How Will Sam Altman Respond to Claude Computer Use?

So, Anthropic has just launched computer use research preview again. They seem to be tiptoeing around this for Claude Pro and Mac subscribers only on Mac OS. Letting Claude move move the cursor, click, type, complete forms, all autonomously. You can even assign them from your remote control, and that's exactly what we're going to do. Command AI. Basically, you with your phone, you're controlling your agent that spins off more agents to do stuff on your computer at the same time. So, theoretically, you could have hundreds of agents do different things on your computer. That's crazy. I I've got Claude open here. Open Final Cut Pro. Render the video that opens. Mark the Notion video database as edited. And text me that you've finished editing it. It's doing its thing as well. It's doing its good. From from Here we go. Here we go. Isn't it asking for permission? So, yes. So, now it's asking for Final Cut Pro permission. I was hoping that would light up, but there we go. So, it's asking for Final Cut Pro permission. Even better. Right. I'm going to go get it. Hang on. Stay here. Stay here. So, you can see it's kind of Yeah. orange border You see the orange on the on the border. It's doing things. So, I can see I can confirm that it's rendering. There's a little icon at the top there. Oh. Did you hear that? That sounds like a text message. Claude is short. Edited. Done. And then See that bottom one? See how there's two there? This is the problem. This is a new feature that got released just the other day, plugins, that it can use iMessage. It It needs to message yourself. So, the blue message is me messaging, and then the gray is me receiving it. So, I've been using it. Every single message has got has doubled up. There's where it failed, so that's fine. But, you've got the blue, and then you've got the the gray. But, I don't get the excitement about the iMessage integration that they launched yesterday. However, how [ __ ] cool was that? Yeah, that that That's really cool. Like, what I'm excited about is to see what will be the ChatGPT's version of this because they they hired the open claw uh creator. Presumably, they're building something like this dispatch computer use from Claude, and they will probably have a different way of doing it. And I would like to see a different way because I I want to pick one and be like to see which one is better. Um but, it's crazy how Claude has been putting out just updates after updates. Like, it feels like every day they have an update. too. Not just like Features. Yeah. The thing is is that Claude they've climbed a ladder. They've They started with the bare minimum, and they've really been developing it They've stacked their progress on top. They could not have got to computer use without co-work. They could not have got to co-work without Claude code. So, they've slowly been building up this this ecosystem this architecture that supports some of these really really good features. Whereas, OpenAI really don't. They're just all over the place, which just a quick one. We covered last week. You would shot off, but we covered last week the fact that they were deprioritizing um efforts like uh Sora and Atlas. This week, they actually discontinued Sora. Like, they've got They're getting rid of it. Now, they didn't say anything about Atlas, but that was part of that list and I wonder if they're going to actually remove Atlas from it. Like, they are really really They're They're on red alert without saying it. They're They're absolutely on red alert. But, all that to be said, they just don't have a a substantial ecosystem or foundation of products that can leverage each other. Whereas, Claude or just Anthropic are time and time again demonstrating that they just they they support in the workforce and they're supporting real productivity that real people actually want, not some AI slop generation video platform that no one ever [ __ ] used anyways. So, it's just I don't know. I'm I'm definitely biased. That's a really good point. Yeah, that's No, that's really a good point. Like, the best product by OpenAI, what has it been? It's been ChatGPT pretty like Yeah, and it's just a chat app. Yeah, true. Okay, good good point. Codex Codex app I I use Codex app. Codex app is cool. Um but, like, mass adoption ChatGPT has been a mass adoption. But, what I was trying to say, like, in my head I was like comparing it to Claude code in a way that Claude is built on top of this essentially a terminal with AI on top of it, right? Like, just give AI access to a terminal to do stuff and they they are they've been building building on top of that. And I'm surprised that OpenAI just never did this in a I mean, they have Codex in the terminal, Codex CLI, and that that was it. And they never built a UI for it. Like, you can't do any anything automation or any like computer user or like any like code anything code work with ChatGPT. This There is scheduling or something to effect. Or auto You can do automations in the Codex app. There is an automations thing at the top left there, but like um but yeah, it's not it's not all that exciting, to be honest. You got it? So, you have automations, but there is nothing in it that is like that exciting. These are mostly about code. Um what is new automation? So, I could I could make a new automation. They did release a few months ago that was really interesting. That was to do with automations. Do you remember it? It was like a node-based thing. Oh man, they they they made so It seems like Sam Altman is just going around and saying, "Guys, I have an idea. Let Let's build this thing." And they they build it, but it's half-baked. Idea is half-baked. People don't I mean, it's also something some companies do. They they throw so many ideas at the wall and see which one sticks. It seems like the baseline for Claude has been just so good. Like with Claude code and the model The model's being really good. They They are really They have this really nice personality of not over explaining things to you. They are just to the point. That's the model. And then on top of it, they they use your computer. And on top of that, now they they're building this ecosystem around it. Um and OpenAI had a few dropping ball moments with the Department of War thing and mass uh uninstallments. Yeah, we are we are on the side of Claude, like on this show. This is our This show has been brought to No, not really. I wish they would. Yeah, yeah, that would be good. I I have to upgrade again, by the way. I have to pay to Claude more. What, Max? I I I I was on the Max, but then I went back on the 20 bucks. Yeah, I see. Yeah, I'm staying on 20 bucks just cuz I've got so much I mean we've got I've got two I've got maybe two more months of free Codex, I think. Cuz I Well, Codex has been amazing in terms of like token usage. I've been using a lot of Codex whenever I get the chance to code. But with this patch, I noticed because I can set up multiple conversations, I drain my limit so quickly. Like sometimes it happens in just a few conversations because I say, "Go do this research." And then I say, "Go do another research." And That's how much you get out of the 20 bucks. This is part of a larger conversation on my show, Command AI, which we stream live every single week. We discuss the news and all things related to AI in the world of design and web. Catch us next week and join in the banter. Command AI.