Perplexity AI Agents on Mac Mini vs OpenClaw

0:00Perplexity just announced that they're running AI agents on a Mac mini, but over these last 30 days I've been doing the same exact thing. I'm going to be talking about Perplexity's new personal computer and what it means for future builders. So when I saw this, my first reaction was like, "Wow, cool, new product." Just like any other day in the AI space. And I was more like, "Okay, yeah, you know, this tracks as far as like what I've been doing and what I've been trying to build for my automations." You know, because the Mac mini as an AI agent hub is something that I have been living inside of basically for these last 30, 35 days. And it's now one of the biggest AI companies in the world just validated the whole concept publicly at their developer conference. So Perplexity I have used in the past for web search and

0:47now they're completely changing the game with this new release. So I'm going to break it down for you what Perplexity actually announced and how it compares to what I'm currently running and what you should think about if you're deciding between these two paths, whether you're going to choose Open Claw or you're going to go down the Perplexity route, which also costs $250 a month. Because they're not the same thing. Even though they look similar on the surface, Perplexity had their ask in 2026 that that developer conference and they dropped a bunch of stuff. But the thing that caught everyone's attention and TechCrunch, 9to5Mac, Macworld, Digital Trends, all of them was something called the personal computer. And every single one of those outlets mentioned in Open Claw in the same

1:32breath. So everyone's trying to minute this. You have Anthropic, you have now Perplexity, you have all of these kind of more automatic automated agentic workflows being copied. So which it does say something, right? It validates Open Claw, it validates a lot of these more agentic workflows that there is a huge demand for this. So here's what personal computer actually is. It's a dedicated Mac mini setup that runs 24/7. Sounds familiar? It connects basically your local apps with Perplexity's cloud AI and it can route tasks across 20 specialized models. It hooks into over 400 applications, which is pretty cool. And the CEO, uh, Srinivas, said something like, I

2:17thought it was generally interesting, he said, and he quotes, "A traditional operating system takes instructions and an AI operating system takes objective." Interesting. And I think that's a really clean way to put it. That's exactly the shift that's happening right now. So, like on paper, this sounds pretty incredible. It's kind of that aha moment that we all had when we got Open Claw set up and it automated one task for us. And same thing with this, you know, but here are some things that we do need to talk about because Open Claw is more open source and Perplexity is not. So, the first thing right after that is the price and that's going to be really off-putting for a lot of people because you're going to need to sign up for the $250 account. So, that's the max

3:02subscription. You get 10,000 monthly compute credits. I don't know how far that will take you. So, and I don't know how fast those credits burn in, you know, the development process and getting everything set up. So, depending on what you're doing, a $200 a month plan is the floor and that's just to get in the door and right now it's not even available for most people. You got to join the waitlist. So, you can't just go in and start using it today. So, if you compare that to what I'm running here on the Mac mini behind me with the Open Claw setup, yes, things break. Yes, it takes hours to get things working. And the thing is though, it's free. And if you're using Open Weight models through Open Router, you can get your first automation going for next to nothing. So, the nice part is for a lot of people it's open source. There's people working

3:48on this and they're releasing updates constantly. So, I'm running it on the Mac mini right now today. I'm not on a waitlist. My agents wake up every morning 6:00 a.m. scrape the trends, write the content, generate some thumbnails, some marketing copy, got posts ready to post on X, and I paid nothing for the platform itself. Like I do pay to use the AI models. Like I use Opus for the heavy lifting, and then I use Sonnet 4.6 for the cheaper tasks, and smaller cron jobs. But, the agent infrastructure itself cost me nothing. Other than my time, right? The difference that matters a lot depending on who you are. And I think most of us are just trying it out. You know, we're not enterprise level people. We're just trying to get our startup off the ground. We're trying to get our business

4:33shipped. We're trying to figure out how we could take our current business and integrate AI. So, we're really trying to just get our feet wet. So, I'm going to tell you what exactly my setup actually looks like because I want you to understand this isn't theoretical for me. I actually have a team of AI agents running at Open Claw right now. Atlas is the main chief of staff. This is the coordinator agent that executes all the tasks. This is Opus 4.6. He oversees the whole operation. Nerizza scrapes daily trends and drafts the scripts. Ghostface handles more of the video clips. Inspector Deck generates my X posts. And then you got a Method Man handle the other pieces of the content pipeline. These agents run on a cron job every single day. I wake up and I have content

5:18ready to go. Four videos daily, five X posts, blog posts, thumbnails, all generated through kai.ai, and we use the Nano Banana 2 model, which is like 8 cents per generation. The whole thing runs in the background autonomously. I've been running this in production over a few weeks now, and it's not just a demo. It's not a sandbox. It's in my actual business. There are a few other things that I have it doing, like podcast outreach, and help handling and optimizing my website for SEO. So, when it comes to Perplexity and it comes out and says, "We're putting AI agents on the Mac mini." I don't know how I feel about it. I'm not I'm not skeptical about this concept, and I know the concept works. A lot of us are living it today. But, what I'm thinking about is

6:03the trade-offs. I want to be honest with you here because there are some real ones. Perplexity is making a strong argument around security, which is awesome. And I want to give them credit for this because they thought about it. Their system has user approval for sensitive actions. It has full audit trails. It has a kill switch. They're also releasing something called Comet Enterprise, which has CrowdStrike security baked in. And they launched a computer for enterprise version that connects to Snowflake, Salesforce, HubSpot. They're actually clearly thinking about the security-conscious buyer. The enterprise buyer, the person who needs to show their IT department and compliance document. So, really gearing it more for people that are in

6:48Fortune 500 companies, people that need that utmost security locked down. So, I'm not enterprise, obviously. I'm a solo builder. My security approach is a bit different. And I have a weekly security audit in a cron job that fires off on the Mac mini. It checks agent activity, flags anything unusual, but it's something that I built and maintain myself. So, if you're not technical enough to do that, Perplexity's system might be good for you and safer out of the box. But, the real honest thing I'll say about them is if you want guardrails and you don't want to build them yourself, their system has all that built in. So, I'll give them that. But, here's the thing about the cloud dependency that I think that gets overlooked by everyone. Perplexity's architecture is hybrid. The local file

7:34access and cloud-based reasoning, meaning when your agents are actually doing the smart stuff, that's going somewhere and coming back. Now, their search API is processing nearly 10 million requests a day. Their total APIs hit 15 million daily requests, which is up 623% year-over-year. So, they're growing fast and their infrastructure is clearly scaling, but you are still dependent on their cloud being up and their pricing is staying where it is. And the road map is going in the direction you need. So, open claws local, my reasoning depending on which model I'm using, can stay on my machine and go to Claude's API, but the agent system itself, like the actual orchestration, the memory, and the task

8:20routing, like that's mine. That's on my hardware. Nobody can flip a switch and change the pricing on me. Nobody can appreciate a feature I depend on without me choosing to update. I truly believe that for builders who want more control, and they want to be more maybe local first in choosing the right architecture to run your business, not because cloud is bad. I'm not saying anything about that, but because ownership matters when you're building a business on top of something. Now, if you're not technical and you want to get into the space, Perplexity might be a bit more user-friendly and accessible, especially if they nail the user experience, which Perplexity usually does. They're really good at that. Their search product is clean and intuitive. It's the best part about being able to jump onto

9:05Perplexity. It's really easy to use. So, I'd expect personal computer to be just about the same. You know, if you're singing, this is the thing, you know exactly what you're signing up for. You're renting the capability, and at $200 to $250 a month, that's $2,400 a year just for being on the platform. You're not building on something you own. So, that's kind of like, I don't know how I feel about that. I really like more control over kind of my framework, my whole system and setup. So, everyone's doing something different. It might be an option for you. Want to talk about the model situation for just a minute because this is just interesting, right? Perplexity's personal computer supports Claude, Gemini, Grok as the underlying AI

9:50models, which is pretty smart. They're not locking you into one model. They're giving you routing across 20 specialized models depending on the task. That's actually similar uh to how I think about it. Like the smaller tasks get routed in actually Grok 4.2 just released on open router and the test results are really good. So, that's how I'm thinking about it then as well as Perplexity. So, like I use Opus for all the complex reasoning, Sonnet for the faster, cheaper stuff, and the philosophy is just about the same. Like different models for different jobs. But notice what's not on that list. OpenAI. That's insane. Like which kind is is like a statement, right? Like, "Hey, we're not going to put OpenAI on this." So, there's been a lot of noise about Perplexity's relationship with OpenAI, and this shows

10:36the truth. And the fact that none of the GPT models are in the mix here is something that people are noticing. So, there was a rumor that also Apple almost acquired Perplexity at one time before they ended up partnering with Google for Siri. So, if you think about that for just a minute, if that deal had actually gone through, this whole conversation would be completely different. Instead, Perplexity is out here validating the Mac mini as the AI agent hardware of choice. Which if you've been watching this channel for just a couple days, you knew that this was coming. The Mac mini wave is a hit and we're in. They're very difficult to get. I bought mine 2 months ago when it was still $499 on Amazon. Now, they're pretty much out of stock. Also, the Mac Studio 512 GB are

11:22completely gone. And it might be hard to even get your hands on one anytime this year. So, where does that leave you and I? And And And how kind of I'm thinking about it. If you want to get into AI agents and you have money to spend and maybe you want someone else to handle the complexity, maybe Perplexity personal computer is going to be worth that for you when it actually launches. Keep in mind there is a waitlist. So, keep an eye out on that. Their product is probably going to be polished, the security story is solid, and the 400 plus app integrations are real. For non-technical person who wants an AI agent setup without building it themselves, that could be probably the best option. But if you want to own your own setup and you want to get your hands dirty and get building, and you want to actually understand how this works,

12:08building on top of it and not having to pay $200 a month just for the privilege of using it, Open Claw is available right now, today, free, open source. You could go download it and get it going within couple minutes. The community is active as well. They are shipping updates like crazy. The documentation is real, and I can tell you from personal experience that it works in production. And if you want to actually learn how to build stuff like this and have a similar kind of setup on mission control and getting a dashboard going and automating a few agents, that's exactly what we do inside of Shipping School. We have over 50 people that have signed up in the last 3 days, and we do four live calls a week with also one-on-one access with myself. The cool part is I have a Next.js starter kit to get you moving

12:53fast. This is what I used to launch three of my apps last year, and you'll be able to hit the ground running with that starter kit. So, I walk you through setting up agents, building your own pipeline, using Claude code and Open Claw together, which other route you want to go. It's not a course you just watch by yourself, sad and depressed. We actually work with you hands-on. And it's me working with you to actually ship the thing. So, if that sounds like what you need, I'll put the link in the description. You can check it out. So, the big picture here is, I would say, what Perplexity just did is basically validate the whole concept. When a company with 10 million daily API requests runs a developer conference and announces they're building the same thing you've been building in your living room, that's not a coincidence. That's the market catching up to what

13:38builders have already figured out. The Mac mini as the AI agent hub is not a niche experiment anymore. It's where this is actually going. The question is whether you want to be on the rented version or the owned version. And whether you want to wait for the wait list or just get going now. I made my choice a couple months ago, and this is what I went with. I didn't need a Mac mini, but I went ahead and got one anyways. And at 6:00 a.m. when those cron jobs kick off and agents are running, it's just a beautiful thing. The whole content pipeline is churning and everything that I have automated to this point has bought me back a couple hours every week, which in the long run adds up. I got a family. I got four kids. I want to spend my time investing in my family and buying back my time, so

14:24I'm not doing all these redundant tasks myself. The main part is it's working. Now, it's your turn to basically decide what path are you going to go down? If you're just getting started into hearing AI agents and you want to maybe get your hands dirty, and you see yourself doing this in the long run, I would highly recommend get going with Open Claw. Get in the terminal if you want to start working more, you know, in cloud code. Start breaking things, and that's how you learn the best. You start by doing. Stop overthinking it and get going. So, I want you to start building, and that's where Shipping School was birthed out of. Community of builders. Everyone's doing different things, and I want to be able to help people that are really passionate about and see the potential with AI. So, like I said, I'll put the link down below, and we'll see you guys in the next one. Be blessed.