I didn't think this was possible...

If you watched the channel and you know me, I set up OpenClaw on this machine. But here's the thing, it was very difficult and cost a lot of money and the maintenance on this thing. I was spending more time just maintaining my OpenClaw than actually getting value from it. But it turns out there is actually a better way, an out of the box way that is for everyone. I kept hearing about Perplexity Computer. I kept hearing that it is a really good out of the box solution that can do all of the things for you but you don't actually have to spend all that time setting it up and maintaining it. Let me show you what Perplexity Computer is all about today and thanks to them for sponsoring this video. All right, so this is Perplexity Computer. If you've ever used Perplexity, the interface should feel very familiar. And what Perplexity Computer is is you can kind of think of it like OpenClaw but fully hosted and much easier to use. Perplexity Computer allows you to do many of the same things that OpenClaw allowed for. It is a full AI agent. It has access to an environment. It can write code. It can execute code. You can use any of the Frontier models out there with it. It can search the web. It has all these other tools that you can use. But again, it's fully hosted. So if you're looking to have the most control over your AI agent, OpenClaw might still be a better option. But for the vast majority of people out there, I would actually say Perplexity Computer is a better option. All right, so let me show you around and show you what you could do. So first, there are tasks. You can kind of think of these as threads. This is one of the things that I actually had to set up with OpenClaw directly. So by default, OpenClaw ships in a single thread, meaning you're having a conversation with your agent and that conversation can span multiple different topics. Then I set it up in Telegram to have multiple topics because that is the best way to work with your AI agent. It allows the agent to have the optimal context for that thread, so you're not mixing context between different topics. Plus, it's just easier to work as the human on the other side of it. I have different threads for different topics, and I just use the thread for whatever use case that I'm working on in that moment. But, Perplexity by default is threaded, and that is something that makes it just so much easier to use. And remember, Perplexity computer under the hood has a full environment that it can write and execute code within. So, with that knowledge, you can pretty much install anything. You can install different NPM packages. You can have it run other software. You can have it write software that it's going to run. All right, so the thing that stood out most of all with Perplexity computer to me is its connectors. I took so much time and effort, so many tokens spent trying to connect all the different services, all the different products I use, including Telegram and Asana and Gmail and calendar, all of these different things I had to set up manually with Open Claw. And it was fine. I took the time to do that. But, most people again, they don't want to. They want something that just works, and that's where connectors comes in. It is pre-built connections with a bunch of products and services that you already use. And they have a ton. Here's Drive, Gmail, OneDrive, Box, Linear, Dropbox, Notion, GitHub, and on and on. Look, if I just scroll down, look how many there are. These are all pre-built, ready to go. Hundreds of them. And so, a huge gripe with Open Claw was its security. A lot of people think it's not secure. You have to install it on a separate machine, but also you're giving it some of your credentials, and you're building connections with a lot of these services yourself manually. And so, I was fine doing that because I'm technical by nature, and I kind of enjoy getting my hands dirty and getting in there. But again, I think a lot of other people might not want to even have to think about that. And that's what you get with Perplexity computer. That's what you get with these connectors. You just click on it. And you can just add the connector and authenticate, and they've pre-built it for you, and it's all wrapped in Perplexity security. And you're not dealing with API keys necessarily all the time. You're just authenticating like you would through just logging in with Gmail, for example. So, very, very easy to set up all of these different connectors. You can start to put them all together. I have incredibly complex workflows that I've built out with Open Claw that I've effectively rebuilt with Perplexity. And you could just do it. It's easy. And so, next, they also have skills, of course. And so, just like with Claw code, just like with Open Claw, skills are a great way to tell your agent how to do things. And you can do that just as customizable with Perplexity computer, which is really nice. So, I have my food journal, for example. Perplexity ships with this create skill skill, so you can just describe a skill that you want and add it, and it just works. All right. So, let me just show you how to use it. Right here you can see, here's the standard Perplexity search. Here is Perplexity computer. And it looks just the same, except now it is a full AI agent. So, if I click the little plus right here, you can upload files and images, you can upload files from the cloud. Here are the different connectors that we can have active, so I can just turn on Gmail with calendar. Here's the one I wanted to show you. Select orchestrator model. You have three options. These are frontier models. Opus 4 6, GPT 5 4, and Sonnet 4 6. And for those of you wondering, in Open Claw, I am using a split between Opus 4 6 and Sonnet 4 6 as my orchestrator model. Sonnet is a beast. And of course, you can use it right here and it's far less expensive. And I'll tell you a little bit about how Perplexity charges later. So, here's Sonnet 4 6, Gmail with calendar. I can just type, "What's my next meeting today?" Hit enter. And yes, this is an overly simplistic use case, but I just wanted to show it to you working. Okay, so it's going to connect with Google Calendar. And while that's happening, I can click off and do something else. Right? So, it's just going to go all of these threads running in parallel. And not only do those threads run in parallel, but it actually can delegate out to sub agents from within a thread. All of this, if you've used open claw or used any other agentic system, should feel very familiar. All right, so let me show you end to end what a task completed actually looks like. So, I'm a big UFC fan. For the next UFC card, automatically gather the official fight lineup, odds movement, weigh-in results, recent interviews, late injury chatter from ESPN, UFC, and major MMA coverage. Send me a contact fight night brief on Saturday afternoon. So, obviously, this is going to prepare a bunch of stuff, but it's not going to actually show it to me until Saturday afternoon. So, here we go. So, it's running tasks in parallel. It searched the web. It loaded the skill research assistant. Here, it's doing some time calculations. You can see some of the chain of thought. You can see here's the terminal command. So, this is actually code that it's writing in a terminal and executing. And then at the very end, it says, "I have it set up, but I'm going to pause it and wake it up on Saturday at 1:00 p.m. PST, giving me time to do the final research sweep." So, that's all finished. All right, so let me show you one more. Do you remember that food journal that I made with open claw? I was having all these stomach problems, and so I wanted a way to track the food that I ate and to see what might have been causing my stomach problems. And so, I would take pictures of food. I would upload it to open claw. I basically recreated it in seconds in Perplexity computer. Very easy. So, I have my food journal uploaded to this project I'm working on. I'll tell you more about that later. And it allows me to easily install that workflow, the food journal workflow, directly into Perplexity computer. So, here we go. Load the skill looking for the food journal kit and it finished. And then if you could see at the very end, I sent a picture of obviously I just grabbed this off the web, but it's a hamburger. And it said, "Let me now initialize the food blog." So, initialized. You can see the code right there. Reading the foodlog.md. So, it logged your cheeseburger at 9:20 a.m. It's a little early for that, but fine. Caramelized onions, tomato, lettuce, special sauce. Okay. That's your first entry in the food log. So, just like OpenClaw, I can just do it with Perplexity now. Very, very easy. Now, one of the best things about OpenClaw was the ability to have it in Telegram. And of course, yes, it's available with Perplexity as well. But, Perplexity also has a phenomenal mobile app. All right, so here's the Perplexity mobile app and you can see right here all of the tasks that I've already had are still in there easily. So, having Telegram is less necessary because you could just use Perplexity's beautiful app, but you can also connect Telegram. I'll show you how to do that in a second. So, if I click into the food journal, scroll to the bottom, there it is. There's the burger, ran the task, and you can see it all from right there. So, here's another use case that I literally just used it for and super important because I'm going to use this in another video that I'm recording shortly. So, I said, "Gemma 4 was just launched. Get the benchmark data and put together a chart comparing it against all the other relevant models in its class size." Ran some tasks in parallel. It first thought, "Okay, Gemma 3 N." because it didn't search the web. Then it searched the web and said, "Oh, no, Gemini 4 just launched. Let me start to look around." So, ran a bunch of sub agents in parallel, grabbed all the data. Then, I said, "No, specifically compare it against Qwen 3.6, which just came out, MiniMax, and Kimik 2.5." If we scroll all the way to the bottom, we can see the image right here. So, Gemini 4, Qwen 3.6, MiniMax M 2.5, Kimik 2.5, and we can see the benchmarks right there. So, it put together a really nice visual for that. All right, so let me go back to the calendar app. It says that my next event, my son has a piano recital at 3:45 p.m. today, and it looks like my wife is also on the invite. So, very easy, very quick, and it's just basically done. I don't have to think about all of the integrations and connections, and it just works. All right, so I mentioned Telegram. Let me show you how to do that. So, if you go to connectors, go to the top right, type Telegram, there it is. You just click it, add connector, it's going to connect to your Telegram automatically, no need to go through the hassle of finding the bot and doing all that weird stuff, it's just done. All right, so let me show you one more use case. And again, all of these are ones that I've done with OpenGlue, that I was able to replicate in Perplexity AI. So, I had it find my earnings preview workflow, and I said, "Install it." And I can also just simply describe what I want it to do, but I already created it, so I just wanted it to recreate that functionality. And it did. And so, what happens now? Sunday at 9:00 a.m., it goes and it searches for all the quarterly earnings that are going to happen in the following week. And it knows what I'm already into, the types of stocks, of course, tech. And so, it gives me all the tech earnings that are going to come in the following week. It says, "Which ones do you want to see?" I tell it, and then after the earnings happen during that week, it pulls the transcript, analyzes it, and sends me a report immediately after the earnings call finishes. All of that easily done with Perplexity computer. And so I wasn't actually sure where will I actually receive these notifications because it's proactive. I'm not going to just come and check the thread. I want it to tell me and it said, of course, I'm going to send it to your mobile app. Or I could also tell it send it to my Telegram if I wanted to do it that way. So very very flexible. So let me just tell you briefly how you pay for it because it's not free, neither is Open Claw. So you get a Perplexity plan, you're paying for it monthly. You also get Perplexity search, which is great. You get all of these different things in that package. And then separately for Perplexity computer, you get a bunch of credits for it, but you also pay for usage. And you can actually kind of see some of the usage here. So for that next meeting task that we talked about, 15 credits. For the Gemini for benchmark comparison, which came with an image output, 388 credits. I installed my knowledge base kit, that's 1,500 credits. So you can kind of see specifically how many credits you're using and get a sense of which task cost more and which cost less. So that's how the billing works. All right, so I wanted to show you one more use case, probably the most important, most valuable use case that I have gotten out of Open Claw. I'm going to rebuild it in Perplexity computer. And that is the knowledge base. Any links that I come across, videos, articles, tweets, I throw them into a central repository that gets converted into an embedding and it helps me just remember everything that I come across in case I want to reference it in the future in a video. I also opened up this database to the rest of my team. They're able to throw articles and videos and anything into this central database. We allow our agents to look through it for videos in the future if we ever want to reference previous things that we've ever come across. It's so valuable. So let's recreate it. Create a knowledge base that I can throw links into expos, videos, articles, etc. That get converted to an embedding and then I can search against it using natural language and this will be a persistent database. All right, so I'll frame this as a small persistent web app with an ingestion pipeline for links. Here we go. We have a plan. This is all being powered by GPT 5.5. Build a full stack web app, create persistent database, link ingestion, embedding, search layer, operational controls, QA, and deployment. Let's do it. All right, so it looks like it's building it right now. Interestingly, it looks like it also delegated to Opus 4.7 for some of the build even though we did the planning with GPT 5.5. Okay, so after about 5 minutes it's now done building. It gives me a nice project. I can open it right from Perplexity. I can just drop my links in here just like that, queue it up, and it's going to extract the content, add it to the database, everything that I'm currently doing. If I wanted to then expose it to Telegram or WhatsApp or anything like that, I could do that also and it's just all managed for me within Perplexity. Here's my takeaway from Perplexity computer. Obviously, very inspired by Openclaw. Openclaw inspired a lot of different products, right? If you follow me on Twitter, you're probably getting sick of hearing me say they're building Openclaw. Well, Perplexity did build it and it's really good. It's very, very useful. Now, if you are looking for full control, if you want everything to be local, then Perplexity computer is not that. But, if you want the power of an AI agent, something that is as capable as Openclaw, but not have to worry about the security and not have to spend billions of tokens setting it up and not have to worry about fixing it and making sure that it's constantly maintained and updating it, Perplexity computer is a great option. So, let me know what you think. I'm going to drop links down below. Thanks again to Perplexity. And if you enjoyed this video, please consider giving a like and subscribe, and I'll see you in the next one.