Comment by data-ottawa
1 day ago
The US is really shooting itself in the foot here.
The restrictions on LLM models like Fable has created a viable international LLM market where it was difficult to justify investment two weeks ago.
As a non-US citizen Opus 4.8 is the best American LLM I will ever have access to. That's no longer up for debate or question. Each month that I pay Anthropic is now a depreciating value -- I'm paying for models I'll never be able to access, while other models are able to catch up.
Adding US based identity verification through Persona is also incredibly off-putting. I think it's sufficient to kill my use of Claude altogether.
So the question I have to live with is what do I do instead.
I installed Mistral Vibe last week and I've been experimenting with offloading work to it. I won't pretend that Mistral-medium is close to state of the art. It isn't. It still writes incorrect tool calls.
From the last week about 50% of my LLM tasks actually reduced to "take this work and write about it" and Mistral excels there -- it definitely beats Opus at writing. Mistral nails it, and when it doesn't its so fast to iterate.
There's another say 30% of tasks that's writing queries against a data warehouse. I updated my semantic layer MCPs and Vibe uses them, but it struggles with ambiguity here. It's not a replacement, it's maybe where Opus was a year ago.
The rest of my work involves writing code. That's going to be harder to replace for now. My next step is exploring OpenRouter and other models. I can't decide if I was ever actually happy with Opus's work on this front though -- the understanding tradeoffs when you trust LLMs with decisions stack non-linearly and negatively. I did like Fable on these tasks, I won't lie, I will miss it, but not by any choice of my own.
Despite what’s being implied everywhere, this ID check page has been there since April. Wayback Machine if you don’t believe me: https://web.archive.org/web/20260415064244/https://support.c...
> As a non-US citizen Opus 4.8 is the best American LLM I will ever have access to. That's no longer up for debate or question.
This is a crazy conclusion for a situation that isn’t even two weeks old. LLMs are not the first tech product that have been restricted by export controls. These situations pass. Administrations change. Technology evolves. We’ve had export restrictions on different chips and even cryptography software in the past. It doesn’t last forever.
This is a strange plea for optimism.
Sincerely or not, judgement is yours, Dario has been begging for regulation. He has been talking about how Claude models are distilled by foreign adversaries. And now the regulation is here.
What makes you think this situation that the CEO of Anthropic is asking for it temporary? Do you not believe Dario was sincere?
How ironic it is that multibillion AI companies are complaining that their models are being distilled (read: used without permission) while the current top players trained their models on stolen data...
1 reply →
There are three issues here:
1. Identity verification as a way to validate real-personness and mitigate distillation by e.g. North Korea
2. Identity veriiication as a way to limit model usage to US residents / citizens
3. The level of model which will be subject to identity verification, today and as time goes on
It’s a mistake to conflate the three and form a rock solid opinion of exactly what will happen from here to the heat death of the universe. Everything about AI is moving quickly. I doubt Dario would claim to have a perfect roadmap not subject to change.
My personal guess is that just like export controls on CPUs, this will apply differently to different regions, and will change over time. Especially with US political instability and increasing anti-science policy, I cannot imagine Dario or anyone else would want to surrender the EU and other markets to become a US-only company.
But whether I’m right or wrong, one thing I’m not is certain. I can’t imagine how anyone could be in the current situation.
Do you sincerely believe Dario wants Fable to be restricted to US citizens only?
3 replies →
I’m a US citizen and absolutely will not be uploading additional information just to use a company’s models. This effectively kills my usage of anthropic for anything beyond their 4.8 models.
> I’m a US citizen and absolutely will not be uploading additional information just to use a company’s models.
I can't speak to your specific use of Anthropic's models, but I find it interesting that people will identify themselves (to set up and pay for an account) and provide all sorts of personal (and often sensitive if not confidential) information to these models on a daily basis, but balk at a 5-minute identity verification.
26 replies →
This decision has, effectively, turned LMMs into a supply chain risk.
Before this incident I’d gladly use any anthropic LLM in production features. Right now, this has become a risky decision that can tank my business overnight.
> Right now, this has become a risky decision that can tank my business overnight.
If your business-critical systems rely on SAAS that doesn't have a solid SLA and breach-of-contract provisions that more than cover the damages in the event, you've made "a risky decision that can tank [your] business overnight".
If the software your business depends on can't run indefinitely without getting permission to operate from someone else's systems, then you're perpetually at risk of someone else tanking your business because they decided that you can no longer use that software.
2 replies →
The long term goal of LLMs is to automate most white collar work, so wasn’t that a risk you faced anyway? Like Amazon basics, which took all the easy to replicate goods they saw on their marketplace.
Crypto export ban lasted only about 45 years. :)
You're way off the mark, and probably viewing this as an American.
If it happens once that means it can happen which means it can happen at any time.
When you feel you're a 2nd class person, or 'other', you're not eager to empower your oppressor, quite the opposite.
> When you feel you're a 2nd class person, or 'other', you're not eager to empower your oppressor, quite the opposite.
Completely agree.
You're acting like you can't just switch which llm you are using in around an hour.
I mean...use opus/fable when you can, if down the road your access gets cut then just switch to kimi or whatever.
Yeah, this sucks, but you're being really dramatic and acting like you can't switch llms with basically no lock in. Getting something like your email cut off would be a real thing to be concerned about, but this isn't that.
> These situations pass. Administrations change. Technology evolves. We’ve had export restrictions on different chips and even cryptography software in the past. It doesn’t last forever.
The big problem American policy makers (and business leaders) don't understand is that they tend to minimize or ridicule extremely serious events.
There's a pre-Greenland and post-Greenland annexation threats for European Nato allies, and it is non reversible. EU allies do not forget that the US (the only country to ever call article 5 or to gather NATO allies for operations) has both mistreated the alliance, and has been the only power to threaten militarily EU countries.
Same happens here. Business-level wise, you seem to be talking with a very American-centric point-of-view, like these events are minor and temporary issues and we're all here waiting to throw money at an abusive relationship.
But this is not how we operate in EU. None of us can afford to build their operations based on uncertainty of US export controls. The damage is here and many of us are replacing Claude/GPT subscriptions with shared opencode servers using GLM and DS4.
Might be slightly worse? Probably. But we can work on it, harness it, get experience, and even update back to American models at some point. But we're no longer going to be building assuming US models availability.
> We’ve had export restrictions on different chips and even cryptography software in the past.
But not on a SaaS whose continued availability you'd rely on.
In any case, your optimism is bordering on naivety. The world has seen how the US can easily disregard anything and act arbitrarily - sanctions, tariffs, shutting down access to SaaSes - and this will not be forgotten. As you say, administrations change. Even if next time around there are competent adults in the White House (which really isn't a given), do you really want to bet your business on that not changing 4 years later?
There's a reason why all the big cloud providers are constantly shouting about their "sovereign" solutions. The US has broken everyone's trust and there is no going back on that.
> This is a crazy conclusion for a situation that isn’t even two weeks old...
I think, this is all a culmination of rapidly eroding trust and soft power between US & its allies for the past 3y.
What allies?
You threatened to invade Canada and Greenland.
You surely don't think you're coming back from that?
15 replies →
Over-reliance on a single LLM is probably not a good idea, no matter who owns it.
> past 3y
since January 2025
1 reply →
US controls on cryptography software lasted _20 years_. If there's something I'm absolutely certain of, and I'm certain of very little in the fields of AI and of politics, it's that Fable will be utterly irrelevant in 20 years time.
> Despite what’s being implied everywhere, this ID check page has been there since April.
Well, irrespective over as to whether this is the case, the blog entry from claude came yesterday, aka June:
https://support.claude.com/en/articles/14328960-identity-ver...
So, why the two months delay here? If they felt all was already said, they would not have had a need to repeat what they wrote two months ago already that mandatory age sniffing is required for all claude users.
What makes you believe this is about export controls rather than harvesting data?
What makes you believe he belives that?
2 replies →
Mass surveillance and all the other that gets associated with mass surveillance
Export controls have typically been for physical goods. Don’t remember the last time it was used for an API
You should look up the words "crypto wars". There were absolutely very annoying attempts by the US government to limit encryption, forcing every software maker to maintain two editions of their software: one targeting the domestic audience with no restrictions, and one "international edition" which had to intentionally weakened encryption (as in ship with shorter key lengths).
Have we all forgotten PGP already? (Not an API, but certainly not a "physical good")
> We’ve had export restrictions on different chips and even cryptography software in the past. It doesn’t last forever.
Yeah, in the long everything will happen, from 1,2,3,4,5, and 6 being the winning Powerball numbers to heat death of the universe, but as the colloquial goes "ain't nobody got time for that!"
Once an institution or person has proven that they will take adverse action against you, it is foolish to bank on them again.
I also think this makes OpenAI and Anthropic even less viable. They're tens of billions in debt, losing money every month, have data center commitments in the hundreds of billions, and now they're reducing their market to the US? The only way this can work is with substantial government subsidies.
And a fraction of the US market at that. Requiring ID scans and all of this nonsense to use a chatbot just added a whole layer of friction that many are going to just nope on out of and use one of the endless other services. I suppose the next step will be for the government to try to ban those other services once they mature and start booming.
Definitely damaging for their IPOs
Tragic.
subsidies what like space x or tesla?
shhh, the investors might hear you. If they do then the bubble pops, all those data centre projects grind to a halt, 5% of US GDP disappears, and the USA goes into a deep recession.
Of course, as you say, this would require govt funds to fix. Congress gets asked for funding to save the US economy, and various members of the Trump dynasty pop up in unexpected non-exec directorships for companies receiving those funds.
easy IPO cashout
> As a non-US citizen Opus 4.8 is the best American LLM I will ever have access to.
> Each month that I pay Anthropic is now a depreciating value -- I'm paying for models I'll never be able to access, while other models are able to catch up.
Excellent point... That made me rethink my payments to Anthropic. As one of the foreign peasants who was banned from accessing Fable by the land of the free, it's become really hard to justify giving Anthropic any more money. I'm very tempted to switch to GLM 5.2.
> banned from accessing Fable by the land of the free
It should go without saying, but "the free" in "the land of the free" refers to Americans, not everyone else. Not sure if you were trying to make it sound like a great irony that "the land of the free" would exclude or ban people, but if so it doesn't quite hit the mark. A more cutting criticism would be that the land of the free isn't letting one of its own companies freely compete.
Fair point. The fact that even americans got cut off from Fable is definitely ironic though. I suppose that's going to be fixed now that they're implementing identity verification. And even that runs smack into the concept of freedom since freedom under surveillance isn't real.
3 replies →
It doesn't even refer to all Americans. Just the wealthy ones. And even that is increasingly reduced to just the wealthy ones who support Donald Trump.
1 reply →
What is stopping you from switching to GLM 5.2 now? Have you tried it out yet?
I'm playing with Deepseek a lot more via OpenRouter recently and the only major downside I can see is the usage billing over the monthly plan
Existing setup. I'm already used to Claude Code I guess. I actually spent time and tokens customizing and fixing it. Have a patcher utility that modifies the binary in order to disable telemetry and remove performance reducing language in the system prompts. Every update I spend some Claude tokens dissecting the newest executable and integrating it with my patcher.
The monthly subscription is also a major hurdle for me. The "high end frontier models for low prices" aspect is a major reason. I think I'm getting a lot of value from my subscription, given that the API prices are like a hundred times higher.
However, there's also a psychological factor here. These subscriptions are like the gachas of the software world. I got "addicted" to them. I developed workflows around achieving 100% weekly usage. Sometimes Anthropic randomly resets weekly usage and I scramble to get the most out of it. I'll point Claude at things and then just have it run hundreds of code review agents. I ran out of projects to do this on and started doing it to my favorite open source projects instead, looking for things to contribute.
I think with usage-based pricing I wouldn't use LLMs quite so freely. It'd probably cure my "addiction" too, but the problem is I'm not sure whether that's a good thing, since this "addiction" has been a somewhat positive force in my life. It's driven me to start new projects and also make major progress on existing ones. It brought me out of a slump. I'm a little afraid of moving away from Claude and not being as driven as I was before.
2 replies →
Meanwhile, as an irregular user of heavy AI, I like the usage-based billing of deepseek. Made a bunch of optimizations to my vibe-coded codebase, created some extra modules, using the chat... has costed me 33 cents so far in a couple weeks.
3 replies →
The downside are opportunity costs - not using other models to make better decision, I guess, right?
For the 100$ I'm paying for Claude, I'm pretty sure I can use Deepseek way more than I can use Opus via the plan.
usage billing over the monthly plan when deepseek is over x25 cheaper?
As an American founder I am sorry for this. I think the world needs to band together to take geopolitical risk out of AI.
> The restrictions on LLM models like Fable has created a viable international LLM market where it was difficult to justify investment two weeks ago.
Note the big cut in token prices from China.
[1] https://www.globaltimes.cn/page/202606/1363827.shtml
> Note the big cut in token prices from China.
After the recent 75%+ price cuts by DeepSeek & Xiaomi MiMo (and now, MiniMax), I pretty much packed my Claude bag up and moved over. I see no discernable difference (other than chattiness of its thinking modes) in capabilities for the kind of coding & debugging work I do.
and China is not threatening to invade the EU or Canada. They are the lesser of two evils at this point.
16 replies →
I just got finished setting up an environment with MiMo-Pro-2.5-UltraSpeed, Qwen-3.7-Max (basically used this to sub in for Opus), and of course DeepSeek.
Then GLM came out and that just means everything got even better.
You need to look up loss leader strategy and dumping in international trade because that is what china is doing. And your use of their models over api is giving them training data.
They will flip to being just as bad or worse if they beat America.
But America isn’t deserving trust at this point.
The only viable option now is local AI. Our industry needs to figure out how to decentralize training data, infrastructure, inference and analysis.
4 replies →
They clearly specify that they accept IDs from most countries. This likely means that they've reached a deal (or hope that they'll reach one) with the US government that lets them share Fable with foreigners, but only as long as they know exactly who it is being shared with.
This is off-putting to the HN demographic, but won't change anything in practice. 99+% of people will just do the ID verification and move on.
Persona already had scandals earlier this year:
- https://cybernews.com/privacy/persona-leak-exposes-global-su...
- https://hothardware.com/news/discord-drops-persona-after-use...
It’s off putting outside of HN
The Peter Thiel connection is especially toxic for a lot of people outside the US. Whether it's substantive or just optics doesn't make a huge difference.
1 reply →
> 99+% of people will just do the ID verification and move on
Why? Is Claude really so much better that the additional hassle and privacy invasion is worth it? What's stopping people from switching to one of the dozen or so other AI tools?
Heck, considering the volatility of the LLM industry, shouldn't everyone already be using OpenRouter & friends to avoid getting screwed over by the model-of-the-week - making a switch absolutely trivial?
Fable seemed pretty good, but the thing is, even having access to test the next model _and see if it's worth switching to it_ is worth the marginal hassle/risk of doing the ID flow. A lot of things now do KYC, so we're not really talking about a categorical shift in me sharing ID info with any company vs not. It's just one more app.
> What's stopping people from switching to one of the dozen or so other AI tools?
I’m lazy.
I will simply upload my drivers license to Claude, and continue paying $200/month.
Subscription prices. Anthropic subscription can't be used with third party harnesses. OpenRouter would require me to pay API prices.
1 reply →
Sure if you want to share your ID and information with Persona.
1. https://news.ycombinator.com/item?id=47140632
That remains to be seen. Assuming the worst from the start discourages people from taking any action and is an imprudent way to approach political topics.
Completely disagree that 99% of people will just do it.
A lot of people are already skeptical of the frontier labs - I moved over to Claude from OpenAI when they bent over for the US government. And I'll certainly move on again if Claude start asking for photo ID.
> 99+% of people will just do the ID verification and move on.
In a vacuum yes. But this space isn't vacuum, I'm going with path of least resistance.
It's pretty obvious this has nothing to do with Fable
> Mistral nails it, and when it doesn't its so fast to iterate
I'm starting to see more and more of this: speed matters more than model, skills matter more than model, cost matters more, harness matters more than model. It seems like until we have a step change in models (and Fable isn't it), there's a lot of room to optimise with what we already have.
This is a stark change to the "best model at all costs" mentality from a year ago.
The government mandate forcing them to restrict access of new models to American has really cut their legs under them.
This identity verification is a best effort to kinda stay afloat: they can now offer bleeding edge models to US nationals, but not to the other 95% of the world. Their influence is gonna tank quite seriously if the previous mandate is not reversed.
Realistically, their choices are to either implement this, or restrict access to new models entirely, which is a sure way to fall into complete irrelevance.
The worst part is that the entire ban is zero step thinking. The US is showing the world that even when it has the best tech it's too politically unstable to rely on as a serious partner.
Worse, doing this forces other nations to catch up to compete. Once they have, what happens next? An AI arms race that the US may not win? Someone else opens their Fable class model first and takes the multi-trillion dollar market that could have been run by a US company?
There is no n-step positive outcome for the USA. The only winning move was not to play.
> As a non-US citizen Opus 4.8 is the best American LLM I will ever have access to. That's no longer up for debate or question.
Not only is it up for debate, I find it extremely unlikely to be true.
I don't really disagree with the rest of your post but I very strongly doubt that Opus 4.8 will be the best American LLM you'll have access to this year.
Why, is there any evidence to support your claim?
There is a long history of US export restrictions on technologies. They have all been temporary.
It would be extraordinary if this was the beginning of the first ever permanent technology export restriction.
3 replies →
Polymarket gives 32% chance the ban on Fable is lifted this month.
Even if it's never lifted, probably GPT 5.6 will be better than Opus 4.8 and they won't hype it up as too dangerous to release before releasing it.
Open models will catch up to Opus 4.8 and state of the art will, most likely, continue to get better. It won't be long before "better than Opus 4.8" is not a big deal.
6 replies →
GLM 5.2 is better than opus 4.8 bro what are you talking about
Have you tried one of the Kimi K2 models or the latest GLM models by z.ai? The general consensus is that they're at least at par with Claude's class.
They are but from our evals for example GLM 5.2 (unquantized) performs as well as Opus but uses more tokens and takes more time.
I really wish this would change soon but they are not there yet.
Using even double the total tokens and taking, what, 2-3x the time?, still seems worth it if prices are 5x+ cheaper (which OpenRouter [1] claims is the case).
On NeuralWatt for my personal projects at home (not affiliated, just a happy customer), I get so much more mileage out of GLM than I get out of Claude at work, specifically because it's priced as a hammer I can pound any nail-shaped-object with, not a delicacy I need to carefully budget-analyze to try to figure out if it's worth burning my monthly spend limits on this task.
https://openrouter.ai/compare/z-ai/glm-5.2/anthropic/claude-...
I thought true token use was being hidden by anthropic and openai both
2 replies →
If K2 or GLM 5.2 are on par with Opus 4.8 I'll eat my hat. They're good, but they're not that good. Deepseek V4 Pro has been better than Sonnet for me, but the only model that comes close to or surpasses Opus 4.8 is GPT-5.5.
GLM 5.2 is far better than deepseek V4. Seriously feels like I’m talking to a Claude model. Also burns tokens like one, so there is that. Deepseek is unbeatable on price/quality.
Honestly just give it time. This stuff moves so fast next month the conversation will be different. For folks who don’t like the ID privacy issues, use Deepseek et al and it should be able to get the job done even if the experience takes a bit more wrangling.
The problem with the ID verification is that they can pair introspective conversations with ID. Either that bothers people or it doesn’t.
Main point: we can’t fret about current state models because the ID verification has future implications. Models will change and competition will catch up. Do what feels right in the long run not whether TODAYS model is better at Anthropic.
5 replies →
> As a non-US citizen Opus 4.8 is the best American LLM I will ever have access to.
Your overall argument makes some sense, but I would bet any money this simply is not true. Even if the US maintains some of the restrictions on export (which is in no way a given with how fickle this administration is, let alone the next administration), as LLMs advance, Fable will eventually be considered a lower tier model, and is likely to have restrictions lifted.
That makes it better, paying money to always having access to the second rate models from US AI providers. The problem remains.
Which is why I said the overall argument makes sense.
> paying money to always having access to the second rate models from US AI providers. The problem remains.
Your alternative:
Pay a Chinese or EU company for a third rate model.
The choice is yours.
3 replies →
Worse than that. You're funding models and an ecossystem that discriminates you.
> The US is really shooting itself in the foot here.
Not exactly. The close buddy with the owners of several other AI companies who has a known dispute with Anthropic and who has seized near-absolute power in the US is doing something that damages Anthropic. The fact that it also destroys long term prospects for the US overall is irrelevant because the individual can't think ahead and also doesn't care about the future of the US.
It's nearly the same thing in the end, but helpful to understand the cause.
There’s no way I’m giving a random foreign company my ID.
I’m really enjoying the phone-controlled sessions with Claude, and the models do good work; but if I’m asked to validate my ID, we’re done.
I already have Aivo and Pi somewhat similarly configured. Just give me a reason and I’m giving my money to someone else. In Europe, if you have my ID it’s not that hard to get me into trouble. And there’s no safeguard that Persona won’t be hacked, it’s not like they encrypt the ID with my own keys or something.
Hopefully Persona is a non-starter for many. At this point I'm done with Anthropic, I'll be a non-paying subscriber to any of the US based "Frontier" providers. I've been finding far more value in how the models are used vs leaning only on the brute force of a SotA model. Between the Fable / Mythos FUD and scare mongering that Anthropic continues to prance around with I can't take them even remotely seriously anymore. Just like with early iterations of ChatGPT these founders have acted like they're sitting on AGI. But just like with those earlier versions of ChatGPT we'll look at frontier models a year from now and laugh at how off base Dario has been, again as he's been very off base for the last couple of years. I'm still waiting for half of white collar jobs to be replaced next week...
> As a non-US citizen
...
> My next step is exploring OpenRouter and other models
May I suggest using Cortecs.ai then? OpenRouter is US-based as well and since you have been bitten by this already perhaps it's really time to change course? :)
I don't believe US government wanted to restrict the model use to US citizens only. The bad actors will and can find their ways around it.
They do however want strict measures in place to avoid abuse, and the export control was the only tool they had to stop Anthropic from releasing the model.
Though I also wonder if it's even possible to patch things without severely crippling the abilities of the model.
> The US is really shooting itself in the foot here.
> The restrictions on LLM models like Fable has created a viable international LLM market
The issue is that there is no "international LLM market." America is leading the AI race, and while Chinese open weight models are great, they aren't quite bleeding edge. I routinely use Qwen and GPT-OSS (locally) for things I don't want to share with Anthropic, but they are clearly inferior to SOTA cloud models.
How does it follow that there’s no “international LLM market” just because one party is ahead of another? There’s an international car market even though some cars are better/worse than others.
> America is leading the AI race, and while Chinese open weight models are great, they aren't quite bleeding edge.
this sentiment is far too commonplace in my opinion.
just because the BMW x5 exists doesn’t mean rav4 isn’t used by far far far more people.
the rav4 is close enough for most people to the x5 and way cheaper.
if deepseek is close enough and significantly cheaper, which direction do we think the market will go once the hype trash moves on and people realize how much of the hype trash is just botted algorithms?
Kings come and go. These types of decisions can actually kill a king. Not instantly, of course. But still.
This strikes me as bluster. You use Anthropic because they offer the best model, if China offered a better model you would use them. You’re trying to signal that you deserve Anthropic’s best model but the truth is as long as Anthropic continue to serve a better model than China, you will use it.
Definitionally, slightly better than China should always be fine for export.
> You’re trying to signal that you deserve Anthropic’s best model
You think some people aren't "deserving" of the best models?
Not an argument I made. A summary of the argument OP was making.
> I'm paying for models I'll never be able to access
That is very well put.
If they have a year sub, then yes I agree (even if it’s implicitly always part of the risk of buying so far in advance) but if they are month to month this position is absolutely nonsensical.
You seem to be saying that it's not a problem because you can just cancel your subscription if you don't like it. The fact remains that it's true. It's bad for goodwill in the same way Apple flipping $ -> £/€ is.
The entire internet will be real id verified soon. You won't even be able to send packets without a hardware digital signature linked to a real id. Might as well get used to it.
I bet that if you don't use companies that enforce these verifications, and they start to lose money, things will shift very fast.
No. Google and Apple are already rolling out the infrastructure. FCC is requiring IDs for phones. It will be opt-in for a while but the acceleration of AI risks will inevitably lead to an infrastructure level lock down of the entire internet. 99% of the population will accept it without hesitation because it is no different than what they do already. Everything you do online is already logged and tracked. The only thing moving to digital real id does is prevent criminal activity.
6 replies →
Please do yourself a favour and check out GLM-5.2, Qwen-3.7-Max, MiMo-2.5-Pro, and DeepSeek-V4-Pro.
And then for stuff that you already said you were able to use Mistral for... Qwen-3.6 (option to run locally), MiMo-2.5, DeepSeek-V4-Flash, or... many, many other models to choose from too.
right now the best contender is GLM 5.2, right?
either this or stick with Codex until the US government cripples it too
I’ve been using GLM-5.2 on openrouter with pi and, while I’ve only been playing with it for a couple days, seems stronger than Opus 4.8, nearly on the level of Fable for coding/architecture tasks.
You underestimate how large the US market is.
AI was always going to be geo-politically fragmentary. As soon as we started talking about it like a utility it was clear that every country was going to be strongly limited to the resources it could develop and the infrastructure it could build out. The US and China will still be selling hardware, managing infrastructure and licensing models and yes the domestic market is massive.
you understimate how large the international market is, for example, half of google's revenue is generated from international markets.
Only *half of Google's revenue is generated outside of the US.
You just proved my point.
I don't know what was the hype about Fable. It was crap. This more looks like PR stunt by the US government to prop up the failing product.
Now everyone talks about Fable and wants Fable.
Having used it for limited time when it was available, I don't miss it at all.
I wanted Fable in order to harden my C project against exploits and vulnerabilities. Anthropic downgraded me to Opus 4.8 every single time I tried to do it.
So I don't even know if I miss it. I suppose that's equivalent to never having used Fable at all.
Claude has been doing this to me even without Fable, trying to write a Xaw frontend for the moc music player.
> As a non-US citizen Opus 4.8 is the best American LLM I will ever have access to
No, we are all just waiting for Dario to get scheduled for an Oval Office press conference where he can present a gold trophy to Liberace Hitler and extoll his praises for all the amazing winning he is doing like no one has ever seen before.
> As a non-US citizen Opus 4.8 is the best American LLM I will ever have access to.
Codex-5.5 > Opus 4.8, so that's not true.
Where does 5.5 beat Opus 4.8?
Reviews for one. It's reviews are phenomenal.
1 reply →
This is just not true. Especially for frontend development. But I've compared them very recently one to one and Opus wins every time.
Some people have been saying that 5.5 is on par with Fable and it's just nonsense.
If you haven’t already, check out omnigent for building workflows across multiple harnesses: https://omnigent.ai/
(Disclaimer: I work for Databricks, but do not work on omnigent - though I have submitted some QOL PRs as a community member)
> As a non-US citizen Opus 4.8 is the best American LLM I will ever have access to. That's no longer up for debate or question.
That’s totally unclear. Things are changing fast. No statement from god or potus has come down about the future of LLMs and who can access what. And for what it’s worth, I’m not able to access fable and I’m a US citizen.
> That’s totally unclear.
Lol, welcome to US foreign policy. Or US trade policy. Totally unclear, and it's a feature.
Yes, lack of stability and clarity is entirely why people are steering away from the US. Yes things can change again next week, it's not a good thing.
>That’s totally unclear.
Being unclear is enough for people to steer clear of it...
I can't build workflows on something that can randomly be unavailable for over a week.
At this point the future availability of Anthropic models outside the US is very unclear.
Right, if we saw an open-source Mythos release today, I’d expect it to move the government’s idea of security goalposts.
That's the implied suggestion from the verification that they can prove to only share it with US citizens
My understanding was that until they can make sure it cannot be weaponized against the US, only US citizens will be allowed to use it.
(Read: the US lacks authority to ban use by citizens and doesn't want to risk their hand in court, particularly since lawyers who know anything have all left US government and what remains are complete incompetent jokes who can't even win slam-dunk cases due to repeated procedural errors. The nice thing about blocking non-citizens is they are easy to bounce out of court on standing)
1 reply →
Or its to stop industrialized distilation efforts?
I would suprised if admin doesn't want american companies and their employees to not be over competitive with outside companies.
But I do see them wanting a lever to prevent international rivals from having it.
As is everything coming out of the US these days.
If you have a person as president that changes his opinions faster and more often than their underwear, you're simply not reliable.