US government orders Anthropic to suspend its Fable 5 and Mythos 5 models over national security concerns

Assuming Anthropic is able to restore Fable in the next few days, there's literally zero point doing any meaningful work until it is back.

What can be done in 100 hours with Opus can be done in 1 with Fable.

Hopefully this is figured out quickly.

Long live open source models.

Rest in peace, Fable 5.

http://europe2031.ai was published 2 days ago their scenario has mild compute throttling outside of US and core allies starting in 2029 safe to say we're on a *much* more aggressive path right now

the real permanent underclass was lack of US citizenship all along

@mattshumer_ Dario should be fired. If this starts 3rd AI winter because of gross stupidity of his mouth and his marketing, the AI bubble will burst and it was cause a global depression.

@she_llac Part of the plan

@mattshumer_ You’re actually retarded

@AbbeylincolnSt1 It crossed a long-horizon-capability threshold that completely changes how you can use it. It’s like GPT-2 -> GPT-3.5

@mattshumer_ I genuinely have no more motivation to work on anything technical right now.

- 5.5 is technically competent but it speaks like a raccoon and unless you’re extremely literal, you’re not going to get anything accomplished.

- 4.8 speaks English, but it’s technically retarded.

I agree with this. When it wasn’t suddenly downgrading to opus it was rather incredible. Weirdly dry and simple responses but it was a different animal - if you promoted correctly. I used chat Claude to help me optimize my Claude code prompts. it found errors, opportunities in the code and executed probably at least 50-75% faster eyeballing it. I just skipped 3-4 weeks of work at my prior pace in the last 48 hours. I better get my money back. Burning tons of tokens dealing with the model dropping mid session tonight and dealing with the major bugs and drift in opus 4.8.

@mattshumer_ tell me you’ve never used Fable 5 without telling me

@mattshumer_ We all use Fable 5 for two days, and suddenly Opus 4.8 feels like GPT 3.5.

This is such an exaggeration, while it is the best model I’ve used, it’s not by the landslide you’re saying. Code quality and task adherence is great, however no where near 100x of opus, plus the churning time is significantly longer. Learning how to use smaller models to produce same quality result will be a skill in and of itself in the next few months and one anyone seriously working with AI for any use case should be striving for. The mindset of just waiting for the next model to solve it or do it better is not the approach people should be taking.

@chiotmode feeling the AGI (fearful) (negative)

@mattshumer_ People like to exaggerate things. You can say it's a better model. But it's preposterous to say 100X of Opus.

@she_llac has nothing to do with citizenship if they're going to shut it down for everyone just in case.

@mattshumer_ Fable provoking such an extreme response from the government likely means no model more powerful than Fable/Mythos will ever be allowed for release since the case for a ban would only be stronger. This is it. I expected an S-curve eventually but not like this. Sad day.

@she_llac this is a good thing. the lightcone belongs to the US, not to a bunch of weirdo nerd cultists and not to a bunch of third worlders

@mattshumer_ So WTF am I supposed to do with my weekend now??

@NatesCode @mattshumer_ You are a dinosaur watching asteroids fall. You better pick up some AI gramps. These young tech billionaires dgaf about your c++. That shit is over 💀