spwa4

Really? Weak criticism given that in that time Google was beaten by OPEN models? China is no longer behind Google in AI, Google is now catching up to China and open models. Every linux shop can give their customers better AI performance than Google can, at a cheaper cost than Google's flagship model (either $100k up front and marginal cost only, or about half the cost of Gemini 3.6 Flash).

At least OpenAI and Anthropic can still tell their customers their best model is better than some Finnish student can setup in the customers' basement, even if the difference is pretty small now. But they are better. Google is not.

Surely that's a major change in the situation. "America" is losing the advantage (although even the Chinese models are losing, see next paragraph) but Google is a big step behind 6 other players and 3 open models, so let's just say, Google is effectively behind everyone else that matters. Worse: one of the models that beats Google's best model was trained on Chinese ASICs (that aren't even 2nm. And very few I might add, couple thousand. Google's next model ... has a LOT to prove)

Even Chinese models are losing their advantage. By which I mean that if you check how far Qwen 3.6 27B is behind trillion-parameter models, it's less than a year, down from easily 3 years. If that continues to go down, even Chinese trillion parameter models will become hard to justify (and Qwen 3.8 is being cooked up as we speak, I mean we don't even know if a new 27B is in the works or not, but everyone's very excited)

You still left out that it doesn't matter anymore, just like Anthropic/Facebook/OpenAI only really needed to read massive amounts of copyrighted data only once (and of course, they all did this illegally, which makes their current complaints more than a little ...). Once they have a large model trained on the data, they can just retrieve reasoning traces and copyrighted data from the previous model. In fact that is a training technique long used because it has better results that directly training on the original data.

In other words: even if the US (somehow) denies them access to the current OpenAI/Anthropic models, they'll be able to improve based on what they already have.

Of course, in Europe, we like to have people to buy the CEO position. Ie. a CEO doesn't cost the company money, he (or, sometimes, she) pays to be CEO. This of course then leads to CEOs seeing it as their right to extract a LOT of money, so the investment is profitable. This even sometimes goes in the public sector.

Of course, letting AI do that, seems like just a incredibly bad idea. Not that it's much better with people, but at least, with the average IQ of CEOs you can be sure they're not good at it ...

Yeah. ASML doesn't make the mirrors, but they have exclusive sales rights, exactly as you say. And yes, positioning them, for example, both in the fixed sense (mounting the mirrors in the machine) and in the dynamic sense (rapidly positioning wafers to nanometer-or-better accuracy below the beam, I forgot how quick exactly but it's 10s to 100s of times per second to maintain production rates) is very much not easy either. But it's "been done" (by teams of phds). Producing the mirrors has not been duplicated.

Of course, producing the EUV beam in the first place has also not been duplicated and it's Californian technology.

Intel 18A has a different strategy all together. I mean, it's not like they're saying much about it, but from the little we know, it's very different.

That's what makes this hard to interpret. The whole strategy of intel is that they have their own version of this machine, part of their "18A" design. The only reason they need the ASML machine is because of the "optics" (between quotes because at the frequencies they work at no optics work - only (bad) mirrors do. However you need make the mirrors accurate enough - we're talking far more accurate than the chips, as in sub-nanometer max deviation, they can function as optics with high losses, but low enough that exposure works). No American company can achieve even 10 times lower accuracy at the moment.

So is this giving up on their own version? That would mean their stock price is absolutely not justified at $110, and it should be more like $60-70 if not lower. Or is this "we have technical reasons to use this. We're still going to do 18A and it's still going to beat 2nm", which would mean their stock price is easily justified at $200 and it's high time to buy.

That's my point. It wasn't a clear win. It did some things no other model did. And it sucked at coding and terminal (as in nowhere near SOTA). It wasn't a clear win.

... and it was not SOTA at the time of release. Gemini 3.1 Pro was previewed on 19 February 2026. It just barely beat GPT-5.3 and was roughly equal with Sonnet 4.6. Better according to some, worse according to others. And that lasted about 10 days (until GPT-5.4 came out which was also not a huge jump). And these are large averages. On coding or terminal Gemini 3.1 Pro was not close to Opus 4.6. Also it only matches the GLM 5 open model, more or less.

Last time Google had a "everybody agrees" SOTA model was Gemini 3 in November 2025, it beat GPT-5.1, it really was better and held it for a month.

Currently Google's best model doesn't match open models, in any category (performance, price or speed), in fact the last 4 open weight champions all beat Google's best model.

Here's a plot of the evolution: https://www.reddit.com/r/LocalLLaMA/comments/1v20g29/kimik3_...

Google has not had a single SOTA model since about a year now.

Nobody spends like republicans.

Trump increased the national debt by $7.8 trillion in his first term (rising from ~$19.9 trillion at inauguration to ~$27.75 trillion when he left office), 39% increase (more than WW2, the Manhattan project and the Moon landing combined)

Each and every US citizen will be required to pay $22,400 for the (ahem) accomplishments of Trump's first term.

So far in his second term, Trump has added $3.2 trillion so far, if you calculate by the end of his second term it will be about $8.8 trillion.

Each and every US citizen will be required to pay $25,000 for his accomplishments now.

For comparison: you owe $9,200 for Barack Obama's last term, and $20,300 for Joe Biden's term. Oh and Clinton was by far the cheapest president since WW2, increasing the debt by a measly $600 during his 2nd term.

... but if you want an accurate graph, why not ask the LLM model to put the data points into a graphing library?

Yeah you'd think they'd focus on the noise pollution ... that only exists because these data centers are saving a little bit of money on sound isolation. Now THAT is a necessary fight, and it really, really should be fixed because it's a real problem.

Appears to be closed-weights entirely. No word at all on any weights release.

It went from 108 to 95. It didn't even recover for 2 months and 2 weeks (the next - spectacularly good - earnings were just barely enough to make the stock achieve 108 again ... for a second (well, 2 days, 1 if you don't count after hours). It really took close to 4 months. And nobody really knows if it wouldn't have been 10% higher forever without that particular senior management fuckup in AI.

And it would be FAR from their last AI fuckup. In fact, the Kimi K3 release, which extremely likely cancelled the Gemini 4 pro rollout (probably because Kimi K3 outperforms not just Gemini, but Gemini 4 pro as well. So in case you're looking to now ask "which Gemini did you mean there?", the answer is ALL OF THEM, very likely including unreleased Gemini models)

It outperforms the fucking ASR model Google uses, which is presumably also a Gemini derivative.

This, by the way, cost Google stock ANOTHER 10%, last week.

Why would you think that? You think pension funds and governments didn't invest this time? Because that's the problem, governments either investing everyone's money directly (to fund their own loans and expenditures, or should I say fund their own expenses using pension money without admitting that's what they're doing to pensioners) or indirectly force investments (you'll find pension funds worldwide are legally only allowed to invest in loans to the government and specific things the government allows (like politicians' charities). "For safety", of course, no other motivations there)

We can check if governments went back to their own tricks or not after 2008. What were the interest rates after 2008? Quick check ... yep, they went back to the old tricks.

When "shit hits the fan". Either governments really cut spending, leave pensioners without income and see their new loan interest spike to 10%+ (if they're lucky) ...

OR they do a massive cash injection saving the banks, and increase spending.

Which would you chose? It doesn't even matter. If you were the government and chose the first option, you would quickly find suddenly everyone across the political spectrum uniting to depose you. A decent chunk of people would literally not have any other choice but to do that.

(of course, as per usual governments aren't behaving "correctly" according to economic theory. If you take the old "government as spender of last resort" governments should have radically cut spending in the last 10 years, because there was no need for extra spending. If you take the new "government invests pensions", then of course government spending needs to be investment. In other words, governments only allowed to spend like a normal investor, expecting and demanding a return. Ie. no social spending increases. In practice governments increased spending at any cost, and so we're in trouble again)

Oh and should I mention that for any other entity investing pension money in loans to yourself is a very unique thing in law. It is THE ONLY financial crime where corporate structures do not protect management. You're the worldwide president of ExxonMobil? Doesn't matter. You do with employee pension money what government does with it? (ie. loan it to yourself, in this case ExxonMobil). You go to jail. Directly and long-term. In America. In Spain. In Japan. In fucking Vatican city.

Legal in the US too, obviously, just as long as you're the richest person in the courtroom. ChatGPT knows the full text of Harry Potter, word for word. Hence, ChatGPT is a reproduction of that book and many others (it even knows the chapters of my book, and only got 2 words wrong in the introduction if you can still get it to repro it)

This was illegal when they did it, that didn't matter.

Then it was made legal specifically for these companies.

Unless you're a sucker ("consumer") IP theft is perfectly legal in the US.

It's even worse. Steamboat willie, plus all the stolen Disney characters (Peter Pan, Snow White, Sleeping Beauty, Cinderella, Rapunzel, Elsa and Anna, it's essentially all of them, including some of the music even) are all in the public domain[1]. Go ahead, ask ChatGPT to make a picture of them. Publish your own version, because obviously making a version of Sleeping Beauty/Cinderella/Rapunzel based on the same source material will be pretty damn close to the Disney versions, and see if you get away with it in court. You know, with the law obviously on your side but the money not.

[1] https://en.wikipedia.org/wiki/List_of_Disney_animated_films_...

Amateurs. Everybody knows you should never wipe databases. You should install a cronjob that makes a random, unrecoverable change on a regular (but random) basis.

... for the management of Google that is so eager to present AI results that Prabhakar Raghavan proudly reported to a press crowd telling his daughter about James Webb photographing a planet outside the solar system. This is generally accepted as having resulted in a 10% drop in the Google stock price over 2 days.

(the key being this never happened, and won't ever happen: the James Webb telescope doesn't have even 1/1000th of the resolution necessary to resolve an extraterrestrial planet)

https://www.reuters.com/technology/google-ai-chatbot-bard-of...

https://www.wired.com/story/google-openai-gemini-chatgpt-art...

https://www.popsci.com/technology/google-ai-in-paris/

I wish I could say this only happened once.

For example: https://www.rusi.org/explore-our-research/publications/speci...

So after the 2008 spying scandal (cia got caught using German intelligence bases to steal airbus designs), Airbus ran it's critical applications ... on AWS?

Wow.

Russia's goals have far evolved over the course of the war, and started out comically ridiculous, prompting a great many people to ask questions about the information "bubble" around Putin. Over the past year, I believe most people would agree Russia has lost a LOT of ground when it comes to their war aims, even if not that much territory.

2014: seize Crimea (success, although it no longer serves any purpose) and Sevastopol (total failure, Sevastopol can no longer serve the Russian black sea fleet, and when it comes to the other goal: keep NATO out of the black sea, it's an even bigger failure. Russia has lost control of the skies above the black sea completely and are losing control over the sea of Azov, by withdrawing air defenses. Now go look on the map where the sea of Azov is to see the magnitude of this failure), prevent Ukraine from approaching NATO and EU (complete failure is putting it very mildly), keep Ukraine unstable (succeeded, until Euromaidan revolution, ie. actually this is a failure)

2022: maintain control over Crimea (looks to be failing, as in Ukraine still controls Crimea's water supply), build a land bridge to Crimea (successfull, except that they can't use their land bridge for anything), seize control over Ukraine's government ("denazify" Ukraine), demilitarize Ukraine (as in Ukraine accepts no to have an army. Obviously total and complete failure), and prevent Ukraine and others from coming closer to NATO (obviously total failure, and not just in Ukraine)

In June 2026, ie. about 4 weeks ago, Russia once again confirmed these goals.

Oh and they were going to do this in 3 days. Then in 3 weeks. Then in 2 months. Then they stopped making predictions. Today is day 1608 of their 3-day attempt.

Please keep in mind that "uncut footage of ..." is the same as those instagram videos where people throw a basketball backwards 20 meters while drinking beer, straight into the goal.

It doesn't prove this is average performance. In fact, if it was, that's not what they'd be showing.

Actually where I'm from the state has indeed decided budgets have priority over rights. In other words it works exactly as you say, as you suggest. So first BIG caveat in right to medical help is that only X people get treated for particular things, as budgeted. Even if the equipment and doctors are available (but of course those are getting "right-sized"). Even goes for access to GPs (there's a massive shortage of GPs, but the government is lowering the number of doctors)

You are number X+1 that needs treatment? Sucks. There's a waitlist. Which in a few cases is pretty bad. You have had a stroke and the doctor wants an MRI, to check for bleeding. The MRI machines are being synergized (meaning a bunch of smaller hospitals only have access by moving patients to a bigger hospital). They can't book a surgeon without confirmation of bleeding (well not without entering the waitlist at the end, without priority, and with a stroke in progress you might as well just book an undertaker). In the best case now you're being moved (you pay for this), potentially with a ripped artery. Best of luck.

There's also more temporary accidents, such as refusing to hire foreign doctors, even where that means parts of the country have to do with insufficient capacity in certain departments or even (in the mountains for now) zero capacity. In theory anything with a waitlist over 2 months is considered insufficient capacity, but if you take that definition, close to 40% of all departments have insufficient capacity. And it's getting worse (Ukraine war, tarriffs, EU lending, Iran war, ... all resulted in extra savings and less capacity)

The same is happening in schools. The law clearly states you have the right to kindergarten < 2km from your home, primary school <10km and secondary <20km). For "special needs" education this has become a total joke, even in parts of the capital this is no longer available. But even in normal schooling none of those are still reality. Or other issues, like the fact that the government guarantees temperatures in schools above 18 degrees, and below 32 degrees (heat index, not pure temperature. Essentially it's 32 degrees at 100% humidity, goes up with 1 degree for every 10% humidity drops). Kids have that right in education. Aside from the free school for EU expats' children in Brussels there isn't a single primary school with air-conditioning. Not even the one the royal family uses for their children. The temperature hit 40 degrees, apparently over 45 degrees in some classrooms. They did discipline some teachers for teaching in the grass outside of the classroom after parent complaints, of course.

That's also a way to look at rights, I guess. Namely that they are "weak guidelines" and you don't have any way to enforce them against the government and everything just completely erodes over time.

I guess it's true what they say: you can pay market price or you can have shortages. Pick your poison.

And yet, it seems awfully consistent with letting China export poisoned toys for children. In fact the EU made itself the only arbiter of what can be imported, then proceeding to allow that. Not by changing the law, but by simply not checking.

And of course, we keep having people claiming that they were, despite a decade of complaints about it, including in courts, somehow not aware that they were doing this. Why?

To say nothing of the other problems with Chinese goods: repression, use of slavery, use of child labor, ...

Now they'll do something about China (or so they claim) ... right after they created the same situation for South America, who also have a history of moral failings when it comes to products: repression, use of slavery, use of child labor, ...

This is exactly what I'd expect a business conglomerate, willing to make any compromise, allow any practice, as long as they can deflect blame and make a lot of money on it ... exactly

Well of course. If you want to know what a "right" is, legally, you need to reverse it.

For diet it is quite reasonable: a duty for certain products to be available for purchase (it does make clear that this right does not exist. A few vendors go out of business and your right to eat vegan goes with them)

For owning a home, this would be a duty for everyone else to build you a home, satisfying your requirements. One thing everyone agrees on is that other people do NOT have the right to force you to work. But that is the only way governments would be able to guarantee a right like this.

And, of course, one of the rights people demanded is the right to not participate in any right or duty that goes against your beliefs ... it is, of course, totally impossible to have this right and either of the other ones in society, or your rights would become very brittle indeed.

... and now another one when a (not so cheap tbh) Chinese model matched OpenAI GPT Sol and Anthropic Fable/GPT 4.8 (some scores below, some scores above) ...

Oh and exceeded Google's best internal model capability enough to have them cancel the next Gemini release. Probably. But I'm 95% certain that's exactly what happened. This is the first time I really fear for the future of Google, because OpenAI matched or exceeded the experience of Google search ... and now a Chinese company matched OpenAI ... followed by Google failing to catch up. That really made me take pause.

It indicates the the "secret sauce" of OpenAI and Anthropic is ... nothing, really. Or none of it matters, except access to the hardware needed for 1.5 trillion parameter models, which Chinese firms now have as well. It means there is no "AI takeoff" where other firms can't match OpenAI/Anthropic models. And Google, inventors of transformers just missed the takeoff. There's more secret sauce, but clearly a great engineering firm can find it in less than 3 months.

The fact that this Chinese model has similar cost to OpenAI and ballpark cost compared to Anthropic while we can be quite sure they're minimizing the cost (that's what China is doing everywhere else) also raises questions about the true cost of serving comparable models and with that about the future profitability of OpenAI and Anthropic.

As Google said in 2023, none of them have a moat, open weight models will win.

Google, who probably canceled the release of Gemini 3.5 pro to avoid having their best new model perform worse than BOTH GLM 5.2 AND Kimi K3?

I mean how is this anything but an incredible defeat of Google's supposed inventiveness?

The problem is that we're going to have another deepseek moment when someone uses GLM or Kimi K3 to do this.

Alphafold's weights are public (1, 2, and multimer, all except Alphafold 3) and there are several techniques to use it to do drug design, from Alphafold 1's "32 blank spaces" technique to the ML technique that should receive the Nobel prize for worst named algorithm in history "hallucination".

The popular method was this. You see because Alphafold goes from DNA code -> 3d structure, and you know parts of the 3d layout of a successful medicine (if it interacts, it will get really close. So you just put very low values in the alphafold output and "lock" them), then you reverse the normal learning process. You lock the outcome in place and you use backpropagation to learn the inputs instead of the weights. The inputs ... are the DNA code to produce the medicine. Hence "hallucination". You pretend to already see the medicine's effect and work back from that rather than trying to determine the effect of a given medicine. So generating 100 high-quality candidates for a medicine becomes letting a python script run for a day or two on a single 3090 (it was a while ago) rather than 5 years of research.

Oh and it comes with nice confidence matrices, so you can rank the 100 candidates. You can postprocess also using Alphafold to check many things that disqualify medicines (toxicity is essentially interaction with non-target molecules, which you can simply test for).

You can even, I mean I didn't get it working, but it is possible to "detoxify" a known-toxic protein using hallucination. Figure out which small changes make it non-toxic, and combine that with also making sure it maintains its therapeutic effects. That works. It's absolutely incredible and open source. But you need someone who is comfortable with high-level AI code to make it work.

And this works, with serious and very difficult code changes, with more modern models. Oh and pharma companies are unwilling to hire people capable of doing this (at least in Europe) at wages they'd need to pay to get them. As per usual.

Hitler did create an "economic miracle", which is why he actually went up in popularity, even with the parties he would later ...: https://www.youtube.com/watch?v=yRlbUeji01U

It was all based on a purposefully designed system of lies. Which can work, until someone proves it's a lie to external parties.

You realize the EU let Chinese sellers sell lead-containing painted toys to EU babies for 12 years despite endless warnings, right? They will not act unless forced or it's easy enough.

The EU came from the "European Union of coal and steel". It's a business first, not a government. And yes, they've really deceived a lot of people about this.

That's why we have the DMA ... except for Apple ... except for Google, as if that doesn't negate the entire law.

That's why we have the GPDR, except for (just for the Netherlands) any company the government wants https://www.avgregisterrijksoverheid.nl/ (specifically this negates the purpose of the GPDR. The first purpose of it was to protect your medical data from insurers, taxes, police and courts, so the insurer cannot decide you're committing fraud based on your medical data, or raise prices for you, or ... for example) well "ministerie-van-sociale-zaken-en-werkgelegenheid" has a specific exemption so they can regulate whether unemployment money can be used for medical treatment, and which ones ... And that's just one example.