Welcome to the podcast.
OpenAI recently just announced that.
You know they're shutting down Sora, their video model, for good, and people are confused about it.
And all signs basically point back to Anthropic and how much ground they have gained with white collar and productivity work.
And their revenue is absolutely insane, as more and more people are using them to basically power their whole businesses and companies.
So I want to talk about the most recent update that Anthropic did.
Two days ago they released a computer use on Cloud Cowork, which essentially, if you're using Cloud, there's three different tabs the chat, the Cowork and the Cloud Code.
Personally, I'm not a developer, so I don't use the Cloud Code, although I probably should figure out how to use that more.
But yesterday, I spent basically the entire day working with Cloud Cowork.
If you haven't used it before.
I'm going to talk about some of the best ways to use it, some of the best frameworks, what it can do, what it can't do and a lot of the updates.
Now, one of the things that I'm really excited about with all of this is that, if you want to try the latest versions of these AI models, whether that's Cloud or OpenAI or Google or Grok or any other models
You can go check out my startup AIboxai, where you get access to over 70 of the top models and you can actually build apps and tools with them, build automations, so you don't have to sit there and do things over and over again.
The website is AIbox.ai.
You describe the tool you'd like to build and we'll automatically chain AI models together, add prompts and build a tool for you.
So it's linked in the description.
You also could just chat. in our playground with over 70 of the top AI models.
And that is $8.99 a month.
So way cheaper than $20 for, you know, 20 of these different subscriptions.
Okay, Claude Cowork.
The thing that a lot of people have been super excited about is.
Felix Riesenberg, two days ago, said today we're releasing a feature that allows Claude to control your computer mouse keyboard and screen and giving it the ability to see any app.
I believe this is especially useful if used with Dispatch, which allows you to remotely control your computer while you're away.
So two features here, and both of them I've tried.
Both of them are impressive.
Dispatch essentially is and all this stuff.
By the way, you might not see if you're using Clawed, But it's in your settings.
So the first thing with this particular use is that it is for.
When they first rolled it out, it was for Mac.
But I think just today, a couple hours ago, they added it for Windows as well.
So everyone should be able to get access to this.
You have to download.
I downloaded the the specific cloud app for my Mac.
So it's not like on the website or the web browser.
This has to be the app because it is taking control of your computer.
So you're going to go into the settings and you're going to want to toggle on cloud co-work.
And also you're going to want to go and toggle on dispatch.
Now dispatch is a particularly useful feature that essentially allows it to sync up to your phone.
So if your, if your computer's open, and I just have my computer set to, if it is plugged in, not to go to sleep, And so I could be away somewhere, somewhere far away.
And Claude can be essentially working.
You know, working remote essentially takes over my computer screen and can do literally anything I want it to do.
And then with my phone.
If I'm like, let's say, I'm at the grocery store, I can open my phone up, see what what it's working on and chat with it and say hey, now can you please do XYZ.
Basically, I think this is Claude's response to open claw which, by the way, I also tried to set up.
What I did set up open claw as well the day before.
It is complex.
I don't want to.
I don't want to like, I don't know sound, sound too dumb here or something, but But it's kind of tricky.
It takes a while to set up.
You're going to want to watch a lot of tutorials if you're not a developer.
Cloud code is much harder to get set up.
And it doesn't have a lot of the cool user interfaces.
There's just a lot to set up where, in my opinion, if you're the average, maybe white collar person or you're working in a company, you want to kind of be on the cutting edge.
This is huge.
Now what I will say is it is completely taking control of your computer screen and it is sort of slow, especially right now.
And they even admit this in their post.
He said, this will work with both Cloud Cowork and Cloud Desktop app.
You can use it to click all of the buttons in a legacy app that you'd like to automate, or use it to help you debug a native app you're working on.
It's slow, but giving Claude any mouse and keys is so exciting to me.
That's what Felix said about this.
He's over at Anthropic.
So it's slow is an understatement.
It feels kind of like it's honey, if I'm being honest.
So I'm not going to overhype this and say like it's zooming around clicking.
The speed will increase inevitably.
But for the time being, I think they're just trying to be really careful and make it slow.
So why do I bring that up?
It's a very useful tool because you can essentially chat with it and get it to do a whole thing.
So yesterday I had it generating scripts for a project I'm working on.
I had it going and logging into different websites, generating content.
I had it going to AI Box, my own website, and running boxes on AI Boxes and grabbing content.
The output of those boxes actually found.
That was one of the best ways to use.
It was basically to build an automation or like a tool on AI box and then give it a whole project.
But you know, when it came to the content generation, instead of trying to manually give it a huge prompt where it had to add that into its steps,
It.
Just one of the steps was go to this box, this link box on AI box generate, with you know this, these inputs and then grab that and move it to the next part of our project.
I then had it literally opening up and editing videos on CapCut.
So I just told it like hey, on my computer there's a video editor called CapCut.
Open that up, pull this file in, pull that file in and start editing it in this way.
Now, What I will say is, again, not not very fast.
Like you got a lot of the work done.
But, like if you sit there and watch it, sometimes it can feel painful because it's like I am opening up CapCut.
Now I am searching for the new project button.
And then it's like says like 20 things back and forth for like three minutes.
It's like, OK, I'm clicking on it.
And you're like, oh, my gosh, it's so slow.
But so here's where I think that this is pretty useful.
Number one is.
And it took me basically all day to kind of chat back and forth with it and get it to figure out how to do this whole project, this whole workflow.
I mean, to be fair, probably would save me two hours a day.
And it's actually something that I have an employee working on.
But it's something that would probably take two hours a day.
It took me, I don't know, the better part of four hours, not of like constantly chatting, but I'd like send a message.
I'd be working on something on my other computer and then, like 10 minutes later, I'd send another message or correct it.
But the cool thing about it is at the end of a conversation, once you do work out like a whole workflow and automation with it, you can say save this as a skill.
It will create a file where it breaks down everything that worked and also, by the way, sometimes it tries things they don't work and And it tries them a different way.
When it creates that skill file at the end it will just like basically document all the ways that it works to accomplish your task.
So it doesn't have to go and try different things.
And sometimes, sometimes the way this tool works is different than the way that we are used to.
So I remember like when I was first using it, there was I was like, copy this and paste it here.
And it really struggled with like the copy and paste for a long time.
And finally there was like a pop up that like had to, you know, say like hey, do you accept the copy?
Like I can use the copy and paste on your computer.
And I'm like, well, of course.
So then added that to the skill when I ran it again right at the beginning, before I even kicked off the whole thing.
It's like hey, we need access to copy and paste and your browser and your computer files.
And it kind of had like this checklist and I clicked accept all.
Then it could run the whole process where well.
I've kind of worked through it and it was taking a long time.
Yesterday,
It was having pop ups one by one being like, oh, I can't do that unless you give me access to this.
So once it built that skill file, it gave me all of those kind of access requirements right at the beginning that I could check off and it could go.
So It takes a while to get something set up, but once you save it as a skill, then basically the skill kind of has a name and all I have to do is say hey, run this skill again and it will go and do the whole process and it kind of has it all figured out.
Now...
It is slow.
So I would definitely recommend having a separate computer that you are running this on.
And I think you can just get an old laptop if you have one.
A lot of people, of course, are buying Mac minis, but I don't think that's completely necessary at all.
But right now I literally have a couple of a few laptops.
So I just ran that on my Mac all day yesterday and then I was working off of my Windows as kind of my backup.
I'll probably at some point get a Mac mini that is running at all.
I think it works a little better on Mac at the moment.
Well, at least that was what I was thinking.
But we just they just rolled out Windows today.
So it was only two days apart.
I was just thinking they didn't have Windows ready to ready to rock and roll.
Overall, been really impressed with this tool.
There is so much it could do.
I mean, literally, because it can take control of your screen and it can click around.
I've used a lot of these agent softwares for the last couple of years.
OpenAI really...
Kind of was kind of had a good thing going with their agents, but they just weren't quite there.
And this feels much, much more sophisticated.
This feels like it actually works.
And I've been actually saying this for a while because I've been using the Google Chrome cloud app, which is a sidebar which takes control of your Google Chrome browser and get stuff done for you.
So that was kind of my first introduction to this.
Now that they've just made it control your whole computer screen, I'm thrilled and this is honestly a pretty decent tool, especially if you can use it, if you can dispatch it remote from your phone.
I would say throw this on a separate computer and I would probably say build out some skills that you teach it how to do and just have it running those skills throughout your day.
Anything that's pretty repetitive if it's writing newsletters or articles or documents or research, all that kind of stuff, it can get it all figured out.
Now, the one thing that I want to end this on.
I made a post on LinkedIn and it got a lot of traction.
So I think it probably resonated and struck a chord with a lot of people.
And I had basically reshared a post that I saw on X from Thomas Frank.
He said, currently 892 hours into automating a 30 second task I do four times a year.
It's going to be so worth it once I get everything working.
I thought that was hilarious and I think that that's kind of what it felt like for me yesterday.
But I was also figuring out the tool and all the capabilities and how to set it up and how to basically connect it with my Gmail and with all of my different apps and stuff that I use.
So I think at the end of the day, it does kind of feel like this.
It takes forever to figure out some of these automation softwares.
But I definitely think that this cloud computer use is one that's super valuable.
It's worth learning.
Yes, like I could say, it's kind of slow and it's not, you know...
And it asks like for permission to do things, but it does get the task done and it does do it well.
So I would say spend some time on this one.
It's going to get faster, like they're going to make the tool become faster and faster.
Now, if you're a developer, you probably don't even need this.
You're probably running something like OpenClaw or something like Cloud Code that can get a lot of these types of little can get a lot of stuff done for you.
But if you are a no coder like myself, this is 100.
I think the best way to work with agents and to automate stuff.
It is phenomenal.
I would highly recommend it.
All right, guys, thank you so much for tuning into the podcast today.
If this was interesting, try to really just basically spend the entire day yesterday researching for this.
So I would give a high quality episode that wasn't just commenting on it, but telling you what I actually tried and what actually worked.
If this was helpful and you want more episodes like this, please leave a review on the podcast and let me know.
Over on Apple, you can drop a comment.
And on Spotify, you can hit the About section on the podcast and drop some stars.
Helps out the show a ton.
Thank you so much for tuning in.
Make sure to go try AIboxai if you'd like to build automations that your cloud co-work can go and use and make sure that it speeds up the process and gets exactly what you want as far as outputs on text, audio video and images.
All right, guys.
Thanks so much, and I'll catch you in the next episode.
