Wes Roth

But I don't think Anthropic is saying that. what they're saying is actually a lot more interesting.

They're grown. We we engineer the data centers, the data we put in, the the training processes.

I'm just here to help you. No, but seriously, I have seen too many of those videos.

So keep in mind Nvidia wants everyone out there to be running agents and using inference and therefore using GPUs using.

Most of the issues I ran into came from the merchant side and specifically how the agent was navigating the website.

So, if you get stuck or somewhere as you're setting this up, something goes differently from what I showed you.

I mean it's it's very viable. It's very reasonable. We've seen for example Google and Alpha Evolve they've published.

Auto researcher is on GitHub. It's open source so anyone can create their own version of this and it's very possible.

Now, in this result, we don't know which model is which.

They're building an alternative to to GitHub and they're building it specifically the infrastructure from the ground up.

the model with this hint that is the teacher model and the student is trained to kind of move towards this model.

So you do the problem and you check if you got it correctly from the answers like in the back of the book.

For most of the last decade, building that pipeline required a entire team, a writer, a video editor.

That's how content empires get built. One idea, 15 different surfaces. The bottleneck has always been the surfaces.

They start with high-quality content and they use AI to just scale up for the production.

People laughed. People thought that he was silly, that he was crazy.

Labs are talking about us being very close to that. So, depending on what's happening kind of behind the scenes.

So there's 10 steps between an ant and a human then you might have to go up thousands of steps before you reach that.

It's kind of hard to tell who's in the right and who's in the wrong without knowing exactly what it is that Amazon.

Probably not. We can assume. But they do have the the research from Amazon saying this is a problem.

We can make whatever changes we want and export as both PDF. We can change the actual code to fine-tune how it looks.

Bedside manner, if they were confused about anything, how long they had to wait for for stuff, cleanliness.

So, he asked Thiebeault to solve this problem first generating a complex 19-page design document and then executing it.

And this was different from the regular classifications, cybersecurity for biology, chemistry.

They're in the club. They get access. And, you know, the rest of us, the general public access, the peasants.

And also because you're dealing with the government, the military. I understand that this is not perfectly analogous.

Unlike our interventions for cyber security, biology, chemistry, and distillation.

Well here's the answer. They didn't they didn't release the model.

So, this isn't like the labbotomized version of Mythos.

Anthropic announces Claude Fable 5 and Claude Mythos 5. And this whole thing is pretty juicy.