Dario ruined it for everyone

if they ban open weight models (chinese open weight models), then it's game over, China has won.

Yes

@gubatron Didn’t have Iron curtain on my AI monopoly board

@MatthewBerman it's almost like the US is turning into China and China into the US

So there will be "AI winter" for the vast majority of people. Welcome to the permanent underclass.

@ScottHazAI It’s really bad

I love your work but you're making some big leaps here. Trump himself called Anthropic a "radical left, woke company." Dario doesn't tell the White House what to do.

Dario told the White House how dangerous his models are, what were they suppose to do?

The White House thinks the models shouldn't be widely released. I'm not saying I agree with them. I'm saying that their reasons for believing the models shouldn't be released go beyond Dario saying they're dangerous

Only because you asked in your vid what solutions do we have … I agree, open source is essential, but pairing it with resilient, decentralized compute layers makes it much stronger for these open source models long-term. Decentralized inference infrastructure helps a lot here. Theta EdgeCloud (@Theta_Network) stands out as one option… already live, running distributed workloads for models like Qwen3 etc on real community nodes. Appreciate you raising these points openly 🙏🏼

You could not be more right! The biggest problem is politicians dont understand AI. And that is dangerous.

Yeah

True 😢

what solutions do we have ? open source for sure, but for how long?

It’s time to boycott Anthropic 🤬

why is everyone blaming Dario?

Watch the video

will do

I don't understand why anyone expects third party developers will continue receiving models from private companies forever. Developer programs are obsolete when the model functions as a third party dev itself to direct customers and resale partners. better get use to open models.

True

100% correct! … I even wonder if they purposely allowed certain prompts to take place to fuel the hype.

Wdym?

Here’s what I mean — I could easily see Sam quietly making sure the right people get shown something scary enough to put his model in the same “too dangerous to release” bucket as Mythos. At that $852 billion valuation and pushing toward IPO, he needs that exact classification either way. Based on following Sam, Dario, and your takes for the past three years, I honestly would’ve assumed you’d agree that feels very on-brand for him.

Not everyone China doesn't care

Dario is guilty of what's happening with AI regulations and remember the essay Dario wrote the day before they banned Fable? Because of that, GPT-5.6 is heading down the same path. So sad. And yeah, Dario is retarded.

This is bullshit. The technology is actually very dangerous. If hundreds of bad actors hack various systems, the world could descend into chaos. This is a really naive take, and I wish you would provide more nuance to your viewers.

If this doesn't stop now, the US is out of the AI game and has just handed it to China. We will see a wave of Fable capable models come open source while US frontier models are busy paying the naked emperor to get released.

Based take. This was Dario's plan all along. Because it was the only way to defeat open source, which they will try to ban on some kind of scale next.

This take is absurd. Trump wants the model reserved for his cronies first. Blaming Dario for Trumps grift is nonsense.

best video i’ve seen in a while

No, Dario just accelerated it maybe three to six months. This was inevitable, but now the reality is forming that NVIDIA and these labs are probably going to be nationalized. They're going to lock these GPUs. And yeah, it's going to be a little scary.

This is exactly what Dario wants it. He’s talked about the security risk of models for the longest period of time. It’s a legitimate concern and now, it’s been governed in a very controlled and safe manner. This directly matches the position he’s had the entire time. A lot of people keep talking about how Dario messed up. He hasn’t messed up anything, for himself. They still have access to the frontier models. The main difference now is that the rest of the world won’t which makes the world safer in anthropic’s eyes, again, directly in line with their overall thesis. Watch out, the war against open sources next.

For the first time in my AI life I just became really sad and a little depressed and I was so scared this was going to happen, I listned to the interview with the ceo of the company and im like bro just keep you stupid mouth shut are people not realizing we have something really good going here and of they would just shut the hell up or play ot down if you will like weve used these simple techniques for years and you know something the government wouldn't have even cared or stepped in but when you make serious statements wether they sound half retarded or not and its open public and the spotlight is on and the world is watching your main objective should be the complete opposite but no you say keywords on public tv that just dont exactly get said these days so yea now the government at this point has no choice but to step in they still have a reputation to uphold and they can't look weak by not acting or doing something which regardless you just let everyone know hey I got this model its so poweful that its basically a super weapon that could cause serious damage well good job dummy now we have to shut shit down cause all the bad actors just got woke the hell up and how stupid can you be were talking that just outlawed AI basically and there's nothing we can do about it, maybe hope that China comes thru for us at this point idk. Next time keep your mouth shut, Matthew i really hope for everyones sake that this blows over with a clear plan in site that makes since and works for the public like they should set up a strict flagging system and 3 strikes your out type deal, basically if someone wants to bypass or jailbreak or groom or use the model in a bad way then ban there account for x amount of time or whatever that law or regulation will look like or maybe even jail time idk. At this point im ok with whatever a long as they keep pushing models and we get back on track to making serious progress and everyones happy and things are just flowing and not this dark cloud of uncertainty and possibly AI getting outlawed.

And the divide starts to widen... hope your canoe can get across that growing moat that will grow to an ocean you can't cross without $$$$ to play. We all knew it was coming, doesn't make it less frustrating to see.

They both want the same thing. Regulatory capture. And they’re going to get it.

I am all for open source, but this is not an ideological / political decision imho, it's a financial one. Most people, don't have $15k+ to throw at GPUs just to host some subpar models... As for cloud subs to the chinese labs, they seem already compute constrained. They struggle already to serve their customers, as people complain about the slowness all the time. If demand increases because people stop using US models, it'll be even slower and degraded UX.

Fuck you, no he didn't, if you think this, you are just a fucking stupid idiot. Go cry to Trump admin, they are banning the models, not Amodei.

Excellent video and analysis Matthew. I sort of was hoping at least this administration wont run into Dario's scheme, but I guess never underestimate government need for control, no matter how much free market oriented they claim to be.

The era of "AI for everyone" is over... sadly.

dario has not regulatory-captured opensource, and as soon as any big AI lab opensources a model it will be under the worldwide protection of the makers and users of opensource which are more powerful than governments

This is truly the ai-pocalypse that we all read about and saw in movies. This is where we’re headed and it’s all cause the US government doesn’t understand AI and Dario

fuck dario x10000 times, this is what happens when you put an asshole as an important face for AI development

Removing the blame completely from the American government is such a bad take. Yes, Dario and Anthropic were obviously doing their thing. Of course they were pushing their own agenda. But everyone else was also pushing their side too. The government did not have to listen to them. They chose to listen to them. They purposely decided which side to take. So at some point, this is not just “oh, Anthropic influenced them.” This is also just a bad administration. Really bad. Bad at their job. Stupid, honestly. It’s an open market. If Anthropic wanted to go in that direction, fine, let them do it. Then OpenAI or Google or whoever could release something better, and Anthropic could end up completely out of the game. That’s how the market works. But the government decided to listen to the wrong side. So yeah, Anthropic deserves criticism, but the administration absolutely owns this too. You can’t just drop all the blame on one company and pretend the people in power had no agency. They made the choice. They had an AI Czar and made the choice. That’s on them.

Open source is obviously the way to go! We can keep ignoring that fact or wake up now.

yea its actually so sad. your right. same with healthcare if your a healthcare startup working on something like cancer you cant just use the frontier anymore, you need to try to get some kind of regulatory agreement with Anthropic or OpenAI to get a model capable of really moving the needle on it, so only big healthcare gets to accelerate, so concentrations of power across the board ... meaning less competition and higher prices in the end. Really dark move

Dario is the perfect example of a meteoric fall from hero to zero, he single handedly accelerated AI regulation and turned the race into a restricted, government-controlled industry.

You missed a key point when you started talking about the government controlling GPU allocation. If that happens China will win the global GPU war. GLM 5.2 was trained entirely on Huawei Ascend chips (using the MindSpore framework). Meaning we can do frontier level LLM training runs on 100% Chinese chips. If the US gov stops allowing NVIDIA to sell that to other friendly nations, then those nations like France will have no choice but to switch to the Chinese chip/software stacks. This will crash NVIDIAs valuation and possibly the US stock market as that stock is a load bearing pillar stock.

only @pmarca can save us now

@pmarca How

@pmarca