What happens when your veteran sysadmin coworkers stop reading docs and blindly trust Gemini hallucinations? A catastrophic 2.2x performance drop and a hilarious Reddit rant.
Have you ever felt like you're working in a cult where the holy scripture is a generative prompt? Because the "ChatGPT/Gemini said it's easy" pandemic is getting out of hand.
Today on Coding4Food, we've got a spicy Reddit drama from a 25-year-old sysadmin who is watching his veteran on-prem infra team (average age 45) completely lose the plot and bow down to their new AI overlords.
Our OP works in a business-critical environment managing a mix of Linux and Windows servers. You know, the kind of place where blindly pasting commands into production is basically signing your own death warrant. His older coworkers have always been smart, capable mentors.
Until the AI bug bit them.
The Manager's 1000-Word Slop Bomb: While dealing with an incident OP thought was resolved, his manager casually drops a 1000+ word unformatted raw response from Gemini into the chat. The manager didn't read it. He didn't explain it. He just said, "Yeah, that should solve it."
When OP asked for some context because he didn't want to just forward raw AI slop to the incident response team, the manager looked at him like a "sheep in the rain."
The kicker? The manager actually said: "Just ask Gemini to explain it to you if you don't understand it?"
The 30-Minute Architecture Explanation vs. A CLI Prompt: It gets worse. OP spends 30 minutes at a whiteboard meticulously explaining the whole architecture to a coworker. Afterward, he asks the guy to look up a simple flag for a CLI tool.
The coworker's response? "Can you write me a prompt for that?"
Excuse me? What happened to reading the man pages? What happened to Google? Later, this same coworker proudly presents his "research"—a Gemini chat riddled with hallucinations. OP warns him, "This is a waste of time, it'll be 2x slower." The coworker ignores him because, quote, "Gemini says it will be faster."
Spoiler alert: It ended up being 2.2x slower. And when faced with the cold hard benchmarks, the coworker just stared at OP... exactly like a sheep in the rain.
The post blew up with over 1k upvotes, and the community had a field day roasting these new-age AI cultists:
LLMs are fantastic tools. But they are essentially highly confident, eloquent bullshitters. They are the eager junior developer who wants to impress you so badly that they'll just make things up.
When experienced engineers lose their critical thinking skills and replace documentation with a chat prompt, we're all doomed. If you blindly copy-paste commands into a critical on-prem server without knowing what they do, you are no longer an engineer; you're just a highly paid typist waiting to be replaced by the very tool you're worshipping.
Use AI, but please, run the output through your Brain API first.
Source: Reddit - My older coworkers have accepted AI as the source of truth
