dcre

Because it’s a web site.

This example disproves your point. And LLMs do not just do what we tell them to do. They are perfectly capable of asking “are you sure? this has X, Y, Z consequences you may not like.” They do it all the time.

Didn’t read the whole thing, but it starts out a lot better than I expected. Most critics are in denial about the state and trajectory of the technology. This critic is not, he just thinks it’s bad.

Ha! I have a heterodox view on the taste thing, which is that it’s mostly cope: the models apply something like taste in nearly everything they do, and these taste posts will disappear as that becomes more obvious. However, besides the substantive disagreement, I think your post is fine since it appears to be your actual point of view, written down by hand. It’s not particularly novel or insightful (most of these takes are basically “there are still some things the models can’t do” — no shit!) but it’s measured and reasonable. Fine.

No! It’s garbage!

Heh whoops. I thought my other comment didn’t post so I wrote this one instead.

This post is awful! Annoyingly written and zero insight. Most posts about this taste concept are bad (and there have been so many), but this is below average.

This post sucks, and it was also written by Claude.

I feel like we need a name for this error. Pretraining fallacy? Coding ability is all about RL.

This is exactly what I mean. You’re thinking about the abstract conditions and not what it would take to secretly stage something like this.

Wish people crying “marketing” would think about what it is they’re claiming.

Think about what you’re saying. If Jarred does it because he wants to, what does marketing have to add? How do they “cash in on it”?

Both!

I don’t understand why anyone thinks management of any kind was involved in this at all. It seems much more plausible to me that Jarred just does what he wants. He has said as much.

> since AI makes rewrites to certain languages relatively easy

This premise makes no sense. AI makes rewrites to any language easy.

I think we actually are seeing an app explosion, just not a consumer app explosion.

A joke but also not a joke.

This reads to me as fully written by LLMs. Pangram agrees. Note the (alleged) author misHQ’s comments on this thread are getting downvoted as obvious slop.

https://news.ycombinator.com/item?id=48706307

Even if it were written by hand, it’s a very poor and frankly stupid essay about an interesting topic. “The model's attention is a fixed quantity, and it has to add up to one, so the more things you make it look at, the less of that attention any single earlier thing can keep.” This is borderline gibberish and it outright rejects the interesting question about LLMs and attention, namely that they have very different capacities from us. LLMs can read an entire OpenAPI schema in seconds and immediately construct valid requests from it. The article first points this out, and then switches to arguing that LLMs have similar limits to us. It’s completely incoherent.

That message is obviously aimed at the government. See the other thread.

All we can say is empirically this is not true. Plenty of people whose full time job is managing rack infra are sick and tired of the customizability of commodity hardware, of the many custom bugs in commodity software, and of the customizability of the price of VMware.

Source code here :)

https://github.com/oxidecomputer/rack-explorer

Thanks, missed it!

He has said every month for the past three years that there is no technical progress left to be made in LLMs and that there is no more room in the market for inference spending to grow.

Here[0] is a fun selection of excerpts from his July 2024 post "How Does OpenAI Survive?"[1]

"I see no signs that the transformer-based architecture can do significantly more than it currently does."

[0]: https://xcancel.com/pathsnotchosen/status/206360940100129633...

[1]: https://www.wheresyoured.at/to-serve-altman/

Jane Street’s profit per employee (profit, not revenue) in 2025 was high single-digit millions per employee.

Yeah, I agree — not a big difference for me between the Opuses. I really just meant Opus 4.x.

Claude Code (or Codex, or OpenCode, or Pi, or Amp — whatever) can do this out of the box without any skills or special tools. The most important thing for making results like this easier to achieve (in any harness) is using the best current models. Right now that's Opus 4.8 and GPT-5.5.

They're only good at it because that's what they're good at? Come on.

They are metered. That's why their ARR went from $9B to $45B in 6 months.

I must have thought I wrote something that I didn't actually write in the previous comment — I can't figure out what "as I said" is supposed to be about.

In any case, maybe I was too subtle. I was talking about Mythos, a model that continues the trend, but which is not available to the public yet. The "overwhelming evidence" is the testimony of the people who have used it. The irrational skepticism was people who don't believe that testimony. In other words, we do know the future, because we know that model and others like it will come out soon.

And there's a whole lot more evidence than that!