Hacker Newsnew | past | comments | ask | show | jobs | submit | niea_11's commentslogin

I think that comment is the reporter sharing anthropic's response and the gif is his reaction to their response


Oh! Yes, stupid assumption of mine. Thanks for catching it.



They have the deep pockets to either win the lawsuit (fair use) or pay the authors.


Every lab has been sued whether they released training data or not.


A video by Le Monde (french newspaper) showing that you can have an ai song ready and distributed to several platforms in less than one hour : https://youtu.be/G3d8wBOLS2c?is=meGcHQZtpcRFRUsj

(The original video is in french but it has an autodubbed english track)


I'm confused by the last part saying that if "weak" models (like gpt oss) find the openbsd bug they are just hallucinating. and also stronger models not finding it is because they dont hallucinate but are not strong enough.

AISLE demonstrated in the last few weeks that small (weak per the author) models can find the openBSD bug (when pointed at the code). And apparently did several runs with the same results. Was gpt oss hallucinating on all those runs?

And what separates a strong model from a weak one? Is qwen3.5 27b weak?

Don't trust who says that weak models can find the OpenBSD SACK bug. I tried it myself. What happens is that weak models hallucinate (sometimes causally hitting a real problem) that there is a lack of validation of the start of the window (which is in theory harmless because of the start < end validation) and the integer overflow problem without understanding why they, if put together, create an issue. It's just pattern matching of bug classes on code that looks may have a problem, totally lacking the true ability to understand the issue and write an exploit. Test it yourself, GPT 120B OSS is cheap and available.

BTW, this is why with this bug, the stronger the model you pick (but not enough to discover the true bug), the less likely it is it will claim there is a bug. Stronger models hallucinate less, so they can't see the problem in any side of the spectrum: the hallucination side of small models, and the real understanding side of Mythos


The env variables names are misleading. They don't require api keys to OpenAI. Apparently, their tool can connect to any open ai compatible api and that's how you configure your crendentials. You can point it to openrouter or nebius.com to use other models.



It's possible that the people complaining now are not the same people that complained about apple's lack of rcs support.


The same thing for Morocco, it started being called Kingdom of Morocco only after the end of the french protectorate in the 50s.


> in the US, whether you are permitted to enter and how long you may stay are determined by the person at the border

I don't believe this is specific to the US. The agent at the border always has the last decision on whether you can enter the country even if you have a valid visa and travel documents.


Border guards in the US tend to be more strict than other countries though. I’ve heard plenty of stories of friends being refused entry into the US, and I don’t personally know of anyone who was rejected coming in to any other country.

It’s devastating when this happens because not only are you thrown out of the country on the next flight, but once you’ve been rejected at the border once, you can never apply for a simple online visa permit again. For the rest of your life you need to go to a consulate with your passport like this article describes.

I was almost denied entry once when I visited Hawaii. The immigration officer heard I was on vacation for a month and had no cash on me (who travels with cash?). Turns out he was worried I’d end up homeless. He didn’t realise that as an Australian, I get paid while on vacation and I had plenty of money in a bank account online.

This isn’t (just) a race thing. I’ve heard plenty of rejection stories from white, Australian men.


The US also has some quite byzantine rules. For example, let’s imagine I fly from Europe to Canada, with a 3 hour transit in US. I stay four months in Canada because I’m a UK citizen and Canada allows me to stay up to 180 days. I return to Europe using the same route via US.

If I did that I would have overstayed in the US. Because the US only allows a stay of 90 days under the Visa Waiver program, and even though I only spent a few hours in the US, the four months in Canada counts towards the 90 day limit. The US also doesn’t have airside transit like most countries.

The correct thing to do would have been to apply for a US transit visa so that I am not entering on VWP, or to route my flight so that it doesn’t transit in US.


I've also noted that issue. I know of several people at my former job who traveled to the San Jose office and were turned back for some inane reason that had never been an issue before that time. US Border Guards have extreme latitude and they seem to like their power a bit too much.

Problem is that a single rejection forces you to go through the Visa process instead of the Visa-free process for the rest of your life: one of the questions on the entry form is (or was) "have you ever been denied entry", next to "are you a nazi war criminal" (not a joke). And answering "yes" on the former guarantees a whole lot of pain. If you answer "no" you're committing a felony.

The USA's entry system is inane, utterly dependent on the person in front of you and downright scary to me.

I've known conferences to move outside the US for this kind of reason, the pain was just too much and too many people couldn't get in.


It also occurs if you visit countries the US doesn't like such as Cuba or Iran.

Visited Iran when sanctions weren't in place, no ESTA for you. You get treated like a criminal, and have to pay more money for a proper visa.


Something I've started mentioning, is the US immigration and customs is also pretty rude to their own citizens. Coming back to the US has been the far more stressful border crossing in my experience.


The files in the repo are a list of magnet URIs which contain in this case the hash plus the title of the content.

Something like this :

    magnet:?xt=urn:btih:BSXYFTH5VCJLB32DYS2FPTJUB2IVNKUS&dn=Fedora-Astronomy_KDE-Live-x86_64-37


Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: