NEW Grok Build AI is INSANE!

**Julian Goldie** (0:00)
Grok Build is actually beast mode, and honestly, everything else I've previously tried with Grok Build was pretty terrible, but this bad boy is looking nice. I'll compare it against what we've created with other models, but this is literally the best build I've seen out of everything that I've tried to create with this same prompt. Now, if we have a look as well, we've got another game right here. Let's check this out. So you can see it's kind of like a very dark first person game. I'm not really sure what it's trying to do. And again, we'll compare that against some other stuff that we've built in a second. Let's take a look at some other builds here. We've got a website that's looking pretty nice and clean. That's looking decent. We have this really cool 3D. It seems pretty cool building like 3D games. That's what I've seen with this. Like websites, you know, similar to everything else I've seen. But coding out, for example, games and 3D stuff, really, really cool RPGs, et cetera. Nice.
And this is actually quite fun to play. It's hard to switch off. So you can see some of the cool stuff that we're building with this. And this is all with Grok Build. Now, the other thing that shocks me about all of this is that you can just log in with Grok Build on OAuth. So if you already subscribed to Twitter, no problem, mate. You actually have the CLI included with your Twitter subscription, which blows me away. So we've been building this directly into the Agent OS earlier today. One thing I will say is like, I was testing out the old Grok Build versus the new one, and this is a huge step up. Now also, what's really good about this is there's a CLI. So you can actually build this into your Agent Operating System as a CLI and have it sit in the next alongside, for example, Claude, Hermes or Gemini, whatever you want, which is pretty nice as well.
Not only that, my friends, you can also plug in Grok with your Twitter subscription again, directly into Hermes Agent, and then you can use it inside the studio. So not only can it code out using Grok Build and the CLI, not only can you build the CLI and add it alongside all your other agents inside your operating system, but also on top of that, you can generate voice notes, you can generate videos, you can actually train Grok to create videos for you. So we've got it trained to create AI avatar videos like you can see. And then also we have images that it can generate too. And it's a pretty good alternative to Minimax on Hermes Agent. Now, Elon actually posted earlier today that he was looking for feedback on this sort of stuff. Honestly, from what I've seen, it is good. It is good. So they said, and this was announced on 25th of May, like today we'll launch in an early beta of Grok Build, a powerful new coding agent and CLI for professional software engineering and complex coding work available now to all Super Grok and X Premium Plus subscribers. Now, I will say some of this stuff not that great. So I want to be 100% honest with you. So if we have a look at this one, for example, I mean, what even is this? Like it's so dark, I can barely even use it. Let me compare that side by side versus Kimi Code. So Kimi Code actually built something similar for us, which you can find over here.
And if we open this up, it's a lot nicer, a lot better, right? Not perfect by any means, but it's getting there and it's a lot nicer than what we had previously. So for some stuff, it's really like mediocre, but for other stuff, it's pretty amazing. Let's take a look at what else he built. Also, this is important to note. So this is the game that we created with Grok Build, as you can see, right? Pretty cool, looks nice, nice and clean, fun little design, et cetera.
Pretty fun to play. Now, if we compare that versus what we created with Opus 4.8, it actually created something better, which I've never seen with Grok before. Like in every single test I've ever done, Claude wins. But this time, you can see how boring Claude's build out for the same prompt is compared to what we created with Grok build over here.
That is very interesting. This is the first time I've ever seen Grok get anywhere near, let alone beat Claude Opus on tests. I would genuinely say it probably beat all of these.