GPT-6 Astra vs Claude Fable 5.1, We Should Pause AI & The Benchmark Wars | This Week In AI

Shane Thomas (0:00)

If there's one thing I'm looking forward to, it's a different writing style, because I'm so sick of just seeing Claude slop, and you know the patterns, and I just know you had Claude write that.

Abhi Aiyer (0:10)

Is it load-bearing?

Shane Thomas (0:25)

This is Agents Hour. It is Tuesday, September 8th, and we're gonna do the news. Astra is gonna be the headline. But before we do that, I saw this come out. It's the trailer for Artificial, starring Andrew Garfield as Sam Altman. The film follows the story involving OpenAI focused on the firing and rehiring of Sam Altman in theaters on Christmas. Did you watch this trailer?

Abhi Aiyer (0:47)

I feel like it's a weird casting decision. I don't know, I'm not gonna watch it on Christmas.

Shane Thomas (0:52)

I'm not watching it on Christmas, that's for sure.

Abhi Aiyer (0:53)

I mean, I may never watch it.

Shane Thomas (0:55)

I probably will watch it.

If for nothing else than just being able to talk about it on the show. Yeah. Countries will fall.

Abhi Aiyer (1:04)

Industries are gonna collapse.

Shane Thomas (1:09)

I'm like, what's with this bunker with all these guns, you know? Yeah, dude. How does this play into the actual story of him getting fired?

Abhi Aiyer (1:18)

How's a gun gonna help you in there?

Shane Thomas (1:20)

Yeah, is this like his bunker, personal bunker?

Abhi Aiyer (1:22)

Probably.

Shane Thomas (1:24)

Seems, uh...

Abhi Aiyer (1:26)

Who are all those people at his house?

Shane Thomas (1:27)

I don't know.

Abhi Aiyer (1:28)

He really got his walk down, though, and mannerisms, like how he walks and like with his head up like this and stuff like that.

Shane Thomas (1:35)

Definitely, uh, studied. He studied for this role, which is good. So there you go. We'll see what happens, see how it lands.

Abhi Aiyer (1:41)

That bunker looks tight.

Shane Thomas (1:42)

Yeah. I mean, that can't be real, right?

Abhi Aiyer (1:44)

Maybe it is.

Shane Thomas (1:47)

We got to talk about Astra because this was announced on September 3rd. This was like the teaser, the teaser clip, but then it actually was announced, I think the next day, right? Create a yellow circle there.

11 million views. Just got a lot of people knowing it's coming and that was it. Then if you watch the actual launch video, which we're not going to do right now, it shows and it's kind of like a teaser because of how big of an upgrade it was for computer use type applications and the video shows people just like talking to it and having it do things on their computer. Of course, in front of a big screen, so it's really visual and pretty impressive. What were your thoughts?

Abhi Aiyer (2:32)

First impressions, I thought one, the computer use stuff was such a perfect angle to get more users into the ChatGPT verse.

There are many people who are not writing code, but that are technical on their computers, Blender, CAD, artists, etc. And now they're pretty much, they're having their moment, right? They're building, also 3D rendering is something that Astra does very well. A lot of Fable verse Astra on Blender models and things like that. The problem before was, people were trying to make these tools work as like tool calls for the agent, but now they don't need to. They can have, you know, the native software and Astra can just run it and use it. A lot of Microsoft Paint jokes out there, but it's quite incredible. I don't know if it's fake or not, you know?

Shane Thomas (3:25)

I have to run the experiment myself. A lot of people like pulling up a picture of themselves and then telling Astra to draw them in paint, and it's so good.

Abhi Aiyer (3:32)

So good.

Shane Thomas (3:33)

I mean, it is obviously way better than any human, I mean, not any human, but 99.9% of the humans could actually design it, right?

Abhi Aiyer (3:40)

And the launch video kind of showed them drawing a rocket in paint and then saying, hey, let's now design this in Blender. I really appreciated all this because it's not just about coding.

Shane Thomas (3:51)

Yeah, it makes it feel like it's built for the real world.

You can build a real world applications on top of a model like this, because you can actually just give it access to your computer. It got me using Codex for a while. I mean, I used MasterCode for everything, of course coding predominantly, but I wanted to have it do things across my computer in Notion without having to use the MCP. It just uses a browser. It was navigating the factory for me and just dragging cards across and doing all the things. It's pretty impressive.