this post was submitted on 03 Jul 2026
344 points (96.0% liked)
Fuck AI
7570 readers
652 users here now
"We did it, Patrick! We made a technological breakthrough!"
A place for all those who loathe AI to discuss things, post articles, and ridicule the AI hype. Proud supporter of working people. And proud booer of SXSW 2024.
AI, in this case, refers to LLMs, GPT technology, and anything listed as "AI" meant to increase market valuations.
founded 2 years ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
Citation needed.
I'm an experienced solo developer who tried embracing the state of the art in agentic programming on a several month long non-trivial project. My income depends on this project succeeding so I'm not fucking around here looking for an excuse for ai to fail. I'm trying to keep up...
My idea is that if this works, I want it working for me.
The results were unimpressive at best. It feels very fast at first but the entropy is insidious and comes even faster.
When you hit the wall where the agent stops being able to understand it's own work and contextualise tasks in a meaningful way, you're left with a pile of incomprehensible shit that kind of looks like code.
You will have nothing. Nothing but a giant token bill, wasted time and a sense of shame.
The last study I saw might have been from 18 months ago, but they found that AI costs programmers more time than they saved AND that programmers completely miscalculated the efficiency of AI, unaware that it was slower.
Whether or not AI is more efficient now, it's demonstrated that individual programmers are pretty bad at evaluating AI efficiency subjectively.
That leaves the reports of CEO's, desperate to justify layoffs from lack of work with AI, desperate to goose their stock prices by being AI-first, and the slavish business press who exclusively repeats whatever they say.
Citation definitely needed.
18 months is an eternity in terms of AI performance. Local models now are better than the frontiers models were at that moment.
In my experience, AI automates the boring part of programming (which is actually writing the code). It leaves me able to focus on user experience, architecture, the fun stuff. Am I more productive? Maybe a little, yeah. Is it magically going to replace my whole team? Only a fool would think so.
How many months have been since "in 3 months, every programmer will be out of a job due to the fast progress of AI"? At least a couple years.
Sloppers overestimate by a huge amount both the current performance of LLMs and their improvement speed.
The only thing that gets larger exponentially as time goes on is the cost. As predicted by actual scientific studies that said LLMs would hit a diminishing returns barrier rather quickly.
I've been a programmer for the last 15+ years, and the only constant over that time has been change. Every few years, there's a new technology that "radically simplifies" programming, such that"business people will finally be able to write the code themselves". Without fail, what ends up happening is that we build more complex systems on top of it and we still need programmers to deal with it. AI is a radical shift, yes, but it still doesn't liberate "business people" from the burden of clarifying their ideas. That is what I do, and the technology I'm using to achieve that doesn't fundamentally change that fact.
That, and leaky abstractions. They always end up leaking, and you need someone who can understand what's going on.
I think every three to six months I read people on HN saying the new X.Y model has finally broken the barrier and is the game changer that will forever transform the industry.
Whatever AI is like today, it's been demonstrated programmers wildly overestimate how much time AI saves them; meanwhile, most everyone else has a vested interest in exagerrating the efficiency of AI and there is very little media presence pushing back against or investigating AI claims.
“The common wisdom” that AI saves a lot of time could just be to result of ignorance and delusion. I'm happy to wait for objective research before making assumptions one way or the other.
And however efficient AI actually is, nobody wants to pay the true cost of it yet.