and/or general dissatisfaction with the Poo productivity or when the Poos insert enough of their own into senior leadership to take over completely and run the company into the ground.
 
  • 2Solidarity
Reactions: 1 users

TJT

Mr. Poopybutthole
<Gold Donor>
44,153
115,647
We have an Engineering Week going on. It's kind of neat seeing all of the shit we're paying for showcased (by poo contractors of course!).

The current bae is the action driven AI Agent that is more personalized to your need. The next bae is the Model Context Protocol which is currently being deepthroated by management as the end to all of our issues and exponential productivity gains.

Currently we are standing up dozens of MCPs (Cursor Directory) and attempting to capitalize on them. Here's the demo that we had today:
  1. Guy sets up a React MCP of some kind.
  2. Loads up his IDE. MCP is able to sync with Azure via CLI.
    1. It can also access Figma (wireframing tool)
    2. Can access Jira
  3. Creates a branch and begins chatting with the MCP AI Agent.
  4. MCP reads the Figma after some prompts and the ticket and generates the feature code. Loading it up on the local machine docker instance.
  5. Dude shows us the feature mostly working or at least a working start.

Big problem here? His demo had a Jira ticket that had stunning levels of detail. I asked him who wrote that ticket because it looked like an actual design document and not an lol Jira ticket like I get. Dude brushed my question aside and ignored me (lol).

I like where this is going but this was a deceptive demo. I told my director that we need a demo that can actually parse data problems and provide potential solutions. Even "problems" with the data are not binary. It's anomalies specific to our industry that the AI would never know to look for. Or better yet, where to look. Time will tell though.
 
  • 1Like
Reactions: 1 user

TomServo

<Bronze Donator>
7,747
12,442
We have an Engineering Week going on. It's kind of neat seeing all of the shit we're paying for showcased (by poo contractors of course!).

The current bae is the action driven AI Agent that is more personalized to your need. The next bae is the Model Context Protocol which is currently being deepthroated by management as the end to all of our issues and exponential productivity gains.

Currently we are standing up dozens of MCPs (Cursor Directory) and attempting to capitalize on them. Here's the demo that we had today:
  1. Guy sets up a React MCP of some kind.
  2. Loads up his IDE. MCP is able to sync with Azure via CLI.
    1. It can also access Figma (wireframing tool)
    2. Can access Jira
  3. Creates a branch and begins chatting with the MCP AI Agent.
  4. MCP reads the Figma after some prompts and the ticket and generates the feature code. Loading it up on the local machine docker instance.
  5. Dude shows us the feature mostly working or at least a working start.

Big problem here? His demo had a Jira ticket that had stunning levels of detail. I asked him who wrote that ticket because it looked like an actual design document and not an lol Jira ticket like I get. Dude brushed my question aside and ignored me (lol).

I like where this is going but this was a deceptive demo. I told my director that we need a demo that can actually parse data problems and provide potential solutions. Even "problems" with the data are not binary. It's anomalies specific to our industry that the AI would never know to look for. Or better yet, where to look. Time will tell though.
now do this with random jira ticket. bet it aint shit
 
  • 2Truth!
Reactions: 1 users

TomServo

<Bronze Donator>
7,747
12,442
fudging around with codeium. gonna try to get them to self host in aws. likely gonna get shit on for that. we need to go nowwww
 

Deathwing

<Bronze Donator>
17,208
8,202
Was the feature relatively isolated? It would be more impressive if it could integrate a feature into an existing codebase and files. Even more impressive if the feature required synchronized merge requests to multiple repos.
 

TJT

Mr. Poopybutthole
<Gold Donor>
44,153
115,647
He had an existing React APP. But it was still an app for this demo and it was clearly limited whatever it was. The whole premise resided on two things though.
  • Ultra well written ticket with lots of very specific technical details.
  • Figma wireframe was completed before work was requested (yeah right. I am sure in some timeline this happens. Maybe.).
The agent did not do any merge or commits on its own. It only worked for him locally like a more advanced ChatGPT for a specified purpose.
 

TJT

Mr. Poopybutthole
<Gold Donor>
44,153
115,647
So, a jira ticket that says "upgrade third-party/postgres" isn't going to cut it?
La Xbox GIF
 
  • 1Worf
  • 1Double Worf
Reactions: 1 users

ShakyJake

<Donor>
8,135
20,463
Our division maintains two products that perform effectively the same function, both acquired when we bought out two competitors about 12 years ago. One acquisition was a surprise due to legal hurdles, but we got both, so now we have two products. One’s a hit in the U.S. market, while the other, with its basic but functional “localization” support, does better overseas.

For years, we’ve talked about retiring one product in favor of the other, but there’s never enough time or budget to localize the preferred U.S. product for global markets. I’m on the overseas product team, by the way.

Now, the latest idea floating around is to give the overseas product a UI re-skin to match the U.S. product’s look. The goal is to ease customers into an eventual switch without jarring them visually. I think this is a terrible plan for a few reasons. First, the overseas product is an ancient ASP.NET WebForms app from the early days of .NET. It’s pure spaghetti code, with a mishmash of patterns and concepts piled on over the years. Re-skinning it isn’t as simple as slapping on a new CSS stylesheet—think hardcoded pixel widths, inline styles, and a maintenance nightmare.

Second, with NLP AI advancing, I’m convinced the way we interact with apps will shift dramatically in the next 5 years. Any effort spent on this re-skin is a waste, as both products will likely need a complete UI overhaul by the time we even consider consolidating them.

This post is part rant, part gripe about our marketing team. It drives me nuts that they dream up these ideas without checking with engineering first. We heard about this through the grapevine, which means they haven’t even bothered to loop us in. At the very least, they should’ve pitched this to our team before starting any serious discussions.
 
  • 1Thoughts & Prayers
Reactions: 1 user

TJT

Mr. Poopybutthole
<Gold Donor>
44,153
115,647
I don't see why leadership can't see the writing on the walls here. The productivity of the AI agent is quite clearly explicitly tied to the amount of total access you give it and the level of detail/context you provide it on your technical infrastructure. The demo MCP had to access 3 separate systems just for that demo. All of which are normally locked behind Okta/SSO/etc.

You want this dumbass AI agent to work all on its own though.
 
  • 1Like
Reactions: 1 user

TomServo

<Bronze Donator>
7,747
12,442
I don't see why leadership can't see the writing on the walls here. The productivity of the AI agent is quite clearly explicitly tied to the amount of total access you give it and the level of detail/context you provide it on your technical infrastructure. The demo MCP had to access 3 separate systems just for that demo. All of which are normally locked behind Okta/SSO/etc.

You want this dumbass AI agent to work all on its own though.
Well we do have dumb automated testing you can set up to mfa and sso to systems. Shit we authorized it in prod. But that is just stability testing and agents running through a website or mobile app.

Now giving something access beyond that. And given how dogshit most requirements are.

We gonna see alot of wasted time at scale throwing 11 AI tools at entire engineering teams
 
  • 1Like
Reactions: 1 user

TJT

Mr. Poopybutthole
<Gold Donor>
44,153
115,647
Phazael Phazael

Speaking of Pajeetification. They are consuming basic administrative jobs now too. I had to talk to our procurement team today. Normally this is a team of 3-5 people for the entire company. Most being women. Whatever, that's fine. This is a truly basic admin job that you can absolutely find people to do in the US for a measly $85k a year on the upper end. You don't even have to vet products to buy you just have to haggle with the vendors a bit when the requestors decide they want it.

Nope. All pajeets now. I was kind of stunned.
 
  • 1Tiresome
Reactions: 1 user