> First, we aim not to release open weights that would extend the absolute frontier of cyber capability. K3.1 Pro's offensive cyber capability is, by design, below what our pipeline can produce, and remains behind that of frontier proprietary models. Second, we intend to progressively raise the cyber capability available in open weights.

So they admit that they could do more, but they won't?

2 days ago I wrote a piece about how we think about jobs: starting with translators, then programmers then mathematicians, see https://rudolf.observer/en/posts/i-was-not-a-programmer-afte...

Not to mention how naive it is to think UBI would work.

But - we do have similar (and real) science like psychohistory: https://en.wikipedia.org/wiki/Cliodynamics

Comments URL: https://news.ycombinator.com/item?id=50116586

Points: 4

# Comments: 3

This is rather just another mechanism to control people - by creating a dependency from entity which is providing this kind of support.

Please don't!

If you sweat a lot, wearing polyester is the worst decision you can make!

Not only it makes you sweat more, but it also smells really bad, as compared to wool or even cotton (or linen if you live in a hot area).

The author writes: "This time I do not need the research, because I have spent the past year looking for a house. (...) And the Polish loan is the worse product."

I wrote a piece comparing what Poles and Americans say about the housing crisis in social media, you might find it interesting, especially that I tried to compare affordability as well [0].

In short, it seems that:

- Warsaw is even slightly less affordable for its residents than Manhattan - but the Americans seem to be more pissed off by the situation.

That was very surprising to me.

[0] https://rudolf.observer/en/posts/who-complains-more-on-housi...

One of the best books and simply the best imho in its genre.

Alec from Technology Connections[0] about a year ago put out a video detailing frustrating trend: many modern DVD box-set re-releases of popular series are being quietly downgraded in quality.

Basically publishers are taking content originally authored for dual-layer discs and cramming it onto cheaper single-layer discs to cut costs. The result is aggressive over-compression, terrible macroblocking, and sometimes outright playback issues.

It really sucks: you try to do the "right thing" by paying for a legitimate, physical copy to own your media and you end up getting ripped off with crap.

[0] https://www.youtube.com/watch?v=Hzz_d9Y44ZE

Care to explain how it would work in practice?

Speaking from Poland, we see the same outcome: doctors rushing patients out after 5 minutes. The reasons may be different, though: since COVID, many doctors here have, in my view, become more arrogant and focused on money. Over the past few months, an uproar over doctors’ pay has swept through Polish social media.

Reports have emerged of doctors billing for overlapping work under multiple contracts. In one case, a doctor’s records showed 72 hours of work in a single day!

Meanwhile, the Polish Chamber of Physicians and Dentists (NIL) continues to defend caps on medical school admissions, adding fuel to the national debate.

Following this logic, any broken device is equally interesting as its behavior is sometimes hard to predict, isn't it?

Maybe they are sapient - you know, like the dolphins from the hitchhiker's guide to the galaxy :)

I wasn't aware of that, so I checked - and indeed, there are some firm regulations there, just not exactly the one you're describing:

- a new generative AI must pass a security assessment and be filed with the "Cyberspace Administration of China" before launch, tested against GB/T 45654-2025, which has hard numeric thresholds:

training corpus ≥96% compliant on manual inspection (≥98% technical), sources with >5% illegal/harmful content can't be collected at all, generated content ≥90% pass rate, ≥95% refusal on questions that must be refused, ≤5% false refusals.

The moment in which hallucination factors into this hard 90% threshold is Risk A.5.a ("inaccurate content grossly inconsistent with scientific knowledge"). The standard explicitly states however, that this risk applies when the model is used for specific service types with higher safety requirements, such as critical information infrastructure, automatic control, medical information services, psychological counseling, and financial information services. Everywhere else, accuracy is a best-effort obligation.

source: https://cset.georgetown.edu/publication/china-gen-ai-safety-...