U.S. government will decide who gets to use GPT-5.6

There might be a community effort at some point. This happened in chess where the community recreated and then improved on Alpha Zero. You could run small training chunks on your machine. Some people donated thousands of hours of server time.

I believe until the hardware designs catch up to be more commodized ala cryto mining evolution from GPUs to ASICS for specfic algos. Designs (like Google TPUs equivalent) would also need to evolve to be more memory dense to be able to handle them. Untill then it seems will be system time shares for the larger models , probably with a bring your own model and pay as you go.

Has anyone tried to run a data center as a Co-Op?

Was looking for this. It’s a great short story (it starts with real events and then predicts the future) leading to European fealty to US . Seems increasingly plausible.

Europe has been burning furniture to keep the house warm for so long that an entire generation now thinks that chopping wood is for suckers.

Has the US government limited access to any non-LLM ML models?

> Any country that developed sufficiently advanced models will pursue the same path.

Looking at most of the available evidence, Mythos is an incremental upgrade over other models and nowhere near the implied advancement that this seems to point to. I guess you could be right in that a sufficient advancement would cause this type of withholding of it, but I think it's kind of silly to think that the US has reached that level.

> AI is and always has been a winner-take-all strategy

People claim that AGI is. AI is turning out to be a fairly competitive but “normal” product. Companies carving out niches on cost, quality, and speed.

If it was a winner take all OpenAI’s head start would have been decisive. For years ChatGPT was far ahead of everyone, then Anthropic released Opus 3, then OpenAI released 4o, then in mid 2025 it seemed like everyone had strong reasoning models including Google with Gemini 2.5, and now Claude is probably the best coding model. So taking the top spot is not a guarantee you can hold it.

Also the top model becomes a prime target for distillation, making it easier for competitors to keep up.

I think you view this situation from the US point of view and assume that China has the same guiding principles and values in their foreign policy, for which it doesn't. They might do what you said, of course. But they very well might also treat LLMs as another goodwill investment like the Belt and Road Initiative (https://en.wikipedia.org/wiki/Belt_and_Road_Initiative) and export the capability to partner countries, for example, in Africa, to strengthen relationships.

This isn't how things typically work. For instance the US is increasingly adversarial towards China yet China continues to largely power the US economy both through manufacturing and market access, which makes up an increasingly large share of all revenue for many US companies. Why? Because that position as a dependency is not only directly economically beneficial to themselves, but also provides leverage which can be utilized in extreme circumstances.

This isn't really going to achieve much besides incentivizing the growth of non-US models and minimizing market access for US-based models. I was expecting the US government to try to ban foreign models, which is also a self-own but orders of magnitude less than this. All this will do is greatly diminish the influence of the US in the future, and minimize the benefits they might reap from a global LLM explosion. It'd be like if 30 years ago, China decided that their manufacturing could only be used by white-listed individuals. Their economy and influence wouldn't be even a fraction of what it is today.

> AI is and always has been a winner-take-all strategy

AGI sure. But I don't think we're going to get there in our lifetimes, if ever. There are too many structural and physical limitations. One everyone seems to be catching onto now is that they're running out of human data and are incestuously feeing AI output to itself as input.

Current state AI is barely an improvement over where it was fifty years ago. We just have stronger hardware and more content to train on. We need a new paradigm. One that hasn't come in half a century.

Do you have any sources for the claim that LLMs meaningfully help in the production of LLMs?

> Restricting their usage to a few companies takes away that advantage away from other companies, thereby, limiting new vendors from competing with OpenAI.

wait, are you saying a competitor needs access to an OpenAI model in order to build a competing model?

> So you're a new vendor with a GPT-5.6 or Mythos class model. How do you suppose the regulations are going to work? First you need to get on the list of companies that are allowed to release models, and then you need to have a whole system for limiting access. Both are going to be hugely expensive, on top of training new models already being insanely expensive to begin with.

So your thesis is that someone has the resources to create a model at the level of Mythos or GPT 5.6 but bot have the resources to jump through the concomitant legal requirements?

Surely that’s unlikely?

It's worth keeping in mind the broader context here politically. The president and his cabinet are generally pursuing policies of a strong unitary executive branch that centralizes "technocratic" functions under the control of politicians and political appointees. The majority in Congress doesn't appear to mind and seems to be actively sitting back from legislating on topics like this. The overall effect is that of creating a system in which the rules are deliberately uncertain and the only reliable way to get approval is by aligning yourself or your organization politically. It's a powerful technique for ensuring political compliance in the corporate world.

it's not regulatory capture, because they are not regulating what customers can use - it's limiting what some companies can serve. it's way less impactful to the market as a whole.

It is as if people can't imagine the concept of sanctions despite their liberal use over the years.

Right. That's what sanctions can accomplish...

Countries who've made the mistake of allying with the US might face sanctions or some sort of threats. People will just use Chinese AI then. This is the US biting itself in the ass.

Let me frame it like this;

If GLM-6 is made publicly available[1], will that change your predictions of how they will behave in the future, or your understanding of their motivations in the present?

[1] I am certainly no expert of Chinese AI policy, but I fear you are attributing US values and goals to the Chinese govt.

From where I stand, they appear to be very different though. China is seeking to increase influence, whereas the US (seems to be) seeking to become more insular. China is actively persuing the world-leader role, while the US dismantles soft-power tools like USaid, and alienates countries pursuing obviously flawed military excursions.

Why would the Chinese gov...spoil their rival's tulip bubble by dumping even fancier free tulips on the market?

I think this is plain old regulation, not regulatory capture. For it to be the latter we'd need evidence than it was the incumbents who designed this oversight which they then asked the white house to rubber stamp.

How does this benefit the regulated? I'd say it dooms the regulated:

* with the government requiring the regulated to obtain approval to add each customer, they're losing a massive number of their customers

* and even if a customer gets approved, every such customer now sees that access to the regulated can (and has been) shutoff with no notice if the gov doesn’t like the provider or customer - it's now a massive supply chain risk for any customer to use a regulated provider

* the regulated losing a massive part of their customer base for both of the above reasons means significant impairment of their revenue, as well as their valuation, and staying ahead of their competitors and open models requires massive ongoing investment

Open models are mere months behind.

That's just vanilla regulation. Yes it tends to create market inefficiencies, by definition. Capture entails industry corrupting and directing regulatory bodies.

And the regulated may even publicly complain about the regulations, to increase the illusion that it isn't regulatory capture.

Exactly! The thing that squeezes out new entrants isn't only the compliance cost, it's that your whole roadmap ends up resting on access you don't control.

We already saw the terms moved under us once with Fable, the retention policy changed and some requests started routing to a weaker model, none of us small operators had any say in that. Now access itself is a government decisions.

For anyone building on top of these APIs that's the real barrier, not the rule-following overhead but the fact that the ground can shift mid-flight and you can't negotiate with whoever's moving it.

Which is exactly why open weights start looking less like ideology and more like risk management.

Anthropic and OpenAI: Dear Uncle Sam, it seems laptops became too powerful, can you please do something

Uncle Sam: For national security reasons starting from now on every purchase of GPU model with higher than X Petaflops will need written permission by the US president

Anthropic and OpenAI: Look poor citizens, we are willing to share our capacity with you in limited form, by using our LLM you can avoid spending 35 years in the waiting list to buy a GPU, by the way, to simplify pricing, here is our new pricing with 5700% increase. Enjoy

Extrapolating the progress in both hardware and model efficiency, that will take decades

its not about whether you can shock anyone, if anyone is driving cars outside, you can't say you have SOTA-horse to go from A to B.

When models are good, expectations are adjusted accordingly to deliver things on par with the whole industry, you can't just say, I have built my own Intel Pentium II, now I will try to use it to compile Electron App and run 3DS Max there

Please go ahead. If you are right, you can raise billions of dollars for your startup.

I think I understand now. Don't do it for me. In fact, I'd rather you found something else to do.

I take it you've never known anyone who has tried to run even a small consumer AI company. Do you think that they all independently come up with the same censorship rules, despite no law mandating them? Why do you think that is? That they are just that well-thought out that everyone happens to agree with that same intersecting boundary?

We’ve seen more examples recently. TikTok, wireless routers, polestar cars…