Now ChatGPT Has It's Own AI OS

Hinzugefügt: 2026-04-29

[00:00:00]I built a brand new AI operating system, not for cloud this time, but for chatpt 5.5. And what I want to do is I want to show you what we built out in codec. So you can do the exact same thing that everyone is talking about on YouTube and starting to build is this whole idea of having an AI operating system. something where you can basically talk to a chief of staff and run your business from a single prompt or just interacting with co-work for example or code in Claude.

[00:00:32]But now in this case I've added it so you can do the exact same thing in chat 5.5 and specifically codeex which is what I'm in right here. So I'm going to basically show you how I set it up and then we're just going to go and test it and see what happens. Now, if you haven't been kind of paying attention, uh, in the last while, there's been a lot of people on YouTube talking about having an AI operating system. And what this is, of course, is just a way for a person and or a business to run all of the things that you would normally do as part of your personal life and or your business without having to repeat yourself all of the time, without trying to figure out how to do things. It's just about, hey, I want an email marketing campaign. I want to launch this product. I want to create this course or I want you to organize my trip and not give any context because the operating system has been learning from you, improving, and you've already given it the skills andor in my case the employees that it needs to get the job done. Now I've used the analogy for example here and I'll just pull it up of an org chart and how I actually do this uh right now is uh for my example I basically created a chief of staff. I call him Dean. He is the only person that I interact with when I'm using Claude and he will be the only person I'll be interacting with in codeex 5.5 you'll see. And I broke mine into departments where I've got a marketing manager, community manager, education manager, and a sales manager. And they have a bunch of digital employees in their departments. And I think of skills, which you may have heard of as Claude skills. I think of them as just really, really, really well-trained employees. Call them agents, call them employees. They're just really good at doing one thing because they've been trained on it. And I have departments because many of these employees do separate tasks, but they're in same categories and they may have to work together. And they're in departments because some of the more complex things that you would ask and or and do in a business might go from department to department. And this is basically set up so that you have it. And it uses at the bottom here, we've got a shared memory layer. And I've covered this in a number of videos, but this is a living learning organization. It's running on an operating system that is powered by AI, in this case, Claude. Now, the question becomes, CHPT 5.5 came out, and it's supposed to be really, really good with agents, and it was designed specifically to work with agents. Plus, they've got that new um uh image 2.0, which does some stunning work. So I thought, well, maybe we should try and see if we can get it uh running or my AI operating system running in codeex.

[00:03:29]And guess what? It works. Uh I created it or in codecs. And again, I'm on the Mac right now in the desktop app. And um you could basically see what I've got here. I've got my Dean style operating system that works across Claude, Codeex, and Gemini, and future ones. So I've got this set up now that it is independent of what large language model that is being used. So it figures out which large language model it is using andor installed on and it does the exact same thing. So I have my operating system here which has the memory, the rooting, all of the rules and tool policy all set up. It's got the different departments for my case marketing, sales, education and community. It's got memory of the company, my brand, all the active things that I'm working on, all the decisions I made and preferences and log files.

[00:04:26]We've got Kaparthy's wiki, which was talked about, I guess, two weeks ago about a way to have a second brain or knowledge for your operating system.

[00:04:34]We've got all of the skills, connectors, and all of these adapters. And the cool part is is this is now platform agnostic. Doesn't matter which one you're using. We now have a plugin that works. And this is going to be really cool because each time or each week something new comes out and you're figuring, "Oh, that one's better now than the one I'm using." So, what do you do with your AI operating system? Now, we don't have to worry about it because this one works across all of the major platforms. And those are the three that I pick. Of course, we could probably set it up for another one, but I wanted to have something that is portable irrespective of all of the changes that are happening because it's my learning business. It's my iterative learning, reinforced learning business. The more I use it, the more skills, the better employees, the more I get the departments organized, the more that they work through things, the more they learn, the better memory they have, it's only going to go get better. So, what I did is I basically built it out. And in my case, what I want to be able to do is let's go and try it. So, um, let's see what we have here. I'll just kind of show you what was actually built. So, on this one, okay, here we go.

[00:05:52]Here's all of the files that were built out for me here. Um, and basically it just says all I need to do to get started is I'm just going to copy this one. And it's installed, ready to go.

[00:06:05]I'm just going to open up a chat and I'm going to say, "What should I work on today?" And it should recognize that.

[00:06:12]So, let's go do a new chat. I'm going to open this up and I'm working here in uh permissions. So, I'm going to do default ones or auto review or full access. I'm just going to do default ones here. And on this one, let's go and add files or more. I'm just going to say, "What should I work on today?" And let's see what happens.

[00:06:36]There we go. I'll use the Dean operating system to uh flow to look at what's active and turn it into practical today short list. So there we go. I found D in a workspace rather than file sitting at top level. So I'm reading it now. And again, this is the first time I've run this. So this should be pretty cool to see what happened. It's recognizing all of my agents that I've got built, the goals, the system, the offers, my preferences, active stuff, and decisions that have been made. Uh, and then recommended next actions.

[00:07:08]Uh, make a week cult plan and save the outputs. No blockers are recorded. So, hey, we're set up. So, I'm just going to say on this one, create a draft.

[00:07:21]Um, let's say uh email campaign um for my AI operating system that runs on chat GPT and Claude.

[00:07:50]Now, this one has all of my background information. It has my memory because when I ported it over, I copied the one that I had in Claude, but I'm working in chat GPT 5.5. I'm not working in Claude, but it has access to the same data. So, here we go. Route it through the Dean workspace is a sales and marketing task.

[00:08:09]It organized by which category or department that should be done. So, Dean is the person that I'm talking to. It knows what employees are available, what departments are available, what tools they have, and let's see what we've got here.

[00:08:24]Uh, renewable draft box update Dean's memory so that it becomes part of it.

[00:08:28]So, this is pretty cool. It's running here. Maybe see if I can get it on another screen so it makes it a little bit easier to see. So, wish I could make this bigger. I got to figure out how I can do the bigger screen here. Let me just try this one.

[00:08:47]There we go. Okay. Hopefully that's a little bit better. So going through the routing sales draft with marketing positioning support. Remember I have my department with their own style and brand information. Since the offer file is still empty, I'm treating a operation uh as the core offer. Writing usable first campaign around that assumption.

[00:09:08]Saving the campaign. It's putting it in work folder. Remove draft. I'll update Dean's memory so it becomes part of the MVP test trial instead of a loose artifact and it's still working these now. Uh it's been working for a minute and 28 seconds. I'm really happy about this. So just think what happens is if you start working and you get an AI operating system, irrespective of what happens with all of the big tools, we're going to be able to have this actually run and it's already come up with some work for us. So let's take a look at it.

[00:09:40]Sorry, we'll bring it up. Um, active can't context.

[00:09:45]So, Claude co-working confirm later. So, each includes a preview text, full email copy, suggested send sequence and notes for refining. And wonder if we open this up, see what happens. There we go.

[00:10:01]Okay. Draft email campaign audience positioning email and the problem. And it's actually got everything that is set up here for our email. Now, I also notice it's even using the smart tags for the email marketing system that we use that are built in there, soft invitation, email followup, suggested sequence, uh, notes to refine and replace placeholders in the thing. So, this is 5.5. I've basically got now Dean set up in chat GPT. Um, this is something I'm going to continue to play with, but I want you to consider the fact that as we move forward and all of these new tools come out. If you have an AI operating system, you don't have to worry about all of the big changes. The tools, the agents, the skills, the connectors, whatever you want to call them, the large language model tools are all now independent of the operating system.

[00:11:01]This is incredibly cool. If you want to see more of this stuff, like and subscribe to the channel. I'm going to continue playing around with this uh and uh this is now something that you can install directly on codeex 5.5.