I suspect if OpenAI or Grok was operating at the same level they wouldn’t find themselves on the sharp end of the government stick
I guess that is how it always starts.
Governments don't give up power, and once there's precedence for use of that power they'll continue to do it and begin eyeing the next power they can claim.
IF LLMs are THAT dangerous and powerful (and that's a huge if that I do not currently subscribe to), then no, no one should have access at all, there is no group of people in positions of power (government or corporate power) that I would consider "restrained"
From that, we can very reasonably conclude that the US government has a specific vendetta against Anthropic in particular, and that this vendetta has nothing to do with the technical merits of their product.
To my knowledge, they have yet to drop that classification, despite heavy court opposition.
Additionally: technical benchmarks suggest that the most recent ChatGPT models are within maybe 10% of Fable 5's capabilities, so this being a pure "capabilities" concern seems unlikely.
Uncertainty: It's possible that we have just suddenly reached the end of public AI releases, though - if ChatGPT 5.6 also gets blocked, that would be very good evidence of a general, non-weaponized policy. Given the recent Executive Order requiring pre-release audits of frontier models, this is somewhat more likely than it was a couple weeks ago.
I still think things add up to "weaponization is the most likely theory" and that one is being disingenuous to dismiss it as a reasonable possibility. But it's certainly NOT the only reasonable possibility.
ChatGPT 2 was once too powerful to release.
AI has been moving faster than culture and thinking around it. Once we've adapted to what these models can do we'll relax a little, and then a new stepwise improvement will start it all over again. It always goes this way.
It had just one the cold war and China wasn't even a shadow of what it is now.
So, the last time an AI-related export control was imposed - NVidia chips deemed to be "too powerful" - how do you think that will work out? If the US is holding all the cards, why is China now refusing their chips?
Don't you think these types of restrictions will weaken rather than empower the US?
I do. But I think the US is currently in a position of strength that they are continually undermining.
The broader discussion about AI and model capabilities died a couple of years ago precisely because it's so underwhelming now. People did adapt. Startups stopped hiring just to get to MVP. Coding sweatshops had huge layoffs and stopped overhiring. The corporate world got better tools for collaborations and meetings. Accessibility tools are still bad, but improving. I would argue that the a11y topic is still very ripe to be the next big thing as it continues to converge with better UI/UX instead of being an afterthought.
The layperson and tech professional alike otherwise agreed that this is a vehicle for blame game, grift, disinformation, etc. This is where all the pushback is and the topic at hand. People aren't dumb. The only people worried about "AI" are the ones who bet too big on it.
95% odds this gets reversed by Monday morning is why
AWS and friends are very good at providing excellent enterprise grade security, but it’s literal child’s play for nation state threat actors to exfil these models.
TEMPEST / EMSEC alone is a wide open door for unclassified datacenters when the Mossad’s out to get you.
I'm sure if proprietary models continue to be a big thing, the methodology of their storage and loading on hardware will be obfuscated quite a bit.
Your hypothetical implies that there is a better alternative, but when those models are "restricted", in practice that means that the only people who have access to them are precisely those who can and will use them for the worst kind of shit. So yes, releasing them to the public is a better deal, ethically speaking, at least then the playing field will be slightly more equal.
Businesses will gladly pay it.
Individuals will be locked out and unable to compete.
Oracle will be able to "prompt a new smartphone" for $500M or whatever and enter the market to compete with Apple and Google. You and I can't afford that and won't be able to compete.
Hyperscalers will hyper scale even faster.
They'll port Linux to Rust, remove the GPL, and have all kinds of new entirely proprietary OSes. They'll be attested, signed, and gradually we'll lose open hardware. Thin clients with binary blobs, highly encrypted, no control, only leased to us.
Within a generation nobody will be able to program or own devices that can program.
That's the scary scenario.
Then they won't survive the termination boundary.
Too bad. Should have had more cash.
You don’t _have_ to buy into the technocracy, there ’s a whole outside going on.
US have fair tax breaks to support the national champions, China have unfair State-backed monopolies.
US have necessary intelligence gathering, China have state-surveillance.
The US government believes that Fable/Mythos is a weapon that needs to be export-controlled, and limited to only US customers. Presumably OpenAI/xAI/Google would face the same constraints, for the same reasons.
OS/foreign models are unaffected - OS because they cannot control who runs them, and foreign because they are not controlled by the US government. We could assume that China will implement the same policy controls, but they see the world differently so might not.
So US AI companies are then limited to the US market, effectively, after about six months (the lag between the current frontier models and the OS models). They have much less incentive to push the envelope to create better models, because the US govt might also ban those completely.
The investor froth around the race to AGI dies, so valuations shrink (the current IPOs may be affected), and presumably the bubble bursts. None of the AI companies can afford to continue building data centres, so that all dies immediately. US GDP drops by ~5% because of that alone.
In a year's time, the US is in a major recession because it gambled so hard on AI. Europe less so, only because it was such a distant follower in that race. China is more-or-less unaffected. The best models are now OS/foreign, and AI is moving forward more slowly, but still moving forward.
Any other scenarios?
Get to Tuesday, restriction is lifted. Get to Friday, restriction is back on. Confusion reigns.
Interesting scenario though: do the other labs attempt to “dodge” the import restrictions by claiming their models are “dumber and not a threat” thereby maintaining larger market access.
If so, doesn’t this basically force a stall in US-based development? EU will keep doing its thing at its pace. Chinese models will get a boatload more popular, but will probably slow down as they can drop to whatever pace they wish.
cynical follow-up: if they’ve plateaued, is this a clever way to avoid the negative consequences of the market implosion that a substantially substandard model release would cause, thereby giving them an “out”?
Of course, the world is not filled with rational actors, and the probability of the current administration allowing the market to tank like that seems next to null, so Occam's razor (or whatever) would point to another TACO inevitably incoming
I'd certainly bet on your scenario if it was reasonable to assume the US and China could get over the 'race to the top or die at the bottom' dynamic
so far ai2027 seems to be playing out to an almost uncanny accuracy, realpolitik obliterates the façade yet again
The thing is, that blatant market manipulation is playing with fire here, as so much of the US economy is invested in the AI bubble.
If the major nations that host companies that create those OS models implement export control on top models, there won't be any new OS models with top capabilities.
Same as the new bridge between Windsor and Detroit can't open until some palms are greased.
Chaos is a ladder, gotta keep climbing
https://en.wikipedia.org/wiki/Anthropic%E2%80%93United_State...
The public may not see more improvements but I'm sure they government will be forcing AI companies to continue improving them for their own use. Those schools aren't going to bomb themselves.
“Oh no it’s not our fault!!1! ~~My dad works at Nintendo~~ the govt told us we weren’t allowed to release it”
The base intelligence does not feel much greater to me.
Fable was designed for _really_ hard software engineering problems. Possibly large, but especially hard. For those tasks, you feel the difference immediately.
And yes, the parent poster is accurate, Fable is just as prone to moronic mistakes as Opus was. Stop being so AI-pilled.
Codex is still a better model, and yes, for the hardest engineering problems. I use Claude for UI/GUIs and Codex for all my backend, because I have 20 years of experience of actual hard engineering, and I can see that Codex writes, cleaner code, and is far more steerable.
Bad engineers think Claude is better because it writes more lines of code and is more "proactive", but lines of code doesn't make a better system.
This is a forum filled with experts. Putting marketing aside, in a forum like this, it is most useful to assess models according to the toughest problems in the domain they were specifically refined on. For DeepSeek, that's math. For Claude, that's programming. Gemini and ChatGPT are generalist. Yes, you can use every model for anything you like. But Fable is a bit special, it's very expensive, and very clearly designed for particular types of tasks.
> Fable is just as prone to moronic mistakes as Opus was.
"Just as" is up for debate, but yes, all models are capable of moronic mistakes. That's not helpful information though.
> Codex is still a better model
You're comparing agentic workflows, which relies on a lot more than just the underlying model. It sounds like you're using it like a precision instrument, which is great! It's very different compared to my use cases though, and the ones that Fable seems to excel at. I'm using it for scientific computing, and you really, really want it to one shot a solution. It's either the right algorithm for the task, or the wrong one. So for the hardest problems, it needs to successfully implement a solution in effectively one shot. I use Codex too, but it's often too careless for the delicate tasks. If it gets it wrong, it is really hard to steer it back. You have to start from scratch.
> Bad engineers think Claude is better because it writes more lines of code and is more "proactive".
Think you missed the mark on this one. Not really an engineer, have as much experience as you do in my job. A solution to my problems comprises few lines of code. Fable actually gets it right, first time, every time (so far), but this is with a very long prompt and a bunch of attachments. No other model has done this for me. Not shilling for Anthropic, just impressed. This isn't particularly subjective for me; it is quantitatively measurable.
Don't assume everyone using AI is going to have the same experience you have, or the same types of use cases. And please don't assume that because others have different experiences that it makes them "bad".
Also, Claude has always been mediocre at creative tasks. For your line of work, I would have already recommended Codex hands down.
Half of HN commentators probably work on basic CRUD. Armchair experts, maybe.
Efficiency will become the next thing to focus on. It was already emerging, but accelerating the focus on efficiency will lead to a ton of excess capacity and even some investments in data centers to go belly up. And ultimately the AI bubble bursting will look a lot like the dot com, with its surplus fiber.
Oh, and this will put gas on the fire that fighting AI and big tech is the next political rally cry. Along with “eat the rich” as they are seen as taking both jobs and money.
Curious to see where it’s all headed and how Trump’s call will impact it.
But I hadn’t considered this fell into that category. Except maybe as a direction from Iran. You make a good point, it may trigger immediate reactions in the market. Not just 3-6 month ones.
I wonder what the counterbalance will be by Monday morning.
What an unbelievable claim. Especially since the vast majority of publicly available benchmarks disagree.
These threads about Anthropic always seem so astroturfed with some of the loudest and most uninformed people around.
It’s still pretty good though
I've used Claude Code and Codex since release and use them both in parallel. Codex is still better (yes even with Fable).
Claude Code is best for UI, nice looking guis, etc, and apparently also best at impressing mediocre programmers who are prone to AI psychosis.
All the best engineers I worked with in past jobs (faang folk, spacex, x.ai, and others) all use Codex, go figure.
You do realize none of this thread was about Claude Code vs Codex?