computerphage

I read it as "was posted earlier this year. It claimed that compute has been doubling every seven months for quite some time, on average"

What do you mean by "folded during wartime"? When did wartime start, and who folded?

Why do you agree?

They claim it has a very low false-positive rate, which is meaningfully more nuanced than "accuracy".

"Detect AI-generated content with 99.98% accuracy." -pangram themselves

https://www.pangram.com/research/how-it-works

In my view, you're the one claiming to be able to tell the difference! You said, "You tend to see that more..."

...which I took to mean that there were threads where you can look at the comments and categorize them into the two categories and then assert that the size of the paid comments category was larger on posts for high margin than it is for low margin.

Like, it makes sense that the incentive is stronger in that case, but that's not what you said. You said "you tend to see that more"

Palisade Research demonstrated this capability, including fully copying model weights, on May 7th, 2026

https://palisaderesearch.org/blog/self-replication

https://arxiv.org/abs/2605.06760

Do you?? Show me an HN thread where half the comments are paid for

That strikes me as a really poorly calibrated expectation. Who benefits? Who's paying? How could I get this job?

This is AI written, according to pangram, and also obviously

Show me a graph of your javelin skill doubling every six months and I'll start asking myself if you'll be the next champion

But how do you explain the fact that, before it got bad, it wasn't /already/ bad. You've gotta have some model for why there was a change if the rate wasn't previously at 100% and the claim is that the number "should be 100% unless it was regulated"

It didn't used to be regulated. How do you explain the fact that not everyone used to be an opiate addict?

Barely passing body-text contrast in dark themes

This has been killing me recently. Apparently I need slightly higher contrast than some people, and these vibe coded UIs are basically unreadable to my eyes

Yeah, when I'm writing code I try to avoid zeros and ones, since those are the most common bits, making them essentially noise

Do you? What's the technical detail here? Why can't you get the model's prediction, even for that first token?

Related: https://www.transformernews.ai/p/ais-next-big-blue-battlegro...

OpenAI wants to not be responsible for "accidents" that kill more than 100 people, despite some critics arguing that their current actions are likely to cause such harms.

Why are people dangerous? You can just not listen to them.

Why isn't the AI story believable? It seems to me that AI is getting more and more productive

What's your point?

The usual thing is that the market ends up around $0.95 for things like that, if the actors are all solid investors. It only takes one overly enthusiastic yes buyer to break that ceiling, the smart money won't "correct" it down to $0.95

There's another idea, which is make contacts that pay out in shares of an ETF, but I haven't seen this idea put into practice

Why do you think they have not trained a new model since 4o? You think the GPT-5 release is /just/ routing to differently sized 4o models?

Yep. Totally agreed that it's well defined. Only pointing out that the technical execution required will shift, which seems relevant because it's likely to make it take much longer than without this effect

I think "blockbuster movie" is a moving target, so it's a bit hard to know

Because before buybacks there were dividends. Did the difference between buybacks and dividends really make the difference between doing basic research and not?

Like, there is an argument that can be made here, but "there's just not enough money in the world to justify this" definitely isn't it

Indeed. There are trillions of dollars /per year/ paid to workers in the US alone.

"due to privacy concerns about privacy"

This strikes me as a particularly funny typo

"Sometimes" and "sense" are both wrong. I don't think this library is very good