News for hackers. Stuff that matters.

tensor

Re:Postgres 19 is late, and it's AI's fault on Monday November 02, 2026 @02:30PM (#50342320)
The one thing that cuts the other way is code churn. Third party libraries may keep adding features you don't use, and those may expose new bugs. You can pin an older version, but unless the library back ports security fixes, you are not going to benefit from the shared development wrt to security.

If you spend a ton of time hardening your own library code and don't have much feature sure, then that can be a win.

Re:Serbia just elected the world's brainiest parliament on Saturday October 31, 2026 @11:51AM (#50317287)
What anti-intellectual horse shit. Carney has a PhD in economics, for one. But there are many: Tomáš Masaryk, Manmohan Singh, Fernando Henrique Cardoso, Ricardo Lagos, Mary Robinson, Lee Teng-hui, Tsai Ing-wen, Mario Draghi, Angela Merkel, Claudia Sheinbaum.

All these people have been been at the forefront of b ringin freedom, democracy, and well being to their populations. If anything, the "fun facts" show that academics can make great leaders.

Re:We're releasing the weights of Mistral Large 4 on Thursday October 29, 2026 @10:29PM (#50299161)
Yes, that is an option, but not the only option. Mistral also provides an api that is hosted by an EU company as far as I understand.
Re:Concerning behavior from a popular open-weight AI model on Monday October 26, 2026 @03:57PM (#50250011)
Uh yes? Definitely wouldnt use for anything sensitive.
Re:Lula wins election after final votes tallied on Sunday October 25, 2026 @11:44PM (#50238848)
None of the things you mentioned have anything to do with whether or not a candidate is fascist. No “mental gymnastics” are needed.
Re:We should be more ambitious with the software we build on Saturday October 24, 2026 @09:36PM (#50227377)
I’ve been using the latest agentic frameworks and have come to realize that the discourse around vibe coding and slop really needs more nuance.

There is a gulf of difference between someone who doesn’t understand how to program “vibe coding” software and an experienced software developer using ai to write code with heavy direction, review, and best practices.

Today both would be said to be “vibe coded” but we really need to differentiate them. Also, if everything I’m hearing from people is true, it sounds like most corporate code is now completely ai written in the latter way I described. It’s interesting we don’t hear commercial software being called slop yet. Maybe it’s coming.

In any case, ai is far more capable than I realized. Not sure what to make of it.

You should leave and revoke your citizenship.
Re:No More Icons in Every Menu by Default on Friday October 23, 2026 @05:08PM (#50215346)
My father can't even use the apple tv box I got him because the UI is so difficult to see. It's maddening.
Apple in fact did withdraw their security features in Britain over a similar law. It's not simply crying wolf, these companies will withdraw.
Important services have already said they won't. It will be a disaster.
Indeed. I've also sent several emails to my MP. Sadly I've received no response. :-/
Re:Every AI model has the same "controversial" take on humans on Thursday October 22, 2026 @06:22PM (#50201893)
They don't really need to, most are trained on roughly the same data: everything they can get their hands on.
Re:Quad9 (DNS Resolver) to Exit France on Thursday October 22, 2026 @05:50PM (#50201534)
These days you can land in jail for simply bad mouthing the government. US needs to strengthen its adherence to their constitutions. Words on paper are useless if they aren't followed.
Re:No More Icons in Every Menu by Default on Wednesday October 21, 2026 @11:41PM (#50190931)
The latest version has already removed most of the worst of the changes, if you set your settings right.
Re:No More Icons in Every Menu by Default on Wednesday October 21, 2026 @11:30PM (#50190857)
The Apple TV one drives me insane. The selected button is so subtly raised that it takes a lot to see. Usually I need to move it back and forth and look for the motion.
Re:DeepSeek just open-sourced its inference engine on Tuesday October 20, 2026 @11:56PM (#50175914)
The new Mistral large is already competitive and it hasn't even finished training.
The latest Mistral preview already matches Deepseek and it's not even finished its training yet. But please do continue.
State support has nothing to do with anything. Also, in practice the EU has its deepseek equivalent model in Mistral.
Re:Hello Toronto – Waymo Is En Route on Friday October 16, 2026 @05:48PM (#50125207)
You will never stop hearing about this until they are restored. Also enjoy your very slow commute with all the bikes now legally required to ride in the car lane and block you.
Re:Tesla Roadster on Friday October 16, 2026 @02:59PM (#50122856)
> If you care about the environment you get an old VW or ideally a Mercedes that runs on used vegetable oil.

This is not realistic nor scalable. The reality is that the US and Canada were built around cars, and even if we had huge transit projects there would still be a need for roads and vehicles due to the size and rural/industrial areas.

Lithium batteries can be recycled. All the studies have shown that electric cars are a huge net win for the environment, and the cries about the environmental cost of building the cars in the first place don't actually match the reality. Especially when you consider that future ev ecosystems should and hopefully will have a strong battery recycling operation.

Re:Tesla Roadster on Friday October 16, 2026 @02:49PM (#50122688)
> - no emissions

Can we just say it out loud? Because it's good for the environment and we don't want our children or their children to have an unlivable planet. This should probably be the number one reason, but yes EVs are better in many other ways too.

Re:Just say no to OpenAI and Anthropic on Monday October 12, 2026 @05:05PM (#50063069)
Sure, but assuming you're on the expensive 200usd a month plan with claude, 5x is 1000 USD a month. I mean, even if you're running 10x agents you're still only hitting $300 US a month, which is not close to 5x.

Anthropic doesn't really publish their limits so it's hard to do a direct comparison when these things come up. I just have doubts that DS 4.1 Flash is remotely close to 5x as expensive. Also, I've seen a lot of people claim that switching to it cuts their costs by a huge amount, which based on API per token costs of Claude seems to be far more likely.

Re:Just say no to OpenAI and Anthropic on Monday October 12, 2026 @04:49PM (#50062876)
8.4M tokens cost me 0.26 EUR on GreenPT. 10 USD is 8.96 EUR. Based on my usage data that would be 289M tokens a day. Anthropic doesn't publish the token limits per plan so I can't compare, but that's quite a lot of tokens.

For my usage almost all of those tokens are cached input, which is why the cost is so low.

Re:Just say no to OpenAI and Anthropic on Monday October 12, 2026 @02:38PM (#50060834)
Using Deepseek v4.1 Flash on something like GreenGPT is probably about a euro a day if you use it all day. Hard to see how that is 5x more expensive.
Re:Just say no to OpenAI and Anthropic on Monday October 12, 2026 @02:37PM (#50060817)
The level of astroturfing and hate on HN whenever a non-US AI story is posted is really quite shocking. HN has always had a US bent, but it's gotten to the point where you can't even have a technical discussion of non-US AI because it will be drowned out by a very vocal anti-everyone-else-but-the-US crowd.

But yes, great piece. It really captures why I use open models on european hardware. It's like we haven't learned by now that competition is good and we should be encouraging a diverse and open ecosystem.

They don't need to equal in badness to want to avoid them both. But I'll note that at least China still believes in science, on that front.

But even from a business standpoint, the US is not reliable. For all I know Trump will ban countries he doesn't like from using US AI just because. And there is plenty of evidence that at this point in time things you say bad about the US administration can cause them to attack you. I have no confidence that they don't have access to e.g. openAI.

Not that I'm using AI for that mind you, but you put it all together and it's just not worth it if there are good alternatives. Which there are. You can use even use Chinese models but hosted by European countries with privacy as a selling point.

Altman is definitely very Trump-aligned, which to me is enough to not trust his company on data security. But also, there have been many many accusations that they potentially spy on users. E.g. stealing proofs from mathematicians and such.

It's just not worth it to me. And the way the US government is increasingly using data it collects to attack people who simply say bad things about it, there is no reason to doubt that such Trump-aligned companies would help ex-filtrate data.

I have fewer issues with Antropic, but also, from a business standpoint it's too much of a risk. Who knows, maybe the next tariff will be on software, maybe he'll ban random companies from using US AI. Too unpredictable and too risky.

Based on the latest news yesterday it seems that the law would require them to implement more metadata collection, which they are refusing to do. Not sure what will happen if this actually passes. Wonder if we will be able to use foreign hosted VPN services and use Signal through those.
Some of us don't use US models for the same reasons. It doesn't leave much out there aside from Cohere and Mistral.
That's not quite correct. It requires the collection of available metadata only. It does not require that encrypted messages be decrypted, unless the provider holds the keys rather than the user. Still, it's not a great bill.

And I guess VPNs are not end to end encrypted. The provider would still be able to access your data if they want to unlike something like Tor. So, it will be the death of privacy focussed VPNs. But presumably things like Signal and Apple's disk encryption will be fine as only the user holds the keys.

Older »