causal

Very funny

Agreed... I'm not at the top.

Did you mean decrease in your last sentence? Or do you simply mean any solution will make the problem worse?

I mean most of us certainly don't have that kind of authority, and that's not really going to stop AI use when it comes embedded in every service these days.

Verifiability is part of it, but I think the "semantic ablation" article on the front page really captures my problem with AI-washed writing: https://www.theregister.com/2026/02/16/semantic_ablation_ai_...

I think any use of AI "unrolls" the prompt into a longer but thinner form. This is true of code too I think, but it's still useful because so much of coding is boilerplate and methods that have been written a thousand times before. Great, give me the standard implementation, who cares.

But if you're doing hard algorithmic work and really trying to do novel "computer science", I suspect semantic ablation would take an unacceptable toll.

No sir

YES this hits the nail on something I've been trying to express for some time now. Semantic ablation: love it, going to use that a lot not now when arguing why someone's ChatGPT-washed email sucks.

Semantic ablation is also why I'm doubtful of everyone proclaiming that Opus 4 would be AGI if we just gave it the right agent harness and let all the agents run free on the web. In reality they would distill it to a meaningless homogeneous stew.

Fair but I have seen workplaces keep phone use largely curtailed. Surely it's not so impossible with AI...right? Right...? :/

Yeah if the prompt is as specific as this post, then that's probably the issue...

Thanks for sharing that anecdote. I think everyone is susceptible to misinformation, and seniors might be especially unprepared to adapt to LLMs tricks.

> there is almost nothing that a human is better at than Opus 4.6.

> You wouldn't expect a Jr. dev to be the best at keeping things dry either.

So a junior dev is better than almost all humans at everything?

I'm countering the basis of your original claim; "there is almost nothing that a human is better at than Opus 4.6". This is simply not true.

I fear that AI will be intelligent enough to negate human general intelligence before it is itself generally intelligent.

This is my greatest cause for alarm regarding LLM adoption. I am not yet sure AI will ever be good enough to use without experts watching them carefully; but they are certainly good enough that non-experts cannot tell the difference.

> violates DRY in unholy ways

Well said

Opus is the very best and I still throw away most of what it produces. If I did not carefully vet its work I would degrade my code bases so quickly. To accurately measure the value of AI you must include the negative in your sum.

Yup and it's clearly getting attention.

> The project is abandoned but it still works well

Hard deal breaker. And alternative keyboards in iOS feel second class in some ways, so we really rely on Apple to get it right.

Switching from Android, I was shocked by how much in fact did not just work. I kept a running list of basic features that were clearly broken.

Especially around text editing. It seems like they made some fundamental mistakes with their text inputs that they are playing hard defense on. I never know if a given field is going to respond to long-press, double tap, or what context menu I will get if any.

That's a bit like saying just give blind people cameras so they can see.

Agreed. I love the elegance of ARC, but it always felt like a gotcha to give spatial reasoning challenges to token generators- and the fact that the token generators are somehow beating it anyway really says something.

Could write a similar post about their cloud UI. Feel like you have so little control over the thing. I think Cursor has an uphill battle by having to go through the API, but they certainly do a better job of making conversations and context more transparent and manageable.

Same thought. If anything I'm usually trying to find ways to reduce how much context is carried over.

"Flow-state inducers" is interesting, and I think you're onto something. There's something so satisfying about getting working code so easily, feeling like you're jumping from one solution to the next.

Yes. Programmers might be especially susceptible precisely because our advanced understanding makes us think we cannot be easily fooled.

But we are also easier to impress: only we understand how difficult it is to one-shot code a working app.

AI psychosis sets in when AI captures enough of your perception to alter your reality. Programmers might be "smarter", but we give AI a bigger set of tools to capture our perception with.

And then there's the fact that we want to be fooled.

Helpful, absolutely, but only if you're solving the right problem. Solving the wrong problem with AI is doubly harmful because it will almost always give you something that runs, but now you are on a path that takes a lot of willpower to give up.

> Always Have an Agent Running

I'm a huge believer in AI agent use and even I think this is wrong. It's like saying "always have something compiling" or "make sure your Internet is always downloading something".

The most important work happens when an agent is not running, and if you spend most of your time looking for ways to run more agents you're going to streetlight-effect your way into solving the wrong problems https://en.wikipedia.org/wiki/Streetlight_effect

That makes sense. I've thought for a while that having an agent that takes initiative rather than reacting to inputs could be really useful, and I imagine it takes a lot of trial and error to make it take just the right amount of initiative.

I'm still trying to understand what makes this project worthy of like 100K Github stars overnight. What's the secret sauce? Is it just that it has a lot of integrations? Like what makes this so much more successful than the ten thousand other AI agent projects?

> amongst smart people i know there's a surprisingly high correlation between those who continue to be unimpressed by AI and those who use a hobbled version of it.

I've noticed this too, and I think it's a good thing: much better to start using the simplest forms and understand AI from first principles rather than purchase the most complete package possible without understanding what is going on. The cranky ones on HN are loud, but many of the smart-but-careful ones end up going on to be the best power users.