Google Gemini + Perplexity Comet Enterprise + Claude Dispatch VS OpenClaw + GPT 5.4 Mini And Nano

Key Takeaways

Google just dropped a new free personal intelligence update to Gemini and it could change AI forever. Yesterday, Google handled every single person in the United States, something that used to be expensive. And I don't mean a small upgrade. I mean I don't mean a a minor tweak. I mean Google just gave free user access to something that could genuinely change how you use the internet every single day for the rest of your life. Google is bringing personal intelligence to all Gemini users. The free ones, not just people paying for subscriptions, rolling it out across AI mode in search, the Gemini app, and Gemini in Chrome. The moment I saw this, I had to talk about this because here's what this actually means. So, I want to make it really simple. Imagine you had a personal assistant. Not a robot, not a chatbot, a real human assistant who has read every single email you've ever sent, who has looked through every photo on your phone, who understands your browsing history, who knows what you bought last month, who knows where you've traveled even, and what food you like. Now imagine that assistant can answer the question you have in real time. That's what Google just gave away for free and I keep saying this and people keep looking at me like I'm exaggerating. I'm not exaggerating. Let me show you exactly what this does and why it matters to every single person watching this right now, no matter what you do for a living, no matter whether you've touched AI in your life or not. Let's go. So, first of all, what even is this thing? What is personal intelligence? Personal intelligence taps into Google Workspace, Gmail, calendar, Drive, and more, plus Google Photos, YouTube search history, Maps, and other first-party apps to provide responses that uniquely relevant to you. Now, what does that mean in plain English? Well, it reads your stuff, your emails, your photos, your search history, your YouTube, and then it uses all of that to give you answers that are actually about you, not generic answers, not answers for some random person, answers for your specific life. Whether you're looking for a specific brand of sneakers you previously purchased or planning some sort of getaway based on your hotel confirmations and past travel memories, personal intelligence helps you find exactly what you need without giving it all the context, right? Without giving all the context. That's the phrase that matters because right now, when you use any AI tool, you have to explain everything from scratch pretty much every single time. You have to say, "Okay, here's the context of the conversation. Here's what it's about, etc." And that moment is over. Google just built an AI that already knows you, already before you say a single word. Let me give you the real-world examples that Google has been using to explain this. And these are examples that may not be impressive on the surface, but when you really think about what's happening underneath them, they're kind of wild. So, Nick Fox, a Google engineer, was standing in line at a tire shop and didn't know his tire size. He asked Gemini, but it went further than just finding generic tire specs. It suggested all-weather tires, specifically referencing road trips that he found at home in his Google Photos. Then, when he needed his license plate, Gemini pulled the seven-digit number from a photo and helped him identify his van's specific trim by searching his Gmail. Now, I know what some of you are thinking. "Julian, I can Google my tire size. I can look at the sticker inside my car door." I hear you, right? But that's not the point. The point is that it didn't just answer his question, it connected things he never asked it to connect, photos, emails, his car, his travel habits, his past. It reasoned across all of it at the same time. Now, that's not a search engine, that's not a chatbot, that's something totally different, something new. You can even ask questions like, "If my life were a movie, what would the title, movie genre be?" Or "Describe my perfect day." And it can actually answer those questions based on your real data. That's not a gimmick, that's a reflection of how much data this thing has access to. And when an AI can describe your perfect day based on your real life, it knows you better than most people in your office know you. Let me stack some more examples together because one example never tells the full story. Three examples show you the pattern. Robby Stein, VP of product at Google Search, described a shopping scenario like this. "If you need a new coat for your upcoming trip, AI mode could automatically take into account the brands you prefer, as well as your flight confirmation in Gmail to identify the destination timing, Chicago in March, for example, and then give you suggestions of windproof, versatile coats that fit both the weather and your preferred look, like a personal shopper who already knows your itinerary and the vibe you're going for." That's example one, shopping that already knows your life. Example two, with personal intelligence, you could be looking for a new bag to match shoes you just bought. In Chrome, you'd see a range of options tailored to your recent purchases and preferred brand and styles, including subtle details like purses and hardware that matches your new gold shoes. That is not a search result, that's a stylist who's memorized your closet. And example number three is the one that gets me, travel planning. So, you're planning a vacation and searching for things to do and places to eat that everyone in, you know, everyone you're going with will enjoy. With personal intelligence, AI mode can draw on your hotel booking in Gmail and past travel memories in Google Photos to suggest a tailored itinerary, suggesting something for everyone, including recommendations like a old-timey ice cream parlor based on many ice cream selfies stored in your Google Photos, right? And these could be like ice cream selfies from years ago connected to a hotel booking you made last week to suggest a restaurant on a trip you haven't even taken yet. That's not a feature update, that's a fundamental shift in what a computer can do for a human being. Not Now, let's slow down for a second because I can already hear the skeptics and honestly, the skeptics have a point worth hearing. Some of you are sitting there thinking, "This feels weird. This feels creepy, right? I don't love the idea of Gmail or and and Google Gemini reading all my emails, looking at my photos, and knowing what I buy." And that's a fair thought. That's not an irrational action and I'm not going to dismiss it. On the privacy side, personal intelligence is opt-in and you must first connect Google Workspace or Photos before activates. Plus, you can disable that any time. Google says it doesn't train directly on your Gmail inbox or Google Photos library and training is contained to limited information like specific prompts in AI mode and the model's responses to improve functionality over time. So, it's opt-in, you turn it on, you control which apps connect, and then you can turn it off. And I want to be honest with you, the privacy debate here is real and it's complicated because there's a real difference between what Google says it does and what it could do. And there's a difference between "We don't train directly on your Gmail" and "Your data is completely safe forever." >> [snorts] >> Google's approach relies on centralized cloud-based training. Apple's architecture, by comparison, is designed with default on-device processing, fire-walling personal data from the cloud unless explicitly escalated. And that's the actual difference. Apple keeps your stuff on your device, Google keeps it in the cloud. Both of them are trying to build AI that knows you. They just have different philosophies about where your data lives whilst they do it. And this matters because this is a choice you're now being asked to make, not in some abstract future, right now today. Do you want an AI that is incredibly useful and knows everything about you and the data lives in Google servers, or do you want more privacy and slightly less personalization? It's a real trade-off. I'm not here to make that choice for you, but I am here to make sure you understand what that choice actually is. Because here's the thing that a lot of people miss in this conversation. The people who benefit most from understanding this are not like tech geeks. They're not the developers, they're not the engineers, they're you. The normal person who uses Gmail every day, the small business owner who does everything from their phone, the parent who has 10 years of family photos, right? Um and that's all of us, that's most of us, right? And we are now sitting on top of a data goldmine that Google is offering to turn into a personal AI assistant. Now, let's talk about what all this actually unlocks in a practical sense. Not the tire example, not the ice cream photo, real concrete things that change how work gets done. Think about the last time you tried to remember something, right? An email from six months ago, a receipt for something you bought online, a booking confirmation you forgot to screenshot, the name of a restaurant you visited two years ago that you want to go back to, a product you looked at once but never bought. Right now, you go digging through your inbox, you search Google, you scroll through your photos. It takes time. Sometimes you never even find it. Personal intelligence can answer questions like, "What was I working on last month based on my emails?" Or "Show me how my photography skills have improved by comparing photos from 2024 and 2026." Or "What topics was I researching six months ago?" That's your entire digital memory made searchable through a single conversation. If you're a freelancer, right, you can ask Gemini to summarize every client you worked with this year based on your Gmail. If you're a small business owner, you can ask it to find every time a certain supplier emailed you and what that discussion was. These are not like extraordinary use cases, but they are the kind of things that eat up your time every single day, the searching, the remembering, the cross-referencing. All of that is now one question away and it's free. That's the part I keep coming back to. It's like a year ago, this kind of functionality didn't even exist at any single price. Six months ago, um it would have been a subscription-based to try it, right? Two months ago, Google was still testing it with paying users. And yesterday, they opened it up to everyone. Google launched personal intelligence as a beta in January of this year, saying it would roll out the feature to eligible Google AI Pro and Ultra subscribers first. A week later, they expanded it to AI mode in Google Search. And now they're opening up to the free tier across AI mode, the Gemini app, and Gemini in Chrome. That's a 90-day window from paid beta to free for everyone. 90 days. So, that's how fast this stuff is moving right now. Let me tell you what that acceleration looks like from the outside. A year ago, AI assistants had no idea who you were. They typed a question, they gave you a generic answer, conversation over. It didn't know your name. It didn't know your job. It didn't know if you were, you know, 20 years old or 60. Every conversation started from zero. Six months ago, the best AI tools started getting memory. You could tell them things about yourself and they would remember for the first time, right? And it was still manual. You still typing things in, but at least it carried over. Three months ago, paid users started connecting their Gmail accounts and Google accounts and Gemini could finally look at your Gmail photos if you gave it permission, right? Incredible for the people who had it, but it was expensive. Yesterday, it's now free for everyone in the US. No subscription, no payments, just a Google account you already have. That's not a trend line. That is a phase change. And here's the thing about phase changes. When something goes from expensive and rare to free and everywhere, it stops being a feature and starts being the baseline. It becomes what people expect. It becomes normal. And anything that doesn't do this starts to feel broken by comparison. Think about maps on your phone, right? There was a time not that long ago when having turn-by-turn GPS navigation in your pocket felt like magic. Now, if your maps app gives you wrong directions, you're furious, right? The magic becomes a minimum. And this is what's happening right now with personalized AI. The bar just moved. Yesterday's paid feature is today's free baseline. And everything that doesn't know you is about feel outdated. Now, I want to talk about what this means for the people watching this who are trying to use AI to get ahead in their work because this is where it gets really interesting. Most people in AI right now, and using it right now, use it like a library, right? So, they go in, they look something up, they get an answer, they leave. It's a search tool. It's not Google, a better encyclopedia. That is the lowest level of how you can use this technology and it's still useful. But it's like buying a Ferrari and only using it to get groceries. When personal intelligence starts to unlock, this really takes it to the next level where AI becomes a tool that knows your context and works inside your actual life. Not a search engine, not a question-answering machine, a system that understands your situation and can help you navigate it. Here's a simple example of the difference. If you ask a regular AI, "What should I pack for a trip to Chicago?" it tells you layers, comfortable shoes, an umbrella, generic, fine, but very forgettable. If you ask Google's personal intelligence the same question, it knows you're going to Chicago in March because it read your Gmail booking. It already knows you prefer a specific clothing brand because of your purchase history, right? And it gives you specific tailored answers that make sense for your actual trip. Same question, completely different experience because it knows you. This is what I mean when I say we're living through something genuinely unusual. Not unusual like a cool new app, unusual like a shift in what tools are capable of. The difference between a tool that responds to you and a tool that actually understands you. I want to steel man the people who think this is all overblown because they exist and I've talked to them. The argument goes like this. Google has had access to your Gmail for years. They've been personalizing search results based on your history for over a decade. This is just another version of that. Nothing fundamentally new, just better marketing. And there's some truth in that. Google has always known a lot about you, right? Targeted ads have always used your data. This isn't the first time they've connected their apps. But here's what's actually different. Before, Google used your data to show you ads and slightly tweak search results. You never talked to it. It just watched you and made silent adjustments. You never asked it a question about your own life. You never synthesized your photos or your email to ask something in natural language. Now, you can have a conversation with all that data. Now, you can ask it questions and it can reason across things you never thought to connect. And this is different, right? This is new. Not because of the data, because of what you can now do with it. Search engine land noted that personal intelligence pushes Google further into fully personalized search using first-party data like Gmail and photos in a way that makes results harder to replicate, against or track, especially in AI mode where outputs may vary based on user history, purchases, and behavior. It's very harder to replicate, right? That's the key phrase here. Google is building something that no other AI company can copy because no other AI company has what Google has. They don't have your Gmail. They don't have your photos. They don't have 15 years of your search history. They don't have what you watch on YouTube, right? This is Google's moat. And they just opened up to everyone for free. Think about what that means competitively. Every other AI company has to answer the question, right? How do we get close to what Google just gave away? The honest answer is for most of them, you can't. Not without the data. And the getting the data takes years, right? Meanwhile, anyone with a Gmail account, which is roughly 2 billion people on the planet, just got access to a version of this today. 2 billion people. That's the scale of what happened yesterday. Let me take a step back and talk about the bigger picture here because this launch doesn't just happen in isolation. It's part of something larger, right? Gemini's personalization features could compete directly with the Siri personalization that Apple plans to bring to Siri later this year. It's connecting Gmail and other apps to Gemini, mirroring some of the functionality that Apple is introducing for Siri, which will also be able to read emails, messages, files, and photos, and more, learning information about the user to complete tasks. So, Google just went free with this and Apple is building the exact same thing due later this year. And Apple reportedly struck a deal where, here's the wildest part, Google's Gemini models and cloud technology will be used to power the next generation of Apple's foundational models and a more personalized Siri coming in 2026. Stop and think about that for a second. Google is building the AI that powers Google, right? And Google is also powering the AI that powers Apple. The two biggest ecosystems on the planet are both moving forward and toward personalized AI that knows everything about you. One runs on Google servers, one runs on Apple's devices with Apple's privacy standards, but the intelligence underneath both of them is increasingly coming from the same place. That is convergence. The entire tech industry is moving toward the same thing at the same time. AI that knows your life, AI that connects your data, AI that can have a real conversation about who you are and what you need. And the question you should be asking yourself right now is, where am I in this, right? Because here's the honest truth about what's happening. Every week, tools that used to require expertise or technical skill are becoming free, simple, and available to everyone. And every week, the gap between people who understand what's happening and the people that don't is getting wider. The people who figure this out early, who learn how to actually use these tools inside their real work, their real lives, their businesses, are going to have an advantage that compounds over time. Not because they're smarter, not because they work harder, because they are using leverage that most people don't even know exists yet. And look, I get the overwhelm, right? I talk to people every day who say, "Julian, there's so much happening with AI, I don't even know where to start." I understand that. The pace is genuinely a lot. Every day there's something new. Every day there's something wild, right? And it can feel like you're trying to drink from a fire hose. But here's what I want you to hear. You don't need to keep up with every single thing. You need to understand the direction. And the direction right now is very clear. AI is becoming personal AI and it's connecting to your data. It's learning who you are. It's moving from a search tool to a life tool. Plus, it's becoming free. That's the direction. Everything else is detail. That's why I built the AI Profit Boardroom, available at the AI Profit Boardroom.com or you can check it out, link in the comments description. Because the people who are doing well with AI right now aren't the ones who know the most about the technology. They're the ones who understand how to apply it, how to take these tools, tools like what Google just launched, and actually use them to save time, to scale their business, to do work that used to take hours in minutes. If you're a business owner and you want to know how to actually use AI in your business, not just read about it, but do it, that's what we work on every single day inside the AI Profit Boardroom. Go check it out, AI Profit Boardroom.com or link in the comments description. Real workflows, real results with a real community. Now, back to the update. Let's talk about what happens next because this not the end of the road for personal intelligence. This is just the beginning. In the coming weeks, Google will start rolling out personal intelligence to free users of the Gemini app in the US with international availability to follow thereafter. So, right now it's US only. The rest of the world is coming. And that means this is about to hit a global audience at a scale that's hard to wrap your head around. We're talking about hundreds of millions of people who will have access to this within the next few months. And it's not just more users, it's more connection points. Right now, it connects Gmail, Google Photos, YouTube, Search, Maps. But think about what else Google has access to. Docs, Drive, Google Calendar, Meet, Google um ecosystem, right? If you use any of those things, and most people use Gmail, um use most of those things, you're sitting on an enormous amount of data this AI could potentially tap into. Personal intelligence retrieves details about your preferences from text, photos, and videos to customize the Gemini AI mode, answer without you having to specify it in the prompt, without you having to specify it. That's the future this is pointed toward. A future where you never have to explain yourself to a computer again, where the AI already knows the context, where you just ask and it understands. That's coming and faster than most people think. Now, let me give you the concrete thing to do with all of this because I never just want to leave you with information and nothing to do with it. If you're someone who uses Gmail and has Google Photos, which is most of you, here's what I would do this week. First, go and try it. Go to Google Search, turn on AI mode, look for personal intelligence in your settings. See what it can connect. See what it knows. Ask it something about your own life. Ask it to recommend something based on your history. See what happens and get a feel for it. Second, ask yourself, what part of your day involves a lot of searching and remembering? What do you spend time looking for? Old emails, receipts, recommendations, trip planning. That's where this tool is going to save you the most time. Focus there. Third, think about the privacy trade-offs seriously. Not in a panic, not reactively, but actually think about what you're comfortable connecting and what you're not. The tool is opt-in for a reason. You control it. You should make that decision consciously, not by accident. And fourth, this is the most important thing, right? Start paying attention to what AI knows about you versus what it doesn't. Because the next two or three years are going to be defined by how well you can use tools that understand your specific situation. The more you understand that, the better you'll be using these tools when it matters. Because here's the thing I keep coming back to. This stuff is amazing when you pay attention to it. Um and it's amazing whether you pay attention to it whether or not, right? The tools are getting smarter. The personalization is getting deeper. The free tier is getting more powerful. That's the direction. That's where this goes. The question is never whether AI is going to change things. That question has already been answered. The question is whether you're going to be ahead of it or behind it when it fully arrives. And it's arriving now. Not in 5 years, not in 2 years, right now. Yesterday with a free update that just made an AI assistant smarter than most personal assistants you get paid for a living. If you're watching this and you're still not sure how any of this works for you, that's okay. That's what this channel's for. That's what the AI Profit Boardroom is for. Because you don't have to figure this out alone. But you do have to decide how you're going to figure it out, right? That decision is on you. The tools are there, the knowledge is there, the opportunity is there. And Google just gave away something remarkable for free. What you do with it is up to you. I'll see you on the next one. If you haven't already, check out the AI Profit Boardroom. Link in the comments description or go to the AI Profit Boardroom.com and you can learn our practical ways to implement AI automation in your business to scale and save time. Let's see what we got in the comments here. Let's go, Julian. Thanks for being here. Rap says hello. Good see you. Sir, with all due respect, you look and talk like an AI generator. I get this every day, every day. But it never ceases to amuse me, so I appreciate you joining. Hill says, "Can I use an AI agent on my business computer?" Um well, I don't see why not. I mean, if it's someone else's computer, then obviously like you want to probably get their permission first before you start using it. But I mean, can can you like actually in a practical sense use AI agents for business? Absolutely. Like we we have a lot running in our own business. So, you know, yeah, for sure. So, Hill, if you're worried about, for example, um security and that sort of thing, then I'd recommend just using uh something that's I I wouldn't use anything open source, right? So, in those situations, I'd probably go with something like uh Manas. Manas is a really powerful AI agent. And the cool thing about that is like, you know, it's running in the cloud, so it it can't access any of your your personal files or anything like that unless you specifically give it access to them. And another one that's really good is uh Perplexity AI computer as well. So, those two options are really good for for getting good results with AI. Perplexity Comet Enterprise destroys OpenClaw. So, Perplexity just launched something that could change how every single person at a company does their job. Not in a vague, theoretical AI is coming sort of way. In a very specific, very practical, deploy it to your team by the end of the week kind of way, right? Perplexity just launched Comet Enterprise, right? And I want to walk you through exactly what this is, why it matters, and why I keep thinking about OpenClaw when I look at it. Because here's the thing. For the last couple of years, the conversation about AI in the workplace has been dominated by two camps, right? You've got the people who say AI is going to replace everyone, and you've got the people who say it's just a fancy search engine. Both of these camps are missing what's actually happening. And Comet Enterprise is one of the clearest examples I've seen of what's actually happening. Let me explain. Your browser is open right now, probably multiple tabs, right? You're doing research, writing emails, pulling up documents, switching between tools, copying and pasting things from one place to another. That's your job. And that is what most knowledge work actually looks like up close. Not strategy, not creativity, not big decisions. It's tab management. It's friction. It's the same 15 tiny tasks that eat up your day before you even get to the thing you're actually wanting to do. Comet Enterprise is a browser. But it's not a browser the way Chrome is a browser. It's a browser where the AI isn't sitting in a tab off the side waiting for you to ask it something. The AI is the browser. It's baked into every single page, every action, every workflow. You're not switching to a tool, you're already inside the tool. The tool is the environment you work in. This is a fundamentally different idea from what most people think of when they think of AI at work. And here's where OpenClaw comes in. So, OpenClaw, for anyone who's been following this channel, is the open-source AI agent framework that's being quietly becoming the backbone of how a lot of serious AI automation gets built. It's the thing that lets you build AI agents that don't just answer questions, but actually do things, right? Take actions, run workflows, move through multiple steps without someone holding their hand. What Perplexity has done with Comet Enterprise is essentially bring that same energy, that same AI that acts, not just answers philosophy into the browser, into the place where most people already spend their entire workday. Think about what OpenClaw-style agents can do. They can research a topic, pull together information from multiple sources, summarize it, and hand it back to you in a usable form. They can take a task, break it into steps, run those steps in sequence, and complete something that would normally take a human 20 minutes. They can operate across tools. They can do things while you're doing other things. Now, imagine all of that, but it lives inside your browser, right? No separate app, no command line, no technical setup. You can just open your browser and it's there. That's Comet Enterprise. Arvind Narayanan, the CEO of Perplexity, posted about this today. He said, "Comet is the AI browser for the enterprise. Can be rolled out to thousands of devices internally, controllable for what where agents can operate." That last part is important. I'll come back to it later. But first, let's talk about why this is such a big deal for regular people. Not developers, not engineers, regular people who work at companies. I talk to a lot of people in my AI Profit Boardroom community who are trying to figure out how to use AI in their day-to-day jobs. And the number one thing I hear is it's confusing. There are many tools. I don't know where to start. I have to log into one thing for this, another thing for that. My company just won't let me use most of the good stuff anyway. Comet Enterprise is designed to cut through all of that. Here's what it actually does. You're an employee at a company. Your IT team pushes Comet Enterprise to your laptop using the same tools they already use to manage software across your organization. You don't have to download anything. You don't have to set anything up. You open up your laptop and your browser is now an AI browser. Your browser can now research things for you. It can automate tasks for you repeatedly. It can implement workflows without you having to go through step by step doing every single thing. And the admins, the IT people, and the managers, they can control exactly what the AI is allowed to do, which sites it can act on, which tasks it can run. There are full audit logs. They can see what the AI did, when it did it, what it accessed. This is not a rogue agent running wild. This is structured, controlled, enterprise-grade AI that actually respects how companies work. That combination, powerful AI on one side, real controls and visibility on the other, is exactly what most enterprise AI tools have been missing. And it's exactly what Open Claw style thinking has been pushing toward. Let me give you an analogy. Think about email. Before email, if you wanted to send a message somewhere across town, you had to write a letter, put it in an envelope, walk to the post office, and wait days. Email didn't just make that faster, it changed the entire concept of what communication at work could look like. It moved from communication from, you know, something you did occasionally into something woven into every single hour of your day. Comet Enterprise is trying to do that for AI, not make AI faster, change the concept of what AI at work looks like. Move it from something you go to a separate app for into something woven into every single hour of your day, into the browser, the place you already are. And the early adopters are interesting. Fortune magazine is using it, AWS is using it, Bessemer Venture Partners, one of the biggest names, is using it. These are not small experimental companies taking a flyer on something unproven. These are serious organizations that need tools that actually work inside real enterprise environments. Now, I want to steal man the skeptics for a second because there are real questions here. The first question is, isn't this just a fancy search engine with extra steps? And I get that. When Perplexity first came out, a lot of people dismissed it as Google with AI sprinkled on top. That was a fair read at the time, but what Comet Enterprise is doing is different. Search gives you information. Comet Enterprise takes action. There's a big difference between a tool that tells you what to do and a tool that does the next big thing. The second question is, do I actually need this or is this just for tech companies? And this is where I think people are really underselling what's happening. Every company that has employees who uses a browser, which is every company, you know, every agency, every firm, is a potential user of something like this. The researcher who spends half their day pulling together a comparative analysis, for example. The account manager who has to track 12 things across six different tools every morning. The HR person who manually uses and processes the same documents over and over. Comet Enterprise is aimed at all of them, not just the tech industry. The third question is, what about Open Claw? If I've been building with Open Claw, if my company's been using Open Claw and Open Claw based automations, does Comet Enterprise make that irrelevant? And the answer is no. This is actually the most important part of the conversation. Open Claw is a framework. It's infrastructure. It's the thing that you build custom automation on top of when you need something very specific, very tailored, very deeply integrated into your existing systems. Open Claw is powerful precisely because it's flexible. You can make it do almost anything if you're willing to put the work behind it. Comet Enterprise is a product. It's pre-built, pre-packaged, ready to deploy. It's designed to give you the benefits of AI agent automation without the technical overhead of building it from scratch. Those two things are not competing, they're complementary. Open Claw is your custom build, Comet Enterprise is your off-the-shelf solution. Depending on what your company actually needs, you might use one or the other or both. But here's a paradigm I keep seeing that is the thing that I think really matters. The gap between what's possible with AI and what regular employees are using is closing like really fast. A year ago, if you wanted AI that could actually do things in your work environment, you either needed to build it yourself using something like, you know, Open Claw, but it'd be something different back then, or hire someone who could. And that was the barrier. That was the thing that kept most people and most companies on the outside of what was already becoming a major productivity advantage. Comet Enterprise is one of a growing number of tools that are taking that capability and making it accessible to anyone. No build required, no technical team required. Roll it out, turn it on, and your entire workforce now has an AI agent living in their browser. And that's enormous. And it's an enormous change in the accessibility of this technology. I don't think most people fully process what that means. Let me walk you through a specific scenario so this isn't abstract. You're in a sales at a mid-size company. Every Monday morning, maybe you spend the first 90 minutes of your week pulling together a summary of what happened with your accounts, who inquired, what competitors showed up in conversations, what news broke about your biggest prospects. You do this manually because you've always done it manually, and nobody's ever built anything to do it for you. With Comet Enterprise, that Monday morning task becomes something that the browser does whilst you're getting your coffee. It researches, summarizes, compiles. It drops into whatever format you need. You're back at your desk and it's done. That's not a moonshot. That's not a case to be for some futuristic pilot program. That's what this tool is built to do right now today for people at companies like Fortune and AWS who are already using it. And I want you to sit with that for a second because what I just described is not an engineer's workflow, it's not a developer's workflow. It's a sales person's workflow. It's the kind of task that has nothing to do with writing code or building software. Pure knowledge work, research, synthesis, communication, the stuff that fills most people's days. This is why I keep coming back to the Open Claw comparison because Open Claw started as a tool that developers used to build AI agents. And over time, the ideas behind Open Claw, the ideas about AI that acts, AI that can chain together multi-step tasks, AI that operates inside real workflows rather than just answering questions, those ideas have been spreading into more and more products, into things that regular people can actually use. Comet Enterprise is the latest and one of the most significant examples that spread. And there's one more piece of this I don't want to skip over. Perplexity also launched something alongside Comet Enterprise called Computer for Enterprise. And this is where it gets really interesting. Computer for Enterprise connects to your company's existing tools, Snowflake, Salesforce, HubSpot, hundreds of other services that your company probably already uses every single day. The AI can query those systems, pull data from them, create dashboards, build models, put together presentations without you having to go into each system separately and do it hand by hand. Think about what that means for how companies operate. Right now, one of the most common complaints in any organization is that data is siloed. Your customer data is up here, your financial data is over there. Your project management is in the third place, and getting a coherent picture of what's actually happening requires someone to either manually pull things together or have a dedicated analysis team to do it. Computer for Enterprise connected to Comet Enterprise's browser starts to make that friction disappear. The AI knows how to talk to all those systems. It can pull them together with connectors. It can synthesize them. It can give you the picture without you having to build the report. Now, this is still early. I want to be honest about that. These tools are not magic. They require setup. They require someone at your company to make decisions about what the AI is allowed to access, what it's allowed to do, how it fits into your existing workflows. This is not a plug-in and everything is perfect situation. But the direction is clear, and the direction is AI is moving from something you buy a subscrip- a separate subscription for and have to go out of your way to use toward something that lives inside the tools and environments you already work in every day. That is where Comet Enterprise is pointing. That is where Open Claw powered automation has been pointing, right? And that is where the whole industry is heading. And here's the question I want you to sit with because this is the question that actually matters for you, not as a tech watcher, but as someone who actually works. If every person on your team had an AI agent living inside their browser, what would change? What tasks would disappear? What things would that eat your week right now would just get done without you having to do them? What would you actually do with the time you freed up? That's the exercise. Not like, for example, AI is coming for my job. That's the wrong question. The right question is, when AI handles the repetitive, the mechanical, the tedious parts of my job, what do I do with the capacity that creates? What do I become capable of that wasn't capable before? The people and organizations asking that second question right now are going to be in a very different position in 12 months than the people still arguing about whether AI is real or not. And that's why I built the AI Profit Boardroom because this isn't a spectator sport anymore. You can't just watch this stuff happen and assume you'll figure it out when it gets closer. It's already here. Comet Enterprise has just launched. Companies like Fortune and AWS are already using it. The gap between the people using this and the people who aren't isn't theoretical, it's real and it's growing. Inside the AI Profit Boardroom, we walk you through exactly how to use tools like this inside your business, in your workflow. Not the tech talk, not the jargon, the practical stuff. The stuff that actually changes how much work you can do, how fast you can do it, and what kind of results you can achieve. If that's what you're looking for, link in the comments and description, or just go to the AI Profit Boardroom.com. Come and join us. Back to Perplexity. I want to zoom out for 1 second and talk about what Perplexity is actually built here because I think people still underestimate this company. Perplexity started as an AI search tool, better search, smarter answers. That was the pitch, and a lot of people thought of it as just kind of like a fancy alternative to Google, a nice to have for people who liked AI flower flavored search results. But look at what they've actually built over the last year. They built Comet, the browser. They built Computer, the AI agent platform. They built Computer for Enterprise, the version that connects to all of your company's tools. And now they've built Comet Enterprise, the browser specifically designed for companies to deploy at scale. That is a full stack. That is a search layer, an agent layer, a browser layer, and an enterprise deployment layer. All connected, [snorts] all built around the same core idea. AI that doesn't just answer questions, but actually does things inside real workflows. That's not an AI search tool anymore. That's an AI operating system for work. And the reason I keep bringing Open Claw into this conversation is because Open Claw represents the same ambition, the fundamental belief that AI's value isn't just in answering questions, it's in taking actions instead. It's in doing things. It's in running the workflow, not just describing it. The difference is that Open Claw is a developer tool. You have to build with it. It's technical. It's hard to set up. You have to use terminal to use it. Comet Enterprise is a deployed product. You just install it. But they're pointing at the same future, a future where AI isn't a chatbot you talk to, is an agent that works alongside you and sometimes instead of you inside the tools and systems you use every day. The future is not coming. It's here. It launched today. And let me bring this home. Three companies are in a full-scale race right now to own the AI layer of enterprise work. You've got Anthropic with Claude Co-work, which lets AI operate on your desktop and your files. You've got Microsoft with Copilot Co-work, which brings the same ideas the Microsoft 365 environment. And now you've got Perplexity with Comet Enterprise, which brings it into the browser. All three of these products are running on similar principles, principles that come directly from the world of AI agents, the open core world, the world where AI doesn't just help you think, it helps you act. The question for any company right now is not whether to use AI. That question is over. The companies that aren't using AI are already behind, and the gap is getting wider every single week. The question is which tools, which layer, which approach fits what you actually do every day. For some companies, the answer will be open core based custom automation. They have complex enough workflows, specific enough needs, technical enough teams to build exactly what they want and what they need. For a lot of other companies, the answer will be something like Comet Enterprise, ready to go, deployable to the whole team, controllable, auditable, built for the way enterprises actually work. And the thing they have in common is the underlying idea, AI that acts, AI that runs workflows, AI that does work, not just talks about the work. And I'll say this one more time cuz I think it's the most important thing I can tell you today. The people who understand this right now, who understand not just that AI is powerful, but how it actually works, what it can do, and where it fits, how to use it inside real workflows, those people are going to look back on this moment as the moment everything changed. Not because AI replaced everyone, but because AI gave some people a capability that others didn't have, and they used it. And the gap that opened up because of that was enormous. You're watching that gap open right now in real time with today's Perplexity launch. The only question is which side of it you want to be on. That's it for today. If it was useful, share it with someone who needs to hear it. Subscribe if you're not already. If you want to be serious and get serious about actually using this stuff inside your own business and workflows, come join us in the AI Profit Boot Camp. Link in the comments and description, or go to the AI Profit Boot Camp.com. We'll see you inside there. Let's see what we got inside the questions here. Danny says, "At first I thought you left the doc open by mistake." Brutal transparency. Everyone reads scripts, but rarely anyone just shares a screen and goes line by line. Thank you very much, sir. Julian is a pro. Thanks, man. And then Drew says, "Love your content, but I was wondering if you can share" I I assume you mean the links of the tools, right? Or the links of the prompts. Everything that I talk about is like inside the AI Profit Boot Camp. So, if you want to learn this stuff, um just go inside the AI Profit Boot Camp and you'll learn from there. And you can get all like the actionable stuff inside there. So, all the tools that we use, um the full AI avatar process that we use as well, we've got inside the AI Profit Boot Camp. So, feel free to check that out. Is this live or pre-recorded? It is live, but I'm glad that I record that well live that it seems real um or seems, you know, pre-recorded. That's pretty cool. I'll take that as a compliment. What do you think about the new update from Nvidia? Absolutely awesome. Yeah, there's so much good stuff at the GTC conference, right? Absolutely amazing. I I love the fact as well like they just take over the whole ecosystem. Like what they're doing is by far the smartest strategy I've seen anyone do right now. So, yeah, it's amazing. Um and fair play to them. You know, like they seem to be doing everything the right way. They seem to be like collaborating with everyone. Um they've done extremely well as a company. Like even for example, just making stuff open source, like awesome. Fair play to them. Um also like not just that they you know, for example, they created Nemacolin, but then also they've got like, you know, this the chips that still go in. They've got like autonomous cars and stuff. It's just they're building out a full ecosystem that's impossible for anyone else to replicate, I think, at this point. Yeah, that's right. So, for example, Nemacolin is designed to be enterprise grade. So, what it does is like sandboxes the agent, and then it's a lot more secure. Claude Dispatch, Claude's biggest AI update this week. Imagine you're sitting at your desk on a Tuesday morning. You've got a meeting in 20 minutes. Your inbox has 47 unread emails. You've got a report due by the end of the day. The spreadsheet you haven't touched. A presentation that needs to be built from scratch and three client files that need to be sorted and organized before your call this afternoon. That's a normal Tuesday for most people who are running a business, growing a brand, managing a team, or just trying to stay on top of their work. Now, imagine something different. Imagine you pick up your phone while you're still in bed at 7:00 a.m. You type four sentences into an app. You put your phone down. You make your coffee. You take a shower. Maybe you go for a walk. And by the time you sit down at your desk, the report is written. The spreadsheet is built. The presentation is done. The files are organized. The emails are triaged and drafted. You didn't do any of that. The AI did. Not just answering your questions, not just giving you suggestions, not just generating text you have to copy and paste somewhere else, actually doing your work on your computer, in your folders, creating real files that are sitting there waiting for you. This is what happened today. Anthropic just shipped something called Dispatch for Claude Co-work. I want to talk to you about why this is one of the most significant AI updates we've ever seen in 2026, and what it actually does, who it changes things for, and what it means for the way every single one of us works going forward. Let's get into it. First, let me tell you exactly what Claude Dispatch is because the name alone doesn't do it dis justice, right? Dispatch is a new feature inside Claude Co-work. It gives you one persistent, continuous conversation with Claude that runs on your computer. You can message it from your phone at any time. You can come back to finish work later. Read that again. You message it from your phone, and you come back to finish work later. This is not a chatbot. This is not a back-and-forth conversation where you get a reply and then have to do something with the reply. You assign Claude a task, go do something else, and then come back to the finished work. Claude runs on your computer with access to your local files, connected and plugins, and messages you um the result when it's done. That's the thing that most people are going to miss when they first hear this. It runs on your computer. Your files stay on your machine. You're not uploading anything to the cloud. You're not giving external server access to some private documents. Claude runs code in a sandbox on your machine. Your files stay local. You approve what Claude touches for acts. And that last part matters, too, right? You approve what Claude touches. This is not an AI that just runs wild through your whole computer. You tell it what folders it can access. You tell it what it can touch. You're in control. It just does the execution. Now, let me zoom out for a second because some of you might be hearing this and thinking, "Okay, cool. Another AI tool. I already use ChatGPT. I already use Claude. What's different here?" Here's what's different. And there are three types of AI tools in the world right now. The first is a conversation tool, right? You ask a question, you get an answer, you ask it to write something, it writes it. You copy and paste the output, you go use it somewhere else. That's ChatGPT. That's regular Claude in the browser. That's Gemini. They're useful, but they're advisers, not workers. The second type is a coding tool. So, Claude Code, GitHub Copilot, Cursor. These are built for developers. They can actually implement code, talk to your computer's file system, write software, but they require a terminal. They require a technical level of comfort that most people don't have. Most of the people watching this, most business owners, content creators, marketers, coaches, freelancers, they're not living in a command line. Claude Co-work sits in the middle. It's a desktop agent designed for non-programmers who want AI system with files, documents, and workflows. Think of it as having an AI co-worker who can actually do your busy work, not just advise you on it, not just generate text you have to manually implement, actually go into your folders, read your files, build deliverables, organize a mess, deposit the finished work into the folder you told it to use. Claude doesn't just give you text responses in the chat window, it writes directly into your folders. When you ask for a presentation, you get an actual PowerPoint file sitting in your directory. When you need some sort of model, an Excel spreadsheet appears. That's a distinction that most people don't fully understand. The output isn't a chat message you copy and paste, it's an actual file, a real PowerPoint, a real spreadsheet, a real Word document sitting in the folder you told it to put things in. Ready to open, ready to edit, ready to send. And now with Dispatch, you don't even need to be at your computer to start this process. Let me give you a really concrete picture of what this looks like in practice. You open up CoWork and Dispatch five tasks in roughly five minutes. Then you walk away, meditate, make breakfast, do a yoga workout, and return to find deliverables waiting. That's from a doctor at Duke University, an orthopedic surgeon, right, who's running this company. And he said he dispatches five tasks before 6:15 a.m. and comes back to finish work. Now, that guy has more on his plate than anyone, you know, all sorts of responsibilities, a company to build, academic public papers to publish, research to review. And he's using this tool to run five parallel workflows before breakfast. If that sounds extreme to you, let me bring it down to earth. Let's say you're a freelancer, a small business owner, or someone who's running a content business. Your morning might look like this. You wake up, you pick up your phone, you open up the Claude app, you type, "I need you to pull the data from the spreadsheet in my reports folder, write a summary of last week's numbers, and format it into a one-page PDF I can send to a client." You put your phone down, you make some coffee. By the time you're back, Claude has pulled data from your local spreadsheet, compiled a summary report, and delivered it. And if you want, it can also search Slack messages and email, then draft a briefing document, or build a formatted presentation built from files in your Google Drive. That's not a demo. That's what this tool actually does. And look, I want to be honest with you because I always do, right? This is still a research preview. That means it's early. Dispatch is currently slow, and it's about a 50/50 shot whether what you try will work. That's not good enough to rely on when you're away from your desk. So, I'm not going to stand here and tell you this is perfect. It's not, it's an early version. Anthropic themselves said as much, but here's what I'll tell you. Every transformative product that exists today started as a research preview. It was a 50/50 shot before, right? Gmail was invite-only for 2 years and didn't even work half the time. The iPhone didn't have a copy and paste in its first version. The first version of Uber was breaking constantly and was only available in San Francisco. The question is never, "Is it perfect today?" The question is, "What does the direction of travel tell us about where this is going?" And the direction of travel is unmistakable. CoWork launched 12 of January 2026 as a research preview for max subscribers. Four days later, pro users got access. A week after that, team and enterprise. Then plugins came out on January the 30th. Now, Windows support, all in under a month. That is the pace of iteration. That is how fast Anthropic is shipping on this product. And now, less than 70 days after launch, they've shipped Dispatch, mobile control of a desktop AI agent. That's not a company that's slowing down. This is a company that's sprinting. Now, here's the thing I keep saying, and I'll say it again because it matters. Most people are thinking about AI wrong. They're thinking about it as a search engine that's just gotten better. They're using it to answer questions, they're using it to generate text they manually implement. They're treating it like a smart version of Google. That's not what this is anymore. What Dispatch represents is the shift from AI as a tool you use to AI as a worker you manage. Think about what that actually means for your business. If you have a team of four people, and one of them is spending 6 hours a week on reporting organization, formatting, document creation, administrative work, what happens to your output when AI handles all of that? You don't fire that person, you free them, right? You free 6 hours of their week. You free that brain for the work that actually requires a human brain. That's the game. It's not about replacing humans, it's about what becomes possible when the drudgery is automated. You can dispatch multiple tasks simultaneously, have Claude organize your downloads, photo assets, draft email responses while it researches a topic, and each one runs independently. Parallel workflows running simultaneously whilst you do other things. That is not the same category of productivity tool as a better word processor or a faster research engine. This is a different kind of leverage. Now, let me put some real numbers on this so it doesn't feel abstract. The average knowledge worker spends 2 and 1/2 hours a day on email, another hour and a half on document creation, another 45 minutes organizing and searching for files. That's close to 5 hours a day on work that produces no direct value, admin work, not not output work, right? 5 hours a day, that's more than half a standard day that most people spend on things that an AI agent can do now. And I know somebody's watching and thinking, "Okay, but can it really do that? My situation is complicated, my files are a mess, my workflow is unique." Here's what I'll say to that. CoWork shines when your outputs and your inputs are incomplete or disorganized. If you have rough notes spread across text files, markdown documents, or Word drafts, Claude can synthesize them into a first-pass report. It identifies over overlapping ideas, organizes sections logically, and produces a readable draft you can refine further. It is specifically good at messy stuff. It is specifically designed for the scenario where your stuff is everywhere, and you need someone to make sense of it. That's not a flaw in how you work, that's a use case this was built for. Now, let me talk about Dispatch specifically for a second because I want to make sure you understand the specific upgrade that happened today. Before Dispatch, CoWork already existed, and CoWork was already powerful. You could open your desktop, give Claude a task, let it run, and come back to finish work. But here was the problem, you have to be at your computer to start it. You had to sit down, open the app, type in the task, and then step away. That meant the only time you could set work in motion was when you were already at your desk, which means you're already in work mode, which means the benefit of walking away was smaller because you'd already started your work day. What Dispatch does is break that constraint entirely. From your phone, you can hand Claude tasks that use everything on your desktop, including things you can't open on your phone. Your phone becomes the remote control for your entire AI-powered desktop setup. You're on the train, you remember you need a report built, you pull out your phone, open up Claude, type a message, boom shaka laka, it's done, right? By the time you get to the office, the report is ready. You're at the gym, you think of three things you need organized before your afternoon meeting. You message Claude from your phone whilst you're on the treadmill. By the time you've showered and dressed, it's done, right? Now, here's something interesting as well. So, Felix Riesberg, the engineer at Anthropic who shipped this, said it himself. "It feels magical to give Claude a mission on your computer and get occasional updates like creating reports from internal dashboards or finding a better seat on your next flight. Everything Claude can do on your computer files, browser tools is reachable wherever you are. Finding a better seat on your next flight, for example, or, you know, that one example jumped out at me because of how mundane it is. That's not a big dramatic business use case, that's just a thing personal assistants would do. A task that has no business eating up 5 minutes of your brain but has always required you to be sitting at a computer actively doing it. Now, Claude does it whilst you're still doing something else, and this is where I want to introduce a concept that I think is important for understanding where all of this is heading. The bottleneck in your business right now is almost certainly not ideas. You probably have loads of ideas, right? The bottleneck is execution time. The bottleneck is the hours in your day. The bottleneck is how many things you can physically touch and move forward in any given week. AI agents like Dispatch are systematically dismantling that bottleneck, not by making you faster, by multiplying what gets done whilst you're doing other things. This is a genuinely different kind of leverage than anything has ever existed before, right? In the history of work. Think about it like this. Before computers, the leverage available to a small business owner was basically just hiring more people. If you wanted more output, you needed more hands. The ratio of people to output was roughly linear. Then computers arrived, and a single bookkeeper could handle the work of 10 bookkeepers with ledgers. A single designer with Photoshop could produce what an entire art department used to produce. So, leverage improved, but it still required a human to be actively operating a computer. Then the internet arrived, right? And suddenly a single person with a website could reach millions of customers, sell products around the clock, automate transactions, and do things that used to require entire departments. But even then, someone had to be online, someone had to be actively managing the website, writing the emails, responding to customers, and creating the content. What AI agents like Dispatch represent is the next step in the progression. What Now, the computer can initiate work, implement work, and deliver finished output without a human operating it. >> [snorts] >> The human sets a mission, the AI implements, the human reviews and approves. That is not a marginal improvement in productivity, that's a structural change in what a solo operator or a small team can produce. And here's where it gets uncomfortable for a lot of people because this is not evenly distributed. The people who understand how to use these tools, who learn the workflows, who figure out how to give good instructions, who build systems around AI implementation, those people are going to pull away from the people who don't, not slightly, by a lot. I keep saying this, I've been saying it for a while now. The gap between people who use AI well and people who don't is not a productivity gap, it's a leverage gap, right? And it's growing every single week. When someone who uses AI agents effectively can complete in one morning what used to take 3 days, and they're compounding that advantage week after week, the distance between them and someone working without these tools becomes enormous fast. And here's the part that I want you to really sit with. The people who are going to lose ground are not like the lazy ones. They're not like the, you know, the the people who don't understand the stuff. They're not even the ones who refuse to learn things. They're the people who are so busy implementing the current workflows that they have time to upgrade it. They're heads down, working hard, producing solid output, and every day they do that without upgrading their tools. So, the gap gets a little bit wider. And that's not judgment. That's just physics, right? Which brings me to something I want to mention briefly because I think it's genuinely relevant to what we're talking about. Inside the AI Profit Room, we spend a lot of time specifically on this, right? The question of how do you build workflows that use AI leverage, not just AI assistant. But there's a difference. AI assistant is using ChatGPT to help you write an email. It's useful, saves a few minutes. But AI leverage is building a system where Claude automatically processes your inbox, drafts your responses, and organizes your files, um and builds your weekly reports, and delivers finished work to you every morning before you've even touched a keyboard. There's a completely different order of magnitude. And the people inside the AI Profit Room are figuring out how to build that second thing. So, if you want to go into that room and learn how to actually apply this stuff to your business, your career, um your future, the link is inside the comments and description, or you can just go to the airprofitroom.com to get access. We'll come back to more of what dispatch means for you practically, but first I want to talk about something that I think is even more interesting than the feature itself. What I want to talk about, um is the existence of this feature and how this tells us where AI is heading over the next 12 months. Because dispatch didn't appear out of nowhere. It's the latest step in a very clear progression that Anthropic has been implemented on, and understanding the progression helps you see where this ends up. Step one was Claude the chatbot, question and answer, smart, capable, and useful. Step uh two was Claude Code, right? Developers could now give Claude Code access to the computer and have it actually write and run code autonomously. This was a significant leap, but it was still aimed at technical users. Step three was Co-work. Anthropic took the same underlying technology that powers Claude Code, the agentic feature, um the ability to plan and implement multi-step tasks, and wraps it in an interface that didn't require technical expertise. The entire Co-work feature was developed in approximately 10 days as a research preview, responding to users who were already pushing Claude Code beyond its intending um and intended use cases for tasks like, for example, vacation planning, document organization, and email management. Think about that. Users were already bending a developer tool to handle their life admin. They were so hungry for this capability that they were using a terminal-based coding tool to plan vacations. Anthropic saw that, built Co-work in 10 days, and shipped it. Step four is dispatch, mobile control of the desktop agent, the ability to set AI in motion from anywhere, anytime, without being physically present at your computer. What's step five? I'll tell you what I think it is. Scheduled autonomous work. With the introduction of scheduled tasks, Claude can complete something for you automatically, something that isn't possible in regular chats outside of Co-work. And that's already there. You can already schedule recurring tasks in Co-work. You can tell Claude every Monday morning at 7:00 a.m., "Pull last week's data from my reports folder, compile it, and drop the summary in my briefing document." It's not theoretical. That exists today. The next step after that is AI that doesn't just respond when you trigger it, AI that monitors, updates, and acts on its own judgment within parameters you set. Imagine telling Claude, "If you see an urgent email from any of my top five clients, summarize it and send a text immediately." That's where this is going. We are moving from tools you use to work as you manage the systems you run. And the businesses that are building on top of these systems right now, while most people are still treating AI as a better search engine, those businesses are going to be operating at a completely different level of efficiency in 12 months. Now, let me talk proactively about this, and because I know some of you are listening to all of this and thinking, "Okay, I'm convinced this is important, but what do I actually do with it?" Let me break down um this by who you are, right? So, if you're a solopreneur or a freelancer, here's where dispatch immediately moves the needle for you. Your biggest problem is usually not capability, it's capacity, right? You're great at your work. You just don't have enough hours to do the work, do the admin, do the business development, do the marketing, and do the client management all at once. Dispatch starts eating the admin category. You can create daily news or work briefs, draft email replies, delete unwanted subscriptions, manage to-do lists via tools like Notion, generate reports and presentations, and organize messy files. All of that from your phone while sitting at your desk. The practical move right now is to identify your top three most time-consuming admin tasks, the things that eat hours but don't generate anything directly, and test whether Co-work can handle them. Start there. Don't try to automate everything at once. Pick the three simplest, most repetitive tasks you do every week. Test dispatch on those. If you're a business owner managing a team, here's where this matters. The question I'd be asking is, "What are my team members spending time on that an AI agent could implement instead?" If someone on your team is spending 2 hours a week building reports, that's 2 hours of human um work spent on robot work. That's not a knock on your team. That's just an inefficiency that now has a solution. One user built an email priority run skill that categorizes emails by work pillar, drafts responses, and produces a dashboard with copy-and-pasteable replies. That's a repeatable workflow. Save it once, run it every day. The skill does the same work every morning without anyone having to remember it. Think about how many of those workflows exist inside your business. The weekly summary, the monthly report, the new client onboarding document, the competitor monitoring, the content calendar, the follow-ups. Every single one of those is a candidate for an AI workflow. Now, with dispatch, your team can trigger those workflows from their phones while they're doing other things. If you're a content creator, this one is personal for me personally. Uh content creation is genuinely one of the best use cases for Co-work and dispatch right now. You're working across multiple platform formats, right? YouTube scripts, blog posts, email newsletters, social media content, podcasts. You have researched scattered across dozens of tabs and documents. You have notes that never become articles. You have ideas that just drop off in a notebook somewhere. If you have rough notes spread across text files, markdown documents, or Word drafts, Claude can synthesize all of them into a first-pass report. It identifies overlapping ideas, organizes sections logically, and produces a readable draft you can refine further. That right there is a content engine. Your job becomes creating, refining, and putting your voice on the work. The raw assembly, the first draft, the structural organization, Claude actually does that. You're still the creative director. You're not being replaced. You're being freed from the part of content creation that doesn't require your voice. The research, the outlining, the formatting, the cross-posting, the repurposing, Claude handles all of that. What's left for you is the thinking and the delivery. That's a great trade. If you're a manager or implementer, here's the lens I use, right? The question isn't, "Can my company use this?" The question is, "What happens if your competitors start using this before you do?" Think about two companies competing in the same market. Company A is running traditional workflows. Their people spend 5 hours a day on admin, reporting, document creation, and organization. Company B starts using Co-work and dispatch across her team. The admin hours drop to one. Their team is spending four more hours a day on work that actually moves the business forward. That difference compounds week after week. After 6 months, company A hasn't changed, but company B has effectively added the equivalent of two or three full-time employees' worth of productive output without hiring anyone. That's not a hypothetical. That's a basic math on what happens when you reduce the friction between ideas and implementation at scale. Now, here's something I want to address directly because I know it's sitting in the back of some of your minds. What about the people who work for you? What does this mean for them? This is the honest question that a lot of AI content creators dodge, and I'm not going to do that, right? The truth is complicated. For most knowledge workers in the near term, the biggest risk is not being replaced. The biggest risk is being managed by someone who knows how to use AI and they don't, right? Because a gap in output between a person using AI well and a person not using it at all is going to make it very hard for organizations to justify, you know, um paying them the same. That's the uncomfortable reality. The answer is not to be afraid of the tools. The answer is to become someone who knows how to use them. The people who thrive in this environment are not necessarily the most technically skilled. They're the ones who learn how to use AI directly, who understand what tasks to delegate, what instructions to give, how to quality check the output, when to trust the machine, when to override it. Those are learnable skills. They're not complicated. They don't require an engineering degree. They require curiosity and the willingness to spend a few hours a week actually using these tools. And that's the whole point. There's a version of 2026 where you are spending 5 hours a day on admin, your competitors are spending one, and you wonder why you're falling behind even though you're working so hard. And there's a version of 2026 where you've learned how to use these tools, you've built the workflows, you've freed up your capacity, and you're operating at a level of output that would have required a team three times your size 2 years ago. That second version doesn't happen by accident. It happens because you decided to invest the time to understand this stuff before it becomes a crisis. Now, let me bring this back to dispatch specifically on what I think the next few weeks will look like. This is still an early version, right? And you can expect more to come within the next few days. It's rolling now to Mac subscribers with Pro coming in the next few days. So, if you're on a Pro plan, you'll have access to this within a few days of when I'm recording this. When you get access, here's what I'd suggest. First thing, update your Claude desktop app. The dispatch option shows up in the co-work section after you update. You'll see it in the left panel. You pair your phone by scanning a QR code. Takes about 2 minutes. After updating Claude code uh the Claude Mac app, a new dispatch option appears in co-work that will prompt you to scan a QR code to pair the session with your iPhone or other mobile devices. Number two, start with one small task. Not like your most important workflow, something low stakes. Ask it to organize a folder. Ask it to pull data from a spreadsheet and write summary. Ask it to search your emails for messages you haven't replied to. Something simple where it gets slightly wrong. Doesn't cost you anything to try, right? Third, pay attention to what it can't do yet because limitations right now uh actually a preview of what's coming. The things that are hard today will be easy in 6 months. The things that are slightly clunky right now will be seamless by the end of the year. Understanding the edges of the tool help you understand where to watch for improvements. There are some real limitations I want to be up front right now. Your desktop must be active. Claude works on your desktop computer, right? So, if your computer is asleep or the Claude desktop app is closed, Claude can't work on tabs or tasks. So, if you want to message Claude from your phone at 2:00 a.m. and have to work on something important overnight, your computer has to be awake and the app has to be open. There's a real constraint. It means you have to think about your setup in advance. There are no notifications when tasks complete right now. All messages live inside a single continuous thread. There's no way to start a new thread or manage multiple threads. That will change, but for now it's one conversation. You have to check back to see if it's done. And as I mentioned, it's still early. Some tasks will work perfectly, some won't. That's the deal with a research preview, but here's what I want you to hold on to. Every single capability that exists inside the final polished version of a product exists in first in a research preview that was clunky and I were unreliable and got mixed reviews, right? The people who got familiar with it early, who learned its quirks, who built their workflows around its capabilities before everyone else, those are the people who had a head start when the polished version arrived. We are at that moment with dispatch. The question isn't whether this technology is going to be reliable and powerful and everywhere in 12 months. It will be, right? The question is whether you're familiar with it when that moment arrives. So, whether you're scrambling to catch up. Let me tell you what I think the world looks like in 12 months if this trajectory continues. Right now, the default workflow for most knowledge workers is sit down at the computer, open apps, do work, close computer, and go home. In 12 months, I think the default for people who are paying attention starts to look like this. You set your AI's agenda for the day when you wake up. You review what it completed overnight. You approve, reject, or redirect. You spend the first 90 minutes of your day on the work that only you can do, the thinking, the relationships, the creative directions. And your AI handles everything else. That's not a fantasy. That is a direct extrapolation from what dispatch does today. The infrastructure for that workflow is being built right now, piece by piece. Microsoft has also seen what Anthropic is doing with co-work and announced a new co-pilot co-work product built on top of Anthropic's technology. Microsoft um actually fell in value like 14% since Anthropic debuted in co-working in mid-January. Let that land for a moment. Um this is one of the largest companies in the world responding to an AI agent product from a company that is a fraction of its size by building their own version of it because they understand what's happening, right? The desktop, the file system, the work computer, those are not small categories. Those are the center of the knowledge economy. And whoever controls the AI layer on top of that infrastructure is going to have enormous influence over how work gets done. Anthropic planted a flag there in January. Dispatch is um deepening that position. And it's why I keep coming back to the same message week after week. The people who are building fluency with these tools right now, not just awareness, actual fluency, are positioning themselves for a completely different level of productivity, output, and leverage than the people who are watching from the sideline. That's not hype. That's what is actually happening. Anthropic shipped a major AI product update in January. They iterated every single week. They added Windows support. They added scheduled tasks. They added a plugin marketplace. And together and today they added dispatch, the ability to run your AI agent from your phone. That is 70 days of shipping, right? Think about what 70 days of this pace looks like in 1 year. I'll leave you with this. Every major productivity shift in history, electricity in factories, computers in offices, the internet in businesses, was met first with confusion, then skepticism, then cautious adoption, then a mad scramble by the people who waited too long. The people who won in each of those transitions were not necessarily the smartest. They were the ones who got curious early, who learned the tools before the tools were polished, who built the habits and the workflows whilst everyone else was still debating whether it was real. This is that moment. Dispatch is a feature, but what it represents is a direction. And the direction is a world where AI is not a tool where you pick up when you need it, but a worker you manage around the clock. Where your phone is the interface to an AI that's always on, always capable, and always implementing. Where the limiting factor in your output is not how many hours you can work, but how good you are at giving instructions and reviewing results. That world is being built right now. Not in a lab somewhere, in a product you can download today. The question is simple, are you going to be ready for it? If the answer is yes and you want to actually learn how to build these AI workflows, how to use tools like co-work and dispatch properly, how to apply AI at leverage to your specific business or career, come join us in the AI Profit Boiler Room. Link in comments and description or just go to the AI Profit Boiler Room.com. We are doing the work in there, right? Real workflows, real results, real applications for real businesses. That's where the work gets done. All right, that's today's episode. And if you found this useful, share it with someone who needs to hear it because the biggest reason people fall behind on this stuff is not that they're not smart, it's that no one explained to them how it works in plain English. Be the person who passes it along and I'll see you on the next one. Let's see what we got on the comments. >> [sighs] >> Ah, here we got we got a good one by Xavier. Oh, we got an awesome one from uh Sabao. You're the goat. Thank you very much, sir. Xavier says, "I would like to know how to create separate sessions I can switch on between projects so I don't get confused." Just use Discord. If you check out my tutorials in the AI Profit Boiler Room on how to use um Discord with Open Claude, you can set up separate conversations. That's what I do so that the context doesn't get mixed. And that way it's way more chill. Is it after? No, this is me. Isn't the AI game getting overwhelming? Yeah, that's um that's part of the challenge, right? That's why I do these daily breakdowns and just show you what works and what doesn't. Drew said he's off. So, thanks for being around. Appreciate it. >> New opening eye GPT-5.4 and Nano. Mini and Nano. So, let me tell you what just happened. Open eye just dropped two new AI models today. On the surface, you might scroll past it, another model release, another set benchmarks, another thing to pretend we understand. But, I want to stop you right there because this one is different and the reason is different. It's nothing to do with raw power. It has everything to do with speed and scale. And that's what happens when AI stops being slow. Here's what I keep saying. The big models [clears throat] get the headlines. The small models change the world. GPT-5.4 Mini and GPT-5.4 Nano dropped today, March 17th, 2026. Mini is live in ChatGPT and Code X in the API right now. Nano is live in the API and the numbers behind these two models tells a story that most people going to completely miss. Let me give you the headline number first. Someone ran an experiment. They used GPT-5.4 Nano to describe 76,000 photos. 76,000. And the whole thing ran fast, reliable, and at scale that would have been completely impractical with the older models. That is not a typo. That is a new reality. I want to sit with you for that second thinking about that before we go any further because that number is a whole story. That is your unlock. When AI gets efficient enough to process 76,000 photos in a workflow without breaking a sweat, something fundamental has shifted. This is not incremental improvement. This is capability for collapse, right? And let me give you the context because context is everything, right? GPT-5 came out August 2025. And to be honest, it was a mess, right? The expectations were through the roof. Sam Altman has been hyping it for years. People are expecting something close to AGI, artificial intelligence, basically human level thinking in a computer. And what they got back um could not count the number of bees in blueberry, right? People were furious. Reddit went crazy. A petition with 3,000 signatures forced Open eye to bring back an older model. And one developer told Wired that GPT-5 felt like something that would have been released a year ago. An NBA executive literally DM'd a journalist to say it failed his two favorite test problems. The word that kept coming back and coming up was over and over was underwhelming, right? And I say all of this not to pile on because I say it because you need that backdrop to understand what's actually happening with today's release, right? Open eye has been playing catch-up since August, GPT-5.1, 5.2, 5.3, 5.4. They've been shipping model after model in a way that the industry has never seen before. And with each one, the story shifted. Less about the flagship, more about the system around it, more about the speed, more about what happens when you stop using one model for everything and start building teams of models that work together. And that is the real story of today, right? Open eye is not just releasing two new models. They're releasing the worker layer for an entirely new way of using the AI. Think of it like a company. Every company has a CEO, right? Someone sets strategy, makes a big calls, decides where things are going. Then every company has workers, people implement the plan, people who do the repetitive high volume, fast turn-around tasks that keep your machine running. You would not have your CEO process every single customer support ticket. You would not have your best engineer answer that, too, right? You use the right people for the right jobs. That is what GPT-5.4 Mini and Nano are for. They're the worker layer. They're the models you run a thousand times a minute so your expensive model only has to think once. The way Open eye describes it is a larger model like GPT-5.4 handles planning, coordination, and final judgment while smaller models handle narrower tasks in parallel, right? Subtasks with sub-agents like searching a codebase or, for example, reviewing a large file or processing supporting documents. Instead of one model doing everything, you build a system. The boss model decides, right? The worker model implements fast in parallel. This is what they call a sub-agent platform and it's where the entire AI industry is heading. Let me make this concrete because abstract explanations are useless, right? Imagine you're a software developer. You have a bug, you need to find it in a codebase that's tens of thousands of lines long. With a big model, you feed the whole thing in, wait 45 seconds, and get an answer. With the new system, a fast light model scans the codebase, takes a second, and then the more powerful model looks at the three relevant files it flagged and gives you the fix. Same outcome, but at a fraction of the time, ten times the speed. That is not a theoretical improvement. That is the workflow that Notion, GitHub Pilot, and dozens of other companies are building with these models right now. Instead of routing every single task through a large flagship model, you can now build systems where the big model plans and coordinates while smaller models handle the actual groundwork in parallel. Searching a codebase here, reading a document there, processing a form somewhere else. Matthew Berman, one of the most followed AI researchers on X, put it simply. He said, "Use GPT-5.4 Mini and Nano for most of your open-core use cases. It's a workhorse model." That's a high praise from someone who tests every model that comes out. Workhorse, not flashy, not hype, just reliable, fast, and capable. And that is actually more capable and valuable than flashy, right? I want to go deeper on the benchmarks because I know someone watching this is going to say, "What can I actually do?" And that's a fair question. So, let's look at the numbers. On SWE-Bench Pro, a test that measures a model's ability to fix real GitHub issues, GPT-5.4 Mini hits 54.4% compared to 45.7% for old GPT-5 Mini and 57.7% for the full GPT-5.4. That is a massive jump. The Mini model is almost catching up to the flagship on real software code engineering tasks. Not toy problems, not trick questions, real GitHub issues, real code, real fixes. On OSWorld Verified, which tests how well a model can actually operate a desktop computer by reading screenshots, Mini hit 72.1%. The flagship scored 75%. The human baseline is 72.4%. Read that again. The human baseline on operating a 72.4 is 4%, right? GPT-5.4 Mini, a small fast model, scored 72% on operating a computer. It is matching human performance on computer use. That is a different category of capability entirely. That is AI that can look at your screen, figure out what's happening, and take action. That is AI that can do the thing a junior employee does when you say, "Just go through those files and send the relevant ones to the filing system," right? This is real work. This is real output produced by a machine. GPT-5.4 Nano scores 52.4% on SWE-Bench Pro and 59% on OSWorld. Lower than Mini, but still a massive leap over previous Nano class models. These are not minor improvements. These are generational jumps in what a small, fast, lightweight model can do. Now, let me talk about the speed piece because I think it's underrated. GPT-5.4 Mini is more than two times faster than GPT-5 Mini. Two times faster on a model that was already fast. Why does this matter? Because in agentic workflows, speed compounds. When your AI is doing 20 tasks in a row, each one waiting on the previous one to finish, a two times speed improvement on each step turns a 40-step workflow that took five minutes into one that takes two and a half minutes. And in real-time applications, customer service bots, coding assistants, live dashboards, speed is not a nice experience. Coding assistants that need to feel responsive. Sub-agents that quickly compete and complete supporting tasks. Computer using agents and systems that capture and interpret screenshots. And multimodal applications that can reason over images in real time. Multimodal means it works with images, too, not just text. You can feed it a screenshot, it tells you what's on the screen. You can feed it a product photo and it writes a description. You can feed it a chart and it summarizes the data. Someone described 76,000 photos using GPT-5.4 Nano. An entire photo entire photo library described, categorized, ready to be searched fast, reliably, at scale. Think about what one that does for e-commerce, for example, for real estate, for stock photo libraries, for any business that has a mountain of visual content sitting unorganized and underutilized. The barrier to unlocking that data just dropped through the floor. Now, let me steel man the skeptic for a second. I can already hear it. Julian, Open eye has been disappointing for a year. GPT-5 was a joke. The community was furious. Why should I believe these smaller models are any different? Is this just more hype? And that's a fair challenge, and I want to address it directly. The disappointment with GPT-5 was mostly about expectations versus reality. People expected AGI, right? They got a slightly better chatbot. The qualitative feel was worse. Shorter answers, less personality, less creativity. The benchmarks looked mediocre next to the hype. By the end of the night of the GPT-5 launch, Open eye's street credibility had dramatically fallen. But, Mini and Nano are not trying to be GPT-5. They are not competing in the smartest AI race. They're competing in the most useful at scale race. And those are completely different competitions. Nobody was upset that GPT-5 was boring, right? Nobody posted a eulogy for the old Nana model because a benchmark for small model models has never been is this AI amazing? The benchmark is does this AI reliably do the thing that I need it to do fast, right? And on that benchmark, these models score extremely well. Perplexity deputy CTO Jerry Ma said after testing both GPT-5.4 marks a step forward for both mini and Nana models in our internal evaluations. Mini delivers strong reasoning while Nano is responsive and efficient conversation workflows. That is not hype. That's an engineer at a major AI company saying the models work. And Perplexity for context is one of the fastest AI products in the world. When their deputy CEO, uh sorry, CTO says a model is solid, that is a meaningful signal. Early adopters integrating these models today include Notion and GitHub Copilot, not random developers experimenting on weekends, real products with millions of users. They do not integrate broken models. They integrate models that work. I also want to address something that comes up with OpenAI every time it releases a faster lighter model. People say, "Well, what about Gemini? What about Anthropic? What about the competition?" And it's a fair question. And the honest answer is that OpenAI is not alone here. The market is ferociously competitive right now. Every major AI lab is racing to get their models into production workflows. And that competition is what drives capability up and friction down for everyone. What OpenAI has that others are still catching up to is the ecosystem. ChatGPT, Codex API, the developer tools, the integrations. When you choose GPT-5.4 mini, you are not just choosing a model. You are choosing a platform that already works with the tools you're already using. But here's my honest take. For most people watching this, these specific model model matters less than the habit of using AI in your workflow. The question is not which model is 3% better on the benchmark. The question is, are you building with AI yet or are you still watching other people do it? Because that gap between people who are using AI as infrastructure and people are still treating it as a novelty, that gap is getting wider every single week. Let me tell you what I see in the AI Profit Bootcamp, um in the automating with AI. The people who are winning are not the ones who wait for the perfect model. They're not the ones who spent 6 months researching every option. They are the ones who started building. They built something imperfect with last year's models. They iterated. They learned. And now they are running workflows that produce output that would have taken a full-time employee a week to deliver. The people who are losing are the ones who keep watching, waiting for the real time and the right time, the perfect time, waiting until AI is good enough, waiting until they fully understand it. I have news for you. The right time was last year. The second best time is today. With GPT-5.4 nano now available, the barrier to entry for building with AI has never been lower. You don't need a developer. You don't need to understand the internals of how these models work. You need a workflow idea and the willingness to experiment. Here's an example of what that looks like in practice, not theory, practice, right? So, uh for example, when I'm creating video content, I can easily just run a topic or I can even just ask the AI to come up with the topic, then create the script automatically. My team can create the AI after video, and then we have a system where we can automate content and we can push it out, right? And that can work across X, it could cross uh across Facebook, it could go across Instagram, YouTube, whatever, right? And so, that helps save hundreds of hours. Um what this means is that you don't need to become an engineer, you don't need to learn to code, you just need to understand the tools well enough to build a workflow that makes you better at what you do and helps you create more output, right? Because that way you can run a business that looks fundamentally different than it did 6 months ago. And that is what today's release enables us to scale. So, if you're a solopreneur, you can now automate the repetitive parts of your work using lightweight fast models, classification, sorting, first drafts, research summarization, formatting. Every one of those things that eats up 2 hours of your day and produces no creative value. Those things can now be handled quickly and automatically. If you're at a company, the case for AI augmented workflows just got stronger, not because the models got smarter, well, they did, but because the speed and reliability argument is now airtight. When you can process a million words in seconds, the ROI on AI integration is not a close call anymore. It's obvious. If you're a developer, the sub-agent pattern is now the standard pattern, not an experiment or advanced technique, the standard. When you build the next system, the assumption should be a big model for thinking, a small models for working. That's the architecture. Mini and Nano are the models that power the worker layer. In Codex specifically, GPT-5.4 mini uses only 30% of the GPT-5.4 quota, letting developers handle simpler coding tasks at a fraction of the resource load for the same work, same output, far less overhead. That is not optional optimization. That is just good engineering to 2026. And this is exactly what we work on inside the AI Profit Bootcamp. Real workflows, real systems, real implementation for real businesses, not theory, not benchmark comparisons, actual step-by-step frameworks and tutorials for taking what's being released, models like these, and turning them into efficiency and growth for your business. If you're serious about not falling behind, if you want to be the person who your industry actually knows how to use these tools rather than the one who vaguely heard about them, the link's in the comments and description, or you can join us at the AI Profit Bootcamp.com. We go deep on this sort of stuff every single week with new updates. Now, let me zoom out and talk about the bigger picture here because individual model releases are one thing, but the trajectory these releases point to is something else entirely. What OpenAI is building is a tiered intelligence system. At the top, you have the most powerful reasoning models, the ones you can tackle genuinely complex problems with, use sparingly for the tasks where you really need the best. Then you have mid-tier levels for general work, good at most things, reliable. And then at the bottom, and I mean that in the best possible way, you have models like Nano, fast, lightweight, built for specific tasks, running billions of times a day doing the jobs that no human wants to do. And that last tier is where the volume lives. This is the world and where it changes, right? When AI runs fast enough and light enough, every piece of software in the world becomes smarter. Your email client categorizes your messages. Your calendar negotiates your schedule. Your document software summarizes every meeting. Your project management tool updates itself based on what your team's actually doing. None of those things require genius AI. They require reliable, fast, lightweight AI. They require Nano. The context window on these models is 400,000 tokens. For reference, the entire text of the Harry Potter series is about 1 million tokens. These models can read and reason over the equivalent of almost half the Harry Potter series in a single pass. A year ago, that context window was a flagship model feature. Now it's the lightest model OpenAI makes. Think about what that means for document-heavy businesses. You know, health care, administration, consulting, any business that lives in spreadsheets, contracts, reports, um and memos. The ability to feed an entire document library into a fast, lightweight model and extract structured information, that's not science fiction, it's available today. Let me give you the trajectory of where this goes because I do not want to just report what happened today. I want you to understand what today means for the next 12 months. Models will keep getting faster. The capabilities that live in the mini and Nano tier today will keep improving until they rival what the flagship models can do now. We've seen this play out multiple times already. The flagship models of 2024 are the lightweight models of 2026. The pattern will continue. What that means for you is that every use case you are deferring because it feels like complicated to automate with AI, that use case becomes viable earlier than you think. The question is not is this AI good enough for my workflow. The question is, can I afford not to be learning at how to use AI when the cost of not knowing is going to compound against me? I've said it before and I'll say it again. The people who are learning AI right now are not ahead because AI is perfect right now. They're ahead because they are building the mental models, the workflows, and the systems that will be fully deployed by the time AI is obviously useful for everyone in the world. When GPT-5.4 nano dropped today, the developers who had spent the last 12 months building with AI knew exactly what to do with it. They had contacts. They had workflows. They had systems ready to plug a faster, lighter model into. They could act immediately. The people who have been watching from the sidelines, they're reading the headline, wondering even what tokens means. That's the gap right there. That's the gap that matters, not the gap in intelligence, the gap in preparation. Let me talk about the multimodal angle one more time because I do not think this is getting enough attention. GPT-5.4 mini is strong at multimodal tasks, particularly those related to computer use. The model can quickly interpret screenshots of dense user interfaces to complete computer use tasks with speed. Computer is a fast, lightweight model interpreting screenshots, taking actions. Let that sink in for a moment. This model does not just read text you feed it. It can look at your screen, any screen, CRM, spreadsheet, legacy software from 2003, a Google Docs file, and it can do things. Not just describe what it sees, actually operate the software. That's the demo the blue people waved OpenAI first showed computer use in their flagship. Now that capability is in a lightweight fast model that anyone can access. The economics of building software automation have permanently changed. Every manual data entry task, um every copy this system from A to point B workflow, every check this dashboard and send me a summary recurring task, every one of those can now be handled by a model that operates your computer, runs fast, and does not need days off. I'm not saying this to scare you. I'm saying this so you can be on the right side of this. The businesses that win in the next 2 years are not going to be the ones who hired more people for repetitive tasks. They're going to be the ones who built systems that handle those tasks automatically, reliably, and at scale. And then deployed their human talent on the creative, relational, and strategic work that actually moves the needle. This is not a threat to human potential, it's a realignment of it. Here's what I want you to take away from today. GPT-5 was a disappointment. I said so, the community said so, the data said so. But OpenAI did not disappear, they shipped fast, five models in less than a year. And what they've been quietly building underneath all the headline grabbing flagship releases is a model system, right? A tiered intelligence system, a set of tools that taken together are genuinely useful for building real things. Mini and Nano are not the hype story, they are the implementation story, they're the part where AI stops being something you follow on Twitter and starts being something you use to run your business. The context window is 400,000 tokens, the speed is more than double what it was 6 months ago. The compute use benchmark is matching human performance. And GPT-5.4 Mini is available in ChatGPT for free users today. Free users. The capability to run AI workflows that can operate software process hundreds of words, code, reason, and respond fast that is now available to people who don't pay, right? That is the world we're living in. And the question is, what are you going to do with it? Let me leave you with the concrete steps. If you're a complete beginner, someone who's never used AI in a workflow, start today. Open ChatGPT, find the most repetitive thing you do this week, the thing that bores you, the thing that you dread, feed it to Mini and see what happens. Don't overthink it. Don't wait until you understand everything, right? Just try one task today. If you're thinking about AI and using it for some task but not systematically, now is the time to start thinking about systems not prompts. A prompt is a one-time ask, a system is a repeatable workflow. Start mapping the tasks in your day that follow a pattern. The same input, same process, same output. Every one of those is a candidate for automation. And with Nano now available in the API, the barrier to putting those models into real workflows has never been lower. If you're a developer building with AI, the sub agent pattern is now the baseline. Every new system you build should have a clear separation between the planning layer and the implementation layer. The planning layer uses the best model you can access for the decisions that matter. The execution layer uses Mini or Nano for the volume work. Same output, far less overhead. That is not optional optimization, it's just good engineering in 2026. If you're a business owner or team leader, have the conversation with your team this week. Not should we use AI, that conversation is over a long time ago. The conversation now is which parts of your workflow are we automating first? Start with the highest volume, lowest creativity tasks. The ones where your people are doing something a machine could drop. Map them, then build them. And if you have been sitting on the fence about really learning this stuff, about going deep on AI workflows, AI automation, AI systems for business, let me just say this. The tools now available to everyone. The only remaining barrier is knowledge, understanding how to put the pieces together, knowing which model to use for which task, knowing how to build a workflow that actually runs reliably, knowing how to automate the right things and leave the right things to human judgment. That is what we do every single day inside the AI profitable in real people building real workflows, sharing what works, fixing what doesn't, and staying ahead of releases like this one so you're never caught flat-footed when something new drops. Because this is going to keep happening, right? GPT-5.4 Nano today, something else next week, something else the week after that. The releases are not slowing down, if anything they're speeding it up. And the people who are embedded in this world learning continuously, building actively, adjusting their systems as the tools improve, those people are building a compounding advantage that's very hard for late adopters to close. The gap is not closing, it's widening. And every week you spend watching from the sidelines is a week of compounding advantage you are handing to someone else. GPT-5.4 Nano, 76,000 photos processed in a single workflow, matching human performance on compute use, two times faster than what came before, available today for free. This is not the future, it's now. The only question is whether you're going to be someone who used it or somebody who talks about the people who did. I know which side I'm on and I'll see you on the next one. Thanks for watching, appreciate it. Let's see what we got here. Love from India, good to see you. Thank you. Robin, how can I contact you? If you want to contact me, feel free to get in touch via the AI profitable in. Is this live? Um, this is live, yes, my friend. All right, thanks for watching. I'll see you on the next one. Cheers, bye-bye.