llm_nerd

I have no idea why you're being downvoted. HN has a fetish for SIMD, but if you are hand-rolling SIMD and you aren't writing an explicit acceleration library, you're doing it wrong. Like, 100% of the time.

Every modern language has a vectorization optimizing compiler, and through some fairly straightforward techniques this is automagic. And contrary to the various replies, unless you screwed something up compilers are really good at vectorizing on whatever hardware you're targeting, including SVE.

EUV was "invented" (demonstrated) in Japan first, though those US labs did a lot of work on the overall lithography system and reflective optics to put it into practice. The US DOE along with the manufacturers made a consortium (EUV LLC consortium) to commercialize/scale that Japanese technology because they could obviously see the promise, and Cymer is a member, as was ASML already via SVG.

Simultaneously Japan created its own initiative (EUVA), and Gigaphoton didn't rely upon the US initiative at all. And again in an alternate universe -- one where Japan isn't protectionist and Cymer wasn't out of financial runway and desperate -- we'd be talking about how ASML is really Japanese tech, etc.

The actual EUV light source is developed and produced in California by Cymer, which ASML acquired in 2013.

Cymer was one of multiple vendors (Gigaphoton, XTREME) providing this sort of light source, and ASML acquired it because it couldn't achieve what they wanted (they needed 100+W and Cymer could only achieve 15W). ASML needed to control and fund the company, and add engineering talent, to achieve what they actually needed.

Oh and ASML bought Cymer because Gigaphoton (another company at basically the same stage of development as Cymer), being a Japanese company, objected to foreign control. Though if ASML knew the protectionist, insular direction the US was going to go they probably would rethink their past decisions.

So this retconning that Cymer is really the secret magic of ASML and the Netherlands is just stealing US ingenuity is hilarious stuff.

And for that matter, it is amazing how so often we hear about "American" tech that is actually overwhelmingly combining acquisitions and manpower from abroad. Like is an iPhone an "American" creation? LOL, no.

No they didn't.

Yes, they did. Like, look around. Clearly the Western world foolishly and very short-sightedly allowed the US the reigns on far too many things, to its disadvantage.

It is unwinding, but it turns out that having decades of intertwining takes a while to undo.

Also anyone that doesn't like the US is free to stop using any American website, device, or service.

What an idiotic, useless bit of pablum to throw in there. Back to 4chan with you.

This is complete nonsense.

https://www.axios.com/2026/07/20/ai-us-china-open-source-kim...

I don't know why you're trying to downplay it.

Ignoring that I have literally zero trust in anything Anthropic has to say on this -- they have been doing the hysterical routine and trying to get every bit of government granted monopoly they can[1] -- those numbers still simply aren't that impressive.

European models are so far behind because...

What a non-sequitur. Europe, like much of the West, foolishly delegated tech, media, payment systems, etc, to the United States. European efforts on this are poorly funded, poorly capitalized, and marginal efforts.

China is very much not Europe. China is looking to leave the US to the dustbin of history, and their efforts are a little more concerted.

[1] Surely Americans are aware that Anthropic and OpenAI are both very close to getting the US government to ban and fully criminalize the open Chinese models, right?

Models don't have some self identity, beyond what is explicitly handed to them via a system prompt. There have been many, many cases of models identifying as different models by different makers as a basic identity hallucination. They train on enormous volumes of data including lots of people talking about certain makers and models (ChatGPT was actually a super common one given that it became the kleenex of the LLM world). Hence why vendors have to specifically tell it to override that, and if they don't you get lots of funny cases of identity confusion.

This isn't the big gotcha some people seem to think it is, and the whole news cycle about that was mostly by people who have no idea what they're talking about. It's actually a meaningless data point. But it's precisely the sorts of people who think that a few thousand free accounts surreptitiously snuck off with Fable.

The distillation explanation is classic American exceptionalism: No one could possibly do anything unless they were copying American leaders (where "American" means a bunch of Chinese, Canadian, Europeans and Indians working in the US).

It's also a bit of securities defensiveness. Pretending that you really do have a super moat, people just keep swimming in it so you just need to add more alligators.

It's farcical. Anyone who has worked on large models knows that the premise that an almost-Fable model was trained with distillation is beyond ridiculous. It's theoretically possible if they spent tens of billions of dollars on API calls, but it isn't the magic that somehow these people keep convincing people it is.

Previously Anthropic has reported on some Chinese firms doing chicken-shit level of API calls, that at most would be doing some Q and A or final fine tuning. The notion that they're training these models via it is fantastically ignorant nonsense that only very ill-informed and gullible people fall for.

Good article. I think a lot of people in the software development field should ponder over "No. 0002. Preserving problems", because a lot of rhetoric is basically advancing this position.

Software development traditionally is expensive, slow, adapts poorly, and so on. This worked in our favour as it gave us high paying jobs where we could always blame a framework, Microsoft, whatever, and estimate a week to move a button. These people are the ones who are the loudest at filling the silence with anti-AI propaganda because they desperately want to hold onto that problem.

This is a country that has overexpanded in infrastructure which will come home to roost sooner

Man, we've been hearing this same tale for decades. China is always at the edge of some massive collapse any day now. Meanwhile the US goods trade deficit is literally worse than ever, the deficit is $2 trillion dollars and the debt is consuming more and more of federal income just to maintain. The US equities market is a giant inflated balloon and no one even pretends to rationalize it anymore. Leadership in the US is now banana-republic level corrupt, in the open and shamelessly. Yeah, this is not the time to be throwing rocks.

has a terrible demographic structural collapse looming with no realistic way to correct

I mean...not only is China becoming a pretty desirable country for many of the world's residents, they could...have more children. If China decided to turn the corner and get back to maintenance levels, I 100% believe they could do it almost overnight.

The US? Yeah, the US is a free-for-all hyper-selfish, short-sighted end-game right now.

Apple has an architecture license agreement that extends "past 2040" (their phrasing), and it requires Apple to pay ARM ~$0.30 per device sold.

Apple can do whatever they want with the cores, but they do have to pay for the privilege and do have a future expiry to worry about, though it's far enough off that it certainly isn't pressing.

A little bit dangerous for a US administration (any US administration) to do a bailout of unloved companies just before a midterm.

You think there will be a free and fair election? Do Americans realize that Trump has openly floated pardons to anyone in his circle? What do you think all of his "every election that I/we don't win is corrupt" rhetoric -- dangerous, grossly unacceptable, anti-democratic horseshit -- is all leading to?

Trump has done brazenly criminal things, repeatedly. He is pardoning anyone who bribes him. He lies with every utterance from his garbage mouth. He doesn't even attempt to pretend that he's delivering his promises now. Congress has completely abdicated any and all responsibility. His entire administration is just shockingly, unbelievably incompetent, from Epstein-Island Nutlick, to Kegsbreath the ChatGPT warrior weakling dipshit.

Remember how outraged everyone was about Hunter Biden selling a painting, or Pelosi trading stocks? ROFL, how bucolic and corruption-lite that is compared to having a crypto-rug pull, inside trader and his family of halfwit runts running around destroying the US for their own family fortunes.

This Thursday evening is going to be eye opening for a lot of Americans that have tried to delude themselves into thinking they're getting lulz for a couple of years. It is shocking that people still pretend you're a democracy, or even capitalist for that matter. The US is post capitalism, and the plutocrats have decided to be done with this whole democracy farce.

It remains shocking that Americans would re-elect this garbage racist self-dealing criminal imbecile again. And I would like to say "you get the government you deserve", only the US is now a worldwide menace so the entire planet will suffer from this idiocracy.

Trump is currently having an armoured facade installed on the front of the Whitehouse, alongside the very well documented bunker complex. Do Americans really not realize what this is actually for?

Campaign? They're friends of the administration, and the US is firmly in the kleptocracy stage now (the last wrungs of democracy are about to be undone this Thursday evening).

They'll give a bribe to Trump, they'll offer up 5% of the stock to Chairman Trump as the People's Stock now that the US is basically a bizarre oligarchy form of communism, and Oracle will be declared a state enterprise that cannot lose money.

The super rich simply do not fail, and they utterly control every aspect of the US now, exactly as the people apparently wanted.

Americans are in a state of profound denial, but things are about to become very real, very quickly.

Most of them are just their website wrapped into a webview.

Such as? Give some examples, which should be easy given that it's "most", right?

In the linked piece it details one that is so exceptionally trash that it is universally hated. I mean, ostensibly it isn't even allowed by appstore / play store rules, and it's a shit, lazy thing to do.

My thesis is that the standards for web teams were often much, much lower than for app teams. Where tolerance for shit, tolerance for slow and inconsistent behaviours, and so on was just much, much more common.

That is why there was such a fracture. And it's why the "webview wrapped website" is universally reviled trash.

Do websites have to be bad? Of course they don't. But the norms of the realm made users jaded.

Yes, I understand where France is relative to Switzerland. But the basic fact remains that not one of the developers or testers or stakeholders encountered this, and it was so rare apparently the very first person to encounter it -- given that it was literally a blocking issue, and not just some transient issue -- is the linked article.

Is it a regrettable bug that made a poor assumption that breaks in special cases? Sure. Is it "Dunning-Kruger" (groan), or some demonstration of how stupid the developers are? Give me a break.

Isn't this a chicken-egg thing?

Businesses started advertising that they had an app because user's preferred apps, having had so many poor experiences with websites and web "apps" that the entire field was tainted.

This is 100% a "made your own bed" kind of thing. Again, the general standards among web apps for years were terrible, and users became accustomed that using a website on a mobile device was a brutal experience. Things have gotten a lot better, and honestly AI tooling should massively improve the space, but people really need to be honest about root causes.

I mean...a local grocery store advertises their "app" and it's just their website wrapped into a webview, and it is just total dogshit. Because they brough the extremely poor standards they have in their web domain into the app domain, and it simply doesn't transfer.

https://dennisforbes.ca/blog/microblog/2026/05/terrible_mobi...

They're doing the strawman "we evolved with it" tactic commonly found among certain camps.

The problem with their argument is not only that no other part of our lives is similar to our "traditional" lives, evolution is basically done with us by 40. So from an evolutionary perspective, dying at 40 of CVD is not corrected for and ultimately doesn't matter. It's the same with the sun exposure nonsense -- evolution again doesn't care if you drop dead of melanoma in middle age, and it had no impact on pumping out a dozen kids and then dying and getting out of the way. In some ways it was optimal.

By "marketing" do you mean "experience"? Because absolutely no one is marketing apps as being superior, but instead this is just the experience of users.

There's a weird conspiratorial thing that people do about this whole topic that is so easily debunked. For instance "Apple wanted apps more than PWAs!". Android powers about 73% of the world's smartphones, yet PWAs are irrelevant on the platform.

Web apps can be incredibly powerful, but there's just a massively lower bar in the web app domain, historically. Like people are used to the website being dogshit, a mishmash of broken functionality, terrible layout quirks, slow responsiveness, and so on. Because that is generally acceptable to the web community, where it is deadly to an app.

Like I think it's hilariously ironic that the website telling us that the app could have been a website is currently completely broken, unable to handle a relatively tiny amount of traffic.

From a worldwide perspective, it is incredibly uncommon, and just from the basics of this issue it is clearly uncommon in Switzerland as well. The standard keyboard layout in Switzerland is QWERTZ (across all of its language speakers). This website is generally only used by the Swiss.

I mean, if a basic authentication mechanism of a critical government website doesn't work with a keyboard layout, odds are incredibly likely that said keyboard layout is uncommon. Like not one of the developers, testers, stakeholders, or even other users had a problem -- and this isn't a problem you can just ignore -- so it probably isn't that big of a deal.

Read a bit further and the reason they had the issue was their uncommon keyboard layout, which is one that requires holding the shift key to access numeric digits. This broke the input filtering logic.

Right at the outset it opines that AI finding vulnerabilities ("allegedly") is useless because people have to verify those vulnerabilities, which is such a logically braindead bit of nonsense that I knew it was going to be worthless, and somehow it got even worse with every subsequent paragraph.

I get that there are people who are scared, and the change is coming fast and hard and there is a lot of doom and concerns, but good god do people think authoring or upvoting shit like the submitted nonsense is going to stem the tide or change the path? Pathetic.

This particular product used Whisper, so that was obviously the right model to compare it against. Further this is explicitly on device, and Nemotron 3.5, as one example, is 2.5GB for the model.

And if someone were broadly comparing all on-device models (instead of just looking at how this new on-device ones compares to what a specific product uses), Nemotron 3.5's WER are actually a bit higher than what they report for SpeechAnalyzer, for both tests.

I don't believe it. I mean, you might, but the story is fantastical nonsense.

Ignoring that the police "confirmed" suspects that they didn't catch, or that you claim the police would need a "court order" to look at cameras the HOA would own (and which require absolutely zero involvement of flock, with the enormous downsides that dystopian, busted company brings), or that you cited "manually going through footage" like it's 1997 and not 2026 where every security system flags every event -- car, person, line crossing -- with instant-accessible timestamps, the biggest problem with your comment is that gangs do not care about security cameras.

The preventative effect of cameras on such activity hovers around 0%. This ridiculous tale that a flock camera was a magic no crime shield is simply nonsensical.

Again, you might actually believe this. You might have misheard some things and come away with this impression. But it's ridiculous.

I'm a big fan of privately controlled, limited access-guarded and audited cameras, understanding that they're useful post facto to figure out what happened, and sometimes to catch criminals (but catching illegal immigrant SCARY GANG NAME criminals, usually in stolen cars and with masks...lol...utterly useless), but your post is 100% selling the tiger repellent rock, and it's simply incredible if anyone actually falls for that.

Flock's effect on crime has mostly been to increase it, by allowing sexual predator cops to stalk random women.

This is such a bizarre story, each detail more absurd and ridiculous than the last, that I can't believe HN is falling for it.

Honestly there are a number of incredibly weird comments like this throughout the discussion. Is Flock astroturfing every discussion about the company?

I'm saying this as a web developer.

Do you think this gives your claim extra merit? You understand it does the opposite, right?

What it actually achieves is that it makes you look like you're raging against the dying of the light, in utter and complete futility.

Yes, LLMs are especially good at the web stack, given that training data is overwhelmed with enormous volumes of such data. It is the one realm that has been completely obsoleted by modern AI, while a few other realms at least have some resistance.

This feels a bit like a reframing of the rather absurd "Eternal Sloptember" nonsense, desperately trying to pivot from Luddism to visionary (and yes, I know this is the "hacked the iPhone" guy. I'm not a cult of personality person and I positively do not care). Also incredibly weird how it repeatedly talks about people moving to San Francisco, which ... what in the world is that nonsense about? People talking about the concerns of AI and automation are in no universe considering moving to SF as the protection...

"Where’s all this new magical software that the productivity improvements should imply?"

This is a recurring gotcha in the anti-AI marketplace of denialism. It's a bit like saying "I saw a fat guy, so why do people keep telling me that GLP-1s change everything?"

It takes time. Like already I would say just about every programmer has replaced a number of tools with random shit they spit out from LLMs. It percolates out from there as some things become products, etc.

And to anyone actually paying attention, and not just feeding their delusions, the impact is utterly enormous. Incontestable. The "Where's the software? / Where's the change?" people are absolutely going to find themselves in the dustbin of history.

It's also fascinating how often people do the stochastic parrot horseshit.

The other night I had to do a large scale compression test with libjxl, which notably is software that has seen an enormous amount of optimization interest and you would assume has little extra to be eked out. I've traced through this software before and the compression path is insanely complex. It's the sort of software that is headache inducing. Anyways, curious what the state of the platform was I grabbed the latest source and asked Fable to look for low-hanging fruit in the lossy and lossless compression paths. It suggested a few, created a test harness to A:B bitwise compare with the original, and implemented its optimizations. It achieved a 14% performance increase in a single pass, using just the remaining quota I had on a subscription as my week drew to a close. And all it did was some high level logical optimizations, some more efficient memory allocations, and so on. All of its code was completely in the style of the project, was no more significant than necessary, and so on. Anyone that isn't utterly blown away by that -- who gets the hype -- is lying to themselves.

Fable's subscription inclusion theoretically ends EOD today. Anthropic put a wishy-washy "if we have capacity we'll continue it" thing, and given how competitive GPT 5.6 Sol is, and it is included in OpenAI's subscription, I fully expect Anthropic to extend Fable or they will have a serious exodus on their hands.

Competition is good.

What constraint limits this to 5 - 6Ghz? Is it the antenna? Processing?

It's a really neat device, but people should realize that it has a very narrow visibility.

While the article is terrible, what they describe about AI knowing common stacks and frameworks is advice that holds for finding software developers to join your team. Like, 1-for-1. It has always been good advice unless you have a strong justification go in a different direction.

Use standard tools, standard frameworks, standing patterns, standard protocols, and so on, and it's incredibly easy to find highly talented members to join your team, running on day one, and enjoy the progress of the industry. It's quite a different tale when you have some massive internal "framework" monstrosity, have weird patterns and standards, and so on, and these are the sorts of places where you usually find some half-baked terrible custom coding language and so on.

They aren't banning it anymore, they just make it count as "extra usage". e.g. you're paying for every token in addition to your subscription.

Further, the claim that the subscription "version" of the model is worse sounds like bullshit (and the sort of anecdotal nonsense that you see on sites like this). Do you have anything substantiating this?

If you feed their "pure math" question to Fable, in its reasoning it rightly determines that it is the sort of thing you find in phylogenetics / algebraic-combinatorics complexity papers. That is what triggers the classifier.

Anthropic is 100% to blame for fear-mongering, but they said it would be blocked from any biology questions -- even high school level -- and they meant it. If the classifier sees anything related to biology, even in its own reasoning about the question, it blocks it.

Saying it's therefore not useful generally is of course ridiculous. Is it annoying? Of course it is.