newtonsmethod

Yes: https://docs.mistral.ai/models/mistral-large-4-0

You can try it on their website, https://console.mistral.ai/playground

It's just that the open-weights aren't yet available (although the long delay is slightly annoying).

I have made no comments on this thread before the one above.

So saying you said something because my non-existent position is "absurd" and "clearly false" indicates massive levels of confusion.

You then go on to imply I haven't read the article, using an acronymised swear (I'm not a pious person but clearly directing a swear to someone makes a message come off as aggressive), even though your understanding of what I said is based off non-existent messages. (This is quite ironic).

Flock is a company. Companies are a type of organisation: it makes no sense to describe a company as AI, as AI is a type of computer system. It may make sense to say Flock is an AI-based company, but calling them "AI" in itself makes no sense.

I can't find anything indicating that an LLM was used in this case, and this county seems to mainly have Falcon cameras (https://www.scribd.com/document/983922612/250730110407Flock-...). Reverse engineering seems to indicate these are powered by non-LLM AIs like Tensorflow vision NNs. These have been around for ages, and notably no-one is alleging these models hallucinated etc.

People also (perhaps wrongly) use "Flock" to refer to the camera system. CrendKing explicitly calls "Flock" a "camera system". A system of cameras is not AI, and the point of the comment seems to be that it's inappropriate to bring up LLM related AI fears about this system.

I don't think ANPR systems should be held responsible for matching number-plates or car models. That's their job, and police have had access to such tools for many many years. It seems relatively clear that Flock was providing a commonplace technology here, that was misused.

What a weird thing to say. The headline is supposed to be the takeaway of the material. People are certainly allowed to discuss an article's choice of headline (which is what people remember), and should give no care to you telling them to go elsewhere.

I very much agree with what you say on automation bias.

Certain technology just has too much potential to be misused / people repeatedly prove they can't use it properly. It also comes with risks of making authoritarian takeover easier.

However, here, it seems they used Flock a bit like a regular ANPR. The actions they took independent of Flock show there are better stories to use to criticise this technology.

How is Flock responsible in this case?? There may be cases where Flock is responsible, but I don't see it here.

Flock is not an AI. That they have AI features on certain products does not mean "Flock" is an AI. That's the fallacy of composition.

Flock does have AI components (this one seems LLM based). Basically every ANPR system will also use some level of AI.

Also, on "reality would like a word". Why not just speak normally instead of using what seems like a high-school grade comeback?

The main Flock related issue here is that officers treat results that come out of the systems as gospel. The same thing has happened with face recognition systems. There is also the broader issue of access controls to Flock, 5A, and whether places should be building up a technology so powerful (surveillance infrastructure can easily be used for evil).

What a strange strawman. You are deciding now to substitute a different complaint than what was originally sense, because it is simply true that if you don't want to use Apple Intelligence, you don't need to.

The complaint now shifts to that you'll still have files on your system, or can't control exactly what models are stored on your computer. Apple has reasons not to give this fine control: what you have on HN is a group of people who are vehemently anti-AI, use Apple products, and still want features very specific to their control of the system.

Apple caters to the general user. If they think a setting is going to be confusing or likely to be misused, they won't implement it. People know the reality that if Apple intelligence is successful and if this feature is implemented, the general user won't know which parts of their workflow use which features. Removing the models will be advertised as a cheap hack to save storage, and people will be confused when they can't turn intelligence on again.

The fact that the core premise of the article (repeated in the title and first sentence) is about "phoning home": > The nice thing about keeping a blog is that I can say at precisely 9:51am, Wednesday 5th February 2025, I discovered that macOS 15.3 had enabled a feature that was phoning home every 15 minutes with personal data. And that I said “no” and turned it off.

The entire basis of this complaint is not understanding the difference between "duration" and "interval". And not being able to click the link that is given to you in the settings menu or just search Apple's docs for the setting ( https://support.apple.com/en-mn/guide/mac-help/mchlfc0d4779/... )

This is literally the main point of the article: "The nice thing about keeping a blog is that I can say at precisely 9:51am, Wednesday 5th February 2025, I discovered that macOS 15.3 had enabled a feature that was phoning home every 15 minutes with personal data. And that I said “no” and turned it off."

I assume they're using a more candid definition where they're not counting all the countries a company may be based, but rather the primary country they're based in.

I don't think they're trying to flex this as a large number. They don't want to give an exact number, as that may change etc / is fuzzy, but also want to give you an idea of the scale.

They say "In the future, we intend to expand our geographical reach much further". I imagine this commentary is somewhat related to the concerns that AI will create an even worse "global underclass". AI developments are first accessible to Americans, then allies, and then later the whole world.

There's a comment on the GitHub thread which also mentions pinning rsync version would be a bad idea. Many of the people affected by reversions are those with workflows vulnerable to the prior CVEs.

I don't actually see much evidence the usage of AI here was an issue. I think you can obviously identify areas where the code isn't perfect. I'd blame this slightly on human prompting, slightly on AI.

But I'm not a sure a human on their own would've done better. There aren't enough resources to make the changes required.

Can you tell me any of the bugs? All claims I have seen so far of people being affected by bugs seem deeply implausible.

The current ones I see are: * Can't build on Linux < 5.6. But people aren't complaining strongly about this, since it requires a build from source / isn't a result of an update from a distribution (and people can wait for updates / implement them themselves). * An issue hit with chroot false + daemon. People running this in an automated manner (as some claims suggest) are already using it in a way fairly likely to be insecure, chroot false is strongly discouraged here, and the whole point of these commits was that they fixed CVEs impacting precisely these people.

> Just because you got shat on the head once it doesn't mean it's fine to be shat on the head every day now.

This happens frequently when there are fixes for CVEs, since regression tests can't catch many things which incremental rollouts can. It happened for example in 2025. People are right to imply the reaction is totally outsized here, and it's almost certainly the case that people are overreacting to AI (rather than the somewhat weak idea that it's because of the frequency of these issues).

What's really funny is that the people opposing AI here are showing far less literacy than those who wrote the code. People claiming to be affected by this bug severely are likely exposing themselves: * One bug is for Linux < 5.6, where it didn't hit distributions. This is a low severity bug where it can't build, where distribution maintainers may be reasonably expected to fix it themselves (although rsync will also fix it eventually too).

* The other bug affects precisely the people impacted by the CVE https://github.com/RsyncProject/rsync/issues/897

You probably don't want to revert in this scenario. If someone is hit by this bug and is running rsync in an automated manner, it is highly likely they're ignoring very many security practices: you're usually not supposed to run native rsync in an automated manner (the main use case is for public users, where you can't SSH etc; since it's unencrypted, you're supposed to check a checksum against a website etc); these cases are hit with chroot false, which is deeply discouraged and leads to far larger attack surfaces.

Yes, the actual issues seem to be:

* People with linux < 5.6 can't build this from GitHub. This to me seems like a fairly minor regression: people using maintained versions of 5.6 (mostly extended security) will have distro maintainers pick up that the build is failing, allowing for it to be corrected in a timely manner.

* Hardening against path-traversals causes failures for users with: no chroot; using the native rsync protocol. Ironically: chroot = no is deeply discouraged; you shouldn't really be using native rsync in an automated manner (and perhaps it seems I wouldn't advise using it at all); the CVEs the commits fix apply exactly to this use case.

https://www.cve.org/CVERecord?id=CVE-2026-29518

Requires daemon + no chroot. " daemon runs with elevated privileges. This vulnerability can only be triggered if the chroot setting is false."

So the workflows affected are those which are the most vulnerable, and yet people are recommending that people revert versions.

* Furthermore, if a regression test picked this up, it would've been written previously.

Many of these results don't rely on repeatedly calling Lean. You have no clue what you're talking about.

> Just look at the donors of Renaissance Misanthropy. If you're actually interested, who funds each project is listed in the PDF here. https://www.renaissancephilanthropy.org/annual-reports

As you can see, it's mainly philanthropic projects of wealthy families.

This is getting insane. You have no evidence for your initial claims and didn't respond to a thing I said, and are now claiming using AI for education is "child abuse". Please get help.

Is your justification in dismissing Fields medalists that they are impressed by funding? Not even receiving it (I assume you say this because Tao is not funded by AI for Math, but rather an advisor for it)?

Not only would it be a leap to suggest that people automatically lose their integrity by taking funds for projects they believe are useful, especially after involvement with adjacent fields, but you are suggesting merely being impressed by a fund is enough to dismiss their views?

You also have no evidence that Renaissance Philanthropies is a front for VC companies. All news coverage indicates that they seek to be an alternative for high net worth individuals engaging in philanthropy.

Many people discovering Erdos results, engaging in Olympiads etc, are doing so with publicly available models and publish the resources used in the process.

Are we back to magic water and magic soil? Does the energy have some morality attached to it?

The emissions per kWh of energy used in providing internet downloads probably is similar to that per kWh of energy used for washing clothes.

I have to tell you something: there is consumer demand for AI.

Unless they mean another "f word", the f word was used in Apollo 10. It's just that the transcripts change it to "freaking": > Oh, shit' What they did, they made it a two - Ain't this smart' Ain't that a smart freaking sack.

( https://apollojournals.org/afj/ap10fj/as10-day2-pt9.html )

It can be found in this video: https://www.youtube.com/watch?v=DQVEsfa15SY

Have you read the DPA? How did you come to your inclusions on its intent? How do you think Biden stretched the definition here?

There is nothing in the DPA implying companies it is applied to can't be acting in a hostile manner, or that it can't be applied when security interests of the US are being threatened. Of course they have no reason to state such a distinction repeatedly in law (claiming it doesn't apply to adversarial companies....), but 50 USC 4566 applies clearly when acts are being made against national interest (this pertains to foreign investment, which isn't the nature of the Anthropic rift, but shows clearly the DPA contains laws with intent of preventing adversarial action against the US).

Even without knowing the intent of the behaviour, it should be quite clear that companies that are vital to national security are more likely to be supply chain risks. Amodei's direct words were: > These latter two threats are inherently contradictory: one labels us a security risk; the other labels Claude as essential to national security. Being a security risk and essential to national security are not "inherently contradictory".

Are you reading things before agreeing with them? Or thinking about them? It doesn't seem obvious these things are contradictory at all. That Politico reports so doesn't make it the case.

It is clear that the DPA can be invoked for companies posing risks to national security:

> On October 30, 2023, President Biden invoked the Defense Production Act to "require that developers of the most powerful AI systems share their safety test results and other critical information with the U.S. government" when "developing any foundation model that poses a serious risk to national security, national economic security, or national public health."

Furthermore, it should be quite obvious that companies very important for national security can act in manners causing them to be national security risks, meaning a varied approach is required.

I have agreed tariffs will have an effect, but I'm not being myopic.

Lot costs, builder profits, indirect labour (commissioning, financial and legal affairs, advertising) all are far less affected by tariffs. Machine costs could make up 15% of your "labour and material" cost but depreciation and repair purchases are still only 30% of this, with of course not all of this affected by tariffs.

It seems wholly reasonable to believe that the long term effects of a tariff policy like these on housing costs could indeed be in the ballpark of 5%, as I claim, because in fact housing development is less affected by this.

I'm not sure if you trust this for consensus, but you could try asking an AI to give an estimate of the long-term impact for you. Here's what Gemini 3 Pro said to "Estimate the increase Trump's current tariffs, if long term, would have on price of new housing developments."

> Total Home Price Impact: This translates to a roughly 3% to 4% increase in the final purchase price for the consumer.

I agree with you, they'll probably worsen the effect, and I assume this is maybe 5% in the long run from all these tariffs (if they're not cancelled).

The claim they went into effect January 1 simply does not seem true.

The finished products tariff was delayed: https://edition.cnn.com/2026/01/01/business/trump-furniture-... For unfinished lumber, I don't think there was any tariff going into effect January 1.

Sure, these tariffs may further increase the house of prices (e.g. be relevant to 15% of the cost of the house, with tariffs ranging from 20% to 50% and sources of materials adjusting to these tariffs), but the say 4% future effect of these tariffs is likely less than the effect of zoning laws, other development restrictions, and rent freezing.

I think you're getting your facts wrong: https://edition.cnn.com/2026/01/01/business/trump-furniture-...

"President Donald Trump has delayed new tariff increases on upholstered furniture, kitchen cabinets, and vanities for a year, pushing their implementation to 2027, according to a White House statement."

I agree with the first half of what you posted, but immediately jumping to blaming tariffs in your last paragraph seems weak (and a slight attempt at a gotcha).

Concrete, gypsum and steel are primarily domestically produced. Similar goes for wood (although a substantial amount is imported, e.g. from Canada - the tariffs range from 25% to 50%). Labour & Materials may make up say 60% of the cost of a house, but only 50% of this is likely materials, with likely a minority of the materials tariffed.

What is likely to actually reduce rent and house prices is making development permission and laws more lax, as well as preventing rent control.

A nonpartisan newspaper could also condemn or blame a political party for an action. But if all of its posts were supportive of one administration, it would no longer be partisan.

You can just look through the old white house accounts. For example this tweet https://x.com/WhiteHouse46/status/1879171105044181097 , "While Congressional Republicans refused to pass a bipartisan border security agreement, President Biden took action and encounters today are the lowest since July 2020."

Looking through the tweets, you'll see it's not nonpartisan and isn't supposed to be.