People who tell you 'AI is changing everything' are lying
thenerve.news
thenerve.news
The 0% success rate in enterprise AI is wild though. If you can't write good internal documentation and define clear processes, throwing a chatbot on top of it just makes the chaos faster. Trying to skip the foundational engineering logic is a huge mistake.
https://hermit-tech.com/blog/ai-mania-is-eviscerating-global...
>The current generative AI is just spicy autocomplete on steroids—doing phrase and paragraph level prediction rather than character/word level prediction. It doesn't "answer" questions, it just extrapolates the statistically most likely answer-shaped string based on its training set.
My reply:
>How do you explain the multiple unsolved math problems that AI has recently solved?
No answer by Stross.
This is to be expected out of a slop machine that has no notion of higher level concepts.
No one claimed autocomplete-on-steroids could not have any use. Some of us claim the cons (destruction of the planet, destruction of human though capabilities) infinitely outweigh the pros (you solved a useless mathematical problem or built yet another note-taking app, congratulations!)
And maybe even more so since they carry the potential for bigger impact.
"perhaps a few of the younger members of those families would have to do their own homework again."
Talk to any professor. AI has had a profound and irreversible impact on higher ed, it has changed students completely, it has changed how professors operate, and it is changing how papers are written. The leading voices in mathematics are having profound discussions on the nature and uses of the technology. To dismiss these radical changes with this kind of pathetic snark is actually just embarrassing.
University CS instructor here. The above is true, but it's not generally for-the-better. We're working with it, of course, since there is no choice. But students need to develop problem-solving skills that LLMs only steal from them. And they need to learn problem-solving skills that involve LLMs.
And so I work hard to help the students develop skill while using the very tool that robs them of it. It's possible, but it's extremely challenging for instructors and students, alike. It's a challenge that I rise to every day to maximize student benefit.
But LLM misuse during school makes the drawbacks completely negate all benefits, making LLMs positively dangerous for students who are mesmerized by the siren song.
In short, the snark isn't entirely unwarranted. Many students are bringing their robot to the gym to lift the weights for them. And when we, with pen and paper, corner these poor souls who haven't learned to think, they simply fail out and are lost. It's a problem.
So... great with the math breakthroughs and the whole "thinking at a higher level of abstraction" thing, but we have to work with the significant dark side of all this.
"You're not imagining it—the brain-rot is real."
From the programming standpoint, it seems like the devs who honed their skills without AI being available to them are the most capable and productive at using it. But we are downsizing to increase spend on tokens, sales people are building apps that are nothing more than untested PoCs and asking how to deploy it into production, our executives believe the technology is some kind of magic...
I hope someday to hire some young developers again, when that day comes: thanks for your efforts getting them ready.
I can't speak to other industries but LLMs have completely changed many if not all aspects of my job. We're producing more code, better tested code, better reviewed code, at a faster speed than ever before.
This will probably be downvoted since HN seems to be full of people who still think LLMs are ChatGPT 3.5-level "intelligence" and aren't willing or able to learn/try new methods. That sounds harsh but the truth is that writing code by hand is going away.
The people trying to shove "AI" into everything without thinking critically are wrong but so are the people who still, confidently, claim that LLMs cannot write good code. That might have been somewhat true 2 years ago but it hasn't been the case since at least Dec 2025 and has only improved since then.
As for the "0%" success rate in enterprise? Ehh, I'm not sure I buy that but I also work for a smaller startup and enterprise companies never cease to amaze me with their waste so maybe it's not so surprising. But it's not because LLMs are worthless, just that those people don't know what they are talking about and/or horribly overestimate how to use them effectively.