This Unbeliavable AI Controls Your Computer

Okay, we need to talk. And as you might be able to tell already, I'm visibly excited by what just came out because the new claw release, and I don't say this lightly, trust me, changes everything. Yes, I know, big claim. Let me back it up with examples. But it's not just about the new model, it's about the applications that use the new model because both of those got an upgrade. We have a brand new Sonnet 4. 5 model that is great. But the ability to use this browser use agent, that's what I want to show you today. It has the ability to remote control your browser just through here. So, let me give you an example instead of explaining and I'll do all of that, right? I'll explain to you what this exactly does, what is in this release in a second, and then I'll give you some practical examples and ideas on how to use this yourself. But the fact is, we've never seen a browser agent this efficient and this effective at the tasks that I actually care about. This thing is reliable. So, let me just show you. Okay, so I set up a command in here

which is like slash shorts ideas from YouTube comments. What this does is it opens my YouTube studio, navigates to the content tabs, and opens the newest video. Okay, so I just gave it this preset prompt. I'm not touching anything. It goes in here. It's going to navigate to the content tab here in a second. It's going to open up the latest comments, pull in all the comments, generate shorts ideas from that and store that inside of a Google sheet. None of this is done by me. And I ran this multiple times and it actually works. Look, it went to the newest YouTube video we uploaded. Now it's looking at the comments. You can see the entire documentation here on the bottom right. It's doing the thing. It's getting the comments. And ladies and gents, like the thing to understand here is we've seen many versions of applications like this. None of them actually worked. The whole chat agent thing is nice. It's a good idea. It just has many hiccups and they run it sort of like in their own computer, in their own browser, which is terrible because every few days it just logs you out of all of your own services. This thing doesn't just work more reliably, but it also runs in your very own browser where you're logged in. Now, obviously, that comes with security risks. And yes, I am on auto accept mode here, so this is kind of just like running on itself and I can like talk to you here while it does it. But my first impression of this and my second and third impression also is that it works unbelievably well. It works so well that it makes me want to retry every prompt that I've done with AI and see if I could do it inside of this little browser extension. Now, if you follow the channel last week, I told you about Gemini releasing their version of it. It's not even close to this. There has been versions of Enthropics computer use before this. It was like a 2-year-old using a computer versus a grown-up operating it right here. That's how I would sort of express it. You can see it filter it for the criteria that I gave it efficiently right now. Look, it went to the filter. It's doing all of this. You can kind of watch it work as I explain, but there's a lot here in this release. I'm going to break it down for you now, but ultimately I just want you to know that this extension paired with the new set 4. 5 model that just came out from Enthropic is unlike anything we've seen so far. And that's not to say that the competitors are not going to do similar moves in the near future, but as of now, this is so far ahead of all the competition. And we have like a whole set of test prompts that we already ran in here and ran in all the competitors and all the manus and the gen spark super agent and the chat GPT agent um using you know all the newest models none of it is even close and here's one more thing that you need to consider like if it hasn't fully landed in terms of potential if you're not really seeing why this could be so powerful while this works let me show you one feature here. Okay I'm going to open up a new tab and this is how it works. There's essentially this browser extension. And if I go to settings here and then I go to shortcuts, you can see I set up this shortcut right here that we're looking at right now. But the ultimate thing is hidden here. If you go to edit down here, you can schedule this damn thing. So I could have this shortcut that looks into my YouTube channel, goes onto the newest video, pulls all the comments, and then extracts ideas from them run every day at 4:00 a. m. And now you could imagine me setting this up for five different channels that I follow or like. And then it would look into their comment sections, collect those comments, put them into this Excel sheet as this will do in a second, um, Google sheet, I suppose, and it would just work on autopilot. And then you could imagine applying what I might be doing here for my content business for every single other activity out there. Matter of fact, I ran a little prompt showing me different things that this could do for me that I will also show to you here in a second. But first, let's break down this release. But I just want you to kind of like think about what this means because like ultimately this is the first time in human history where an AI can really use your computer in a way that you would be using your browser with everything set up, everything logged in, right? And then it's not just about that. It's that Claude can now use the browser and within the browser it can use other AI tools. So obviously here like I'm using AI to not just scan the comments but also write the shorts. I have a second command in here which is like validate. Just if I have any fact or GPT generates something or claw generates something and I want to validate it. Well, I could use this little prompt and triple validated with GP5 Pro. So, if I do something like that, let's say I'm just going to open up a new uh window. Yeah, I'm going to go in here and say /validate. Now, this is one that you would not want to schedule. This is kind of like use as you go. It presets the prompt right into there. And now I could give it the information, right? I'm going to make something up like the US economy grew by 5% last year. And then if I give it that, I'm going to say enter. It's going to go ahead and open up my other AI tool that I have a subscription to, which is chat GPT. It's going to navigate over to the pro model and it's going to triple verify this information. Okay, it's asking me if it should do it. Actually, this is a first. I set it to this mode where it just acts without asking. I would not recommend that though. Yeah, just to be clear, but I don't know. Somehow it asks for confirmation here. Anyway, I'm giving you the kind of like the raw functioning of this, right? We're not going to like edit this so it looks better than it is. I'm just telling you from the testing so far, it's unbelievably good. I think it's kind of hesitating to work because it's already at work in this tab. So, you'll have to sequence these things. But, as you can see here in this YouTube comment analysis, it's kind of like done. And now it's inputting. It's populating this Google sheet as we expect it to. Okay, so enough of the demos. I just really wanted to show you this up front that it actually works and with the sheet as you would expect. But what I'm going to do now is I'm going to give you an overview of what they have released and then we're going to circle back to this and show you a few more examples of how it built an automation for me, how it runs some of these different prompts in a way that wasn't possible before and more. But

before we do that, a quick word from the partner that made this video possible and that's Web Flow. So, one of my favorite things about AI is that sometimes it feels like you're skipping many of the steps that you previously needed. It reminds me of how some people speedrun video games. You get to the desired outcome, but you just get there way quicker. And along the way, you just get to skip a lot of steps. And today, I want to show you a tool that does exactly that. And you might have already heard of them because it's Web Flow, the sponsor of today's video. So, Web Flow is a web design tool that allows you to do everything that is possible with code without actually having to write it. And if you've heard the name before, because they're the industry standard for professional websites that aren't manually coded up. Just a little side note, a ton of top tier web design agencies actually only use Web Flow. And that's for good reason. I would call Web Flow an intermediate level tool as opposed to a lot of no code tools that really aim at the beginner segment of people that just don't have any previous experience going into the tool. It might not be as intuitive in the first hours, but then the ceiling is way higher and the potential for what you can make in Web Flow is crazy. My personal favorite, and I think this is also the reason why a lot of design agencies go with it, is their focus on animations. Some of these crazy websites where everything is just in motion and animated are made with Web Flow. A good animation on your website can take it from, oh, that's a cute website to something a professional designer would charge thousands for easily. And Web Flow just truly finds that sweet spot between making these animations easy to create while still giving you full control of what you want the elements to do. This tool is really what you want to use if you want to take the next step and go from beginner to something that is usable in a professional environment. So, if you're ready to step up your web design game, check out Web Flow today. You can use the link at the top of the description to get started for free. So, let's start

by looking at the release. Okay, on the surface, you might look at this and be like, "Okay, a new model, the benchmarks are a bit better, looks good. I'll give it a shot. " Um, focused at coders might be another thing that you're kind of thinking. And you would be fair to think that because on all the benchmarks, it crushes all the coding benchmarks. It's state-of-the-art. It beats out the GPT5 codeex models. Opus 4. 1 model, which was Aropic's best model. And I also have to add this video is not sponsored by Enthropic, but we are on the early access list and we had access to this over the weekend. So, I got to try the model and I was like, "Wow, this is it's a great model. It works well, right? " But what I did not expect was this clot for Chrome extension that really unlocks the potential because if you look at all the benchmarks, sure state-of-the-art at most of the things here. But the one I want you to pay attention to is this one computer use. Okay, that is its ability to remote control a computer. And that's what it's so exceptional at because it went from like 42 44% to 61. I'm not exactly sure what the GPT would score here, but let me tell you just practically from all the test prompts like the GP agent is kindergarten level compared to this. It just works better. And the biggest weakness even if you find use cases where it works like well, the biggest problem with GPT agent and chat GBT is that it uses like this virtual machine and it just logs you out of the accounts regularly. So I remember I even got used to like ordering groceries automatically like a standard basket like once a week. Um, but in week two it just lost my credentials and it kind of lost the cookies and it ordered the entire order to some completely different address in Lisbon and instead of like five bananas by piece, it got five bundles of bananas. So all of a sudden I ordered like 25 bananas to completely wrong address, paid for it all, never got it back. It just messed up. That's just an example for how it would work cuz it just loses the cookies and the login information regularly in your browser. it holds up way longer because like you use it on a daily basis and if it loses that you would even log back in plus it just works right because it's local. So that is a big advantage and these scores just make a huge difference. By the way this benchmark is different tasks that you would do in an everyday scenario on your computer on your browser. That's what the examples here are. So from successfully completing like 42% of them this jumped up to 61%. Okay. So that's what I want you to pay attention to here. Here you can see the different kind of like areas and it's just better across the field in finance, law, medicine and STEM fields. And basically it makes Opus 4. 1 a bit obsolete right now which is interesting. So if that is a question in your mind as in like okay should I be using Opus or Sonnet 4. 5? It's set 4. 5 right now. This is what I

want to really go into though. The computer use and they don't even make it that big of a deal here. They show two more things and I want to cover those first. So they uh launched a clawed agent SDK. It's basically a new way to access it through the API that has a lot of these things like memory. I think this is really interesting, but I really I genuinely need more experience with this to give you informed opinion on this because this is just something new that I haven't been able to explore. These browser agents I've been excited about since day one, but they never really worked and this is I really have a breakthrough feeling right now. But yeah, this cloud agent SDK just allows you to do things like integrate memory into API calls, which is like super interesting. So essentially there was like a graph down here that showed it was it a different tab here managing context on the cloud developer platform like essentially you can independently manage context like this example is great where it plays katan so you can like let the API play katan but then there's like a second context window that is only responsible for memory and as like the history of the game progresses that is not stored in the main model but it's stored in like this memory unit on the side as the game happens it fills the context window but then at a certain point when the context window gets too full and I hope this is not going over your head. I'm not like going too deep here. But as the context window kind of fills up with the different see that right there it went from 150k to 67k. It automatically throws out all the context it doesn't need along the way. So it doesn't need a 5 billion uh token context window. It just reviews itself regularly and throws out everything it doesn't need. This is unheard of in through APIs. Yeah. Cloud code had this. That's where they learned how to build this, how to use this, and now they're putting it into the API. So, I think this is really exciting for the SDK. I really need like more experience and to hear from more people to give you informed opinion. It's basically out. If you're a dev, you want to look into this. But what I really want to talk about today and what I really feel qualified to judge here is the browser agent, which is the big release here. Um, it's this right here, Claude and Chrome. It's only available on the max

plan and also only I think they said only a select thousand users have it for now. I am blessed enough to be on the early access plan. So I guess I'm one of the users and I've been trying all of these things and more. And again, as you can tell by the intro of this video, it just works well. Basically, what this is it's a Chrome extension that you install in here and then you can open it up. You could let it ask before acting or you could like let it act without asking. It's kind of the dangerous just do it mode, yolo mode as some people would call it. So that's the overview view. Yeah. And this might be a good time to kind of hop back into the examples I started in the beginning because maybe uh right here it should have five new ideas. And as you can see right there, yeah, it found new ideas. All five comments with their hooks and script outlines have been successfully saved to the Google sheet in the short seeds tab, row 7 to 11. So these are the brand new ones. Um, I guess if we were to really scrutinize this, we could say, "Okay, it might be missing a line here. " But also, we didn't let it work in peace. I was like doing stuff here. I remember like during the first run, like when I started clicking to the side, I actually like messed it up and I clicked here and all of a sudden pasted in here. So, it does like its piece. I'll tell you that. But what matters is the result here, right? And like, let's see. So, the date is correct here. It copied everything into one field. That would be a fault, right? Like, let's be clear, that's not perfect. But here the first line kind of worked. Maybe here I would need to work with the prompting definitely. This is like version two of the prompt, right? I would need to keep refining this up here, this prompt with the different steps. But let me tell you from this right here, here's the raw prompt or raw comment. Here's the hook. Why do AI benchmarks ignore everyday tasks? And then there's um kind of an idea on how I could present the short. And honestly, this to me automatically running at 4 in the morning is a very powerful thing. I understand that you might not be running a content business, but just think about the possibilities here. Okay, let's just like zoom out for a second. I want you to think about this. Every single prompt that you've ever done, there's sort of this like assumption that you kind of accept as a reality when you use chatbt for a bit. The assumption is like okay, it's wise, but it hallucinates sometimes and it knows a lot, but there is a knowledge cut off. then like you know there's remedies to all those things but then let's move on like there's also another assumption which is like it can do a lot it's very smart it has a lot of information but it cannot use the apps that I use on a daily basis it cannot use Gmail it cannot use my calendar it cannot use Google sheets or word documents and the solution to that this far has been one word that made you know big waves this year and another word which made big waves a you know questionable impact for many people's life. Those two words is one automation, right? That's the answer to like okay like LLMs are smart but they can't use the things that I actually do. So like let's build automations where the LLM kind of has access to these various things, right? Like right here, this one was actually built by the extension. I'll talk about this in a second. Yeah. So if I if my LLM cannot use Gmail, then I'll just go into here and uh in like NANE or make or Zapier and then I'll just build an automation where it has access to Gmail and then I can do that. The problem is these flows are like mostly deterministic. So they go from one point to the other and your problems are probably more fluid than that, right? On an everyday basis, you have new problems, right? That's the beautiful and also horrifying thing about life. like every day presents new troubles, new challenges and you need to come up with solutions. That's why people have jobs. That's why there's always something to worry about. But here with this extension, it allows you to kind of like be a bit more like water. Whereas automations, they're a bit more straightforward. Now, the second word that I wanted to talk about here that were was sort of a solution to this like AI is powerful, but my problems always vary, the word is agents. But the thing is like in many ways they haven't been able to deliver what the hype is about. And in some ways they've been incredibly useful, but it's just like you human expectations around that word of like who they're just going to like do stuff for you didn't really align with the reality of like what an agent looks like these days. A good agent for most people is like using the deep research function in here, right? That's agentic and it's incredible and it's very powerful. And also I would argue like GBT 5 Pro is extremely powerful and it's agentic, right? it like has the AI reasoning and tool use with like internet access and stuff that's like agentic and that's super powerful and also like cloud code and the CLIs if you're into that entire world are super powerful yet 9% of the people don't use claw code regularly and don't use like advanced automation workflows with like aentic elements in the middle but everyone can use a browser agent where they just click a button and whoop right there you have your little browser agent and you can just tell it to do what you want it to do and it uses your browser. You don't need to set anything up. You don't need a virtual machine. You don't need like a whole like environment where it like stores your credentials or whatever. No, it just uses the browser you're in. And I think that's really powerful especially when you've pried with a model that can actually get things done. And that's why I think this changes a lot. Okay, so this is why I think this really matters. And now I want to get into like more examples. Okay, so now I gave you like the overview of what this release is about. I gave you kind of like my thinking of why this changes things. And now I want you to think about all the prompts that you've run and how they would look differently and behave differently if they had access to everything on your computer. Because you know, let's take the most classic prompt of them all, like writing an email, right, to somebody. Usually writing an email is not just writing an email. It's like gathering a bunch of information that is somewhere, whether it's in your head, on your phone, in your WhatsApp, in your messages, in another email thread, in a Google sheet, in a Google doc, gathering that information, thinking through it, and then formalizing that and communicating it through the medium of email. That's what writing an email is. It's not about the writing. The writing is the arguably the easiest part. It's the thinking. It's the information retrieval that is hard. Well, what can this do? Well, you could go ahead and tell it to open up a specific Google sheet, take in all this information, or even better, you maybe have a Google doc where it kind of just takes that information. And this is exactly their demo where it goes into the Google doc. It takes in the information and it takes uses it as context to then write the email. So maybe if you and your team have a document that they're collaborating on with all the details on some project or some case or whatever you're working on, then Claude can dynamically go into that sheet, take all the content as context, and then use it to write project updates every day. And I'm going to show you step by step how this idea that I've been now talking about for the past few minutes materializes into something that you can do yourself and actually use. Okay? And I think that's what this channel really is about. So I'm just going to show you now. Okay? So, I have a brand new Google document here and we're going to go through this workflow even if it takes a while. I'm going to name it like I'm just going to call it AI advantage. Let's say it's a Vibe coding course. And then maybe it includes uh various links, some context and maybe it includes uh workflow. So, just for the sake of this tutorial, I'm going to take some of the context from the guide library in our community and I'm going to copy some stuff over into this context document. So, for example, let's take this tutorial on building agents with plain language. So, let's take this as context here. Okay, guide info. This is the context. This is the workflow. Right. And this is we enter like very interesting territory here because like I'm using a guide on what to Okay, let's set up this context document and name it actually this. Let's rename it. So build AI agents using plain language with the NAT agent. Okay, nice. Now all of that is in here in my Google doc, right? For links, the links are already in there. Okay, so this is just an example again like just imagine like whatever you're working on all the kind of like surrounding data is in here. So maybe some emails that you communicated or maybe some transcripts whatever like the context of the project might be are in here. Okay, nice. So I'm going to go ahead and I'm going to make sure this is shared. So I'm going to say share. I'm going to make this any accessible to anyone even though that's not necessary because like I'm working on my local browser, right? So I could actually keep this restricted. And now that I have this document, I'm gonna open up a new um claude um for Chrome preset and I'm going to create a little shortcut here. Okay, I'm going to say slash write um project update email AI agent. Okay, for this particular project then I would say okay write a new email and I'm would make this a workflow. Step two, write an email. Step one, open this document and read all of its content into the context. Okay. And then I just add the link in brackets. Step two, write a new email in Gmail to um I'll just say eigormyadvantage. com with a summary of what the document includes and what the first step is. Okay, so these two things. So we want just a summary on what the first step is. Now you could imagine if you want to set a automatic daily updates to your boss, you could just make sure most up-to-date information is in this document. And then in step two, you would formulate this email. And then in step three, I'm going to go ahead. So here's a brand new Google document update or like update archive. So like email and then date, maybe the other way around. Okay. And then let's finish setting this up. So here inside of the settings, I'm going to say step three. Inside of this Google sheet, I want you to track every email you sent. Um, fields, date, email or like not fields, what would this be? Columns. And then that is sort of it sounds good. So the starting point of this is this document and also I need to give it uh the link to this sheet. I hope this works. I'm doing this live with you and let's see if this works. So I'm just going to copy this link again. I'm can keep it restricted as this is in my browser. And I'm going to let this work. Link to the sheet right here. Okay. So it has everything. And as you can see like if this works once I'm going to test it now, but I could just enable this and schedule it to run every single day at I don't know 9:00 a. m. like I enter work and my boss would like or like whatever this email would receive an update from everything from within here. And then I would log the answers into here. Now, this took me a glorious what was this like 3 minutes to set up. Sure, I might have a bit of experience with this stuff, but like I'm also explaining this to you. And also, we've been talking about prompting for like 3 years now on this channel, okay? So, you better have some prompts at this point that you can kind of like rethink into, oh, the prompts can now access my browser and the stuff actually works. Now, how can I transform certain prompts to do stuff like this for me on schedule as schedule tasks? Okay, so we have this set up. Now, all I need to do is to test it is I would go into the agent here on the side and I need to move my camera so you can probably see this. But all I would to do need to do here on the bottom right is say slash and then uh write project updated date email AI agent. Yeah. And I'm going to set this to act without asking. So this nicely runs. And hey, if this fails, so be it. Again, this video is just here to show you what's possible. But as you can see, it opened the document. I'm going to keep my hands up here. It's doing all of this by itself. It opened the document. Now it's going to take in all the context. Good. I see the document is loading. Let me scroll down to read all the content in the document. Guys and ladies, like I cannot emphasize enough how surprised I am by how well this works. And I think one of the big reasons is because they took everything they learned from Claude Code and its success over the past months and implemented that into this new model and this agent. And I wonder if like the competitors are even going to be able to keep up. Well, to be fair, the codec CLI is also pretty damn good now, but like seriously, this is so impressive. In this moment in time, this just works. So, as you can see, it just scrolled through the entire thing and it's taking the screenshots and the content there. Perfect. I've now read through the entire document. Now, let me open Gmail in a new tab to compose an email to Eigor at myadvantage. com. Okay, so navigating to mail. google. com. Okay, so I think it's Gmail is now open. Let me close this notification popup first. So, I think it's open in a new tab. As you can see up here, I I'm even hesitant to like kind of like move my mouse cuz I don't want to like trip it up with that. But, as you can see, like over there, there's a little like Gmail tab that opened up next to this one. And we're going to double check all of this in a second, right? Pop-up is still there. So, like it dismisses the pop-up. One interesting thing that we found during the testing of this in the community, I just had an office hours and like 10 of us hung out for two hours and really kicked the tires on this thing. One interesting thing that we found like on all pop-ups it like disagrees to share privacy information on cookies and stuff. So Mike from the community noticed that as we were testing. Thanks Mike. And it sort of like I don't know maintains your privacy I guess. Also it denies some requests. So if you give it some like blatantly false information and you ask to verify it. It's denied that request which I thought was interesting too. Anyway, so it's doing this now, right? It's clicking on the subject field. I kind of wish we saw it work, right? Should I risk it? I guess I'll risk it. Let's risk it. Okay. So, let me just very maybe let's time it right. So, like so it doesn't click in the message body and compose the email. I hope I don't mess it up. I I don't want to like time this in the middle of a click. So, like let's it's writing right now clicking. I feel like this would be a bad time to click. But yeah, like while it does that and let me move over quickly. I'll tell you in a second what I think you should do with your existing chats and prompts like practically to get to take advantage of this. It's taking screen. Okay. No. Okay. What's going on? Is this the tab? Oh no. Did I mess it up? Let's see. Where is it? Did we lose it? Okay. Oh no. I know. I shouldn't have done it. Oh no, we're back. Okay. So, let's just let it work and then review it in the end. Okay. So, it's running here in the background. Let me give you a strategy for what to do here really because like I think this is I know this is like a lot. Yeah. new release and it opens so many new possibilities, especially with the scheduling and the browser use. But I think what you should think about is like you should take some of your favorite prompts and you should see if there's like apps that you use in your browser or like do this. Take some of your favorite prompts on one side, open up your browser history and look at what kind of apps you used over the past few days. Like what sites do you frequent? Is it Gmail? Is it cheats? Is it other stuff? And see if there's anything that AI could kind of like do for you automatically that you do manually that would help you. That would be a good starting point. That's a good starting point for automation. And I think this will work here too. Secondly, there's this trick that you can do where you set up a project inside of chat GPT. Give it all the context on something there and then you let this extension navigate into the project to do your task. So you don't need to give it all the context there. Matter of fact, all you need to do is make sure that project has the up-to-date information and then it can do things with AI for you. So you don't even need to prompt anymore. You just set it up once. You schedule out the tasks and then it does it regularly. Realize this is not exactly a beginner workflow, but this is why learning all these tools is so essential. It allows you to set up sort of a base, a knowledge base, a task base of things of context about your life. And then tools like this allow you to plug into that and to actually get things done. And breakthroughs like this will enable more and more of it. So now it's actively asking me if I want to send this email. So, it wanted the confirmation for that. I'll just say yes. And then let's see if it worked. And then if it did, I'll show you one more example here that I did earlier. And we can end the video on that. I really wanted to give you multiple examples of things. And I wanted to show you our full workflow, which I think we did here. And now we can judge it. Let's see if it actually worked. I can see the email is ready to send. I mean, and again, apologies. I wish we could kind of like follow along, but also like I want to give you the raw and authentic take on like how this performs in a browser cuz I realize most people won't have access to it yet. So right here I can see the mail is ready to send with all the content. Now let me click the send button. Excellent. The email has been sent successfully. I can see the message sent confirmation at the bottom of the screen. And now it's going to the Google sheet to fill in the information. Can I tab over and see it work? I wish I could. Let's just see it appear here. So it should be working in here. Claude is active in the tab group. It showed me down there. Yeah, the chat is like kind of tied to this window, but it should be working in here. So, yeah, if I want to open the chat, I need to tab over anyway. Okay, so you can see it like it disrespected my header there. Fair enough. It put in the date and the email. But technically, I didn't tell it that there's a header. I just tell it told it like, "Hey, put in the date and the email," which it did here. So, like this that's a problemic problem. That's actually good instructions following almost, I would argue. You know, sure, could it take a screenshot and kind of think about it critically before it did it? Yeah, but also like could I change the prompt to make this work also? Yeah. And I think that's what I asked it for. Yeah, I asked it to Let's review that. Inside of the Google sheet, I want you to track every email you sent. Columns are date and email link to the sheet. Okay, so I didn't mean email address, but email content, but again, I it needs to be more specific. I was kind of doing this on the fly. Um, yeah, I think it did this. Um, I see that I accidentally overwrote the header row. Ah, it's reflecting the data is now in the row one where the header should be. Let me fix this by inserting a new row at the top for headers. Okay, so it's fixing its mistake now. Nice. Look, it's like inserting a new row. And that's what I'm saying. It just there's no better way for me to express this than like it just feels better. Like sure, the benchmark scores clearly show that it is better, but like stuff like this, like look, it just fixes itself. And then you refine the prompt, you schedule this to go out every day at 9 in the morning and you have an automatic project update that you can send to team members. And all you need to worry about is not actually updating the team members, but making sure that like your Google doc that you work in is up to date. Great. Simple thing. Okay. So, I just wanted to finish here and then I'm going to show you one more example super quick automation related and two a prompt that I ran to show me what it can do for my business that you can kind of copy and use for yourself. Okay, there you go. It did that. It fixed itself. Let's see what is it thinking right now. Excellent. A new row has been inserted. And yeah, this is looking good. Meanwhile, I kind of like want to take a peek into my Gmail and see if this um actually went through Gmail. All tasks have been completed. So, see, read the Google doc. Done. Send email to Eigor at my advantage. Done. And then if I go into my sent emails right here, you can see it. Summary NA10 natural language AI agent guide. Higgore. So I asked for two things, right? Like a summary that is this right here. And then the first step, the first step involves downloading into two JSON files and accessing Google doc template. Perfect. It just perfect. Like sure, could I like refine it to enter the data and all the email content into the Google sheet? Sure. That's a prompt tweak. But it did what I wanted. Cool. Two last things that I promised you. First of all, what I did in here, I thought this was really cool. I asked it to do something simple. create an automation that automatically sends a email from my Google account whenever I press a button. Okay, so it created a web hook. So I could have been more specific. I didn't want that which acts as a button you can press via URL. Okay, I meant like a different trigger, but that's fine. And a Gmail note to send the email. And here's what really got me and sold me here because this is in my browser and because this like I was logged into NAND here, it is able to like open this like look at this. It has the OOTH set up already. So this is not like all the other workflows that create automation workflows and you import them and you have to connect everything up and like 80% of the work is left. This can do it and everything is connected up already. So you could even like use this to automate the creation of automation if you have that skill. Guys, this is what I keep telling you like all these AI skills they stack. Like you learn how to express your personal context just in order to understand what prompts would help you and to get more out of them. When you have that, you learn how to build little GPTs, learn how to populate projects. When you have that, you learn the very basics. Well, you can go into more advanced prompting or automation. And you use all of the first building blocks here like your context to automate things away. And when you have automations, you can start like building more automations with this. Or you can start and do what I'll show you here as the last thing in this um video. And I I'm just excited. I just can't help myself. But like the last thing that I want to show you here is that you can kind of go into your work context that I teach how to set up and you can kind of set this up, right? You can go in here and you can make sure that you can use a prompt like this and you can find different high ROI ideas that work for your company. Now, this only works because inside of the project, I have a big document which is like an overview of my company. I have my mission, vision, values. I have all the things that I've taught on this channel and in the community on how to set these things up for yourself. And when you have it, you go in here, you ask like, "Okay, there's this new feature. How could I use it? " And then it's like, "Yeah, number one, you could do a daily AI news radar or app of the week where it actually tests the app for you. " Matter of fact, I showed you these guides in the community. We had this idea of like we could actually now test all the guides regularly and then it could create a Excel sheet or like Google sheet with suggestions on what to update, what might be out of date, what didn't work, right? So we don't have to test the workflows ourself like sure we do it when we initially created it but like maybe once a month we could just let this extension do the QA testing. It could generate a bunch of prompts of the week and like evaluate the results. It could um reply to community questions for us, right? Like it could go into the community as it's logged in a man browser. Um it could test funnels front to back. All of this along with the prompt, very powerful prompt right here that you can kind of like run to see what this could do for yourself. Again, this will work only work if you have like your own context set up. And there

it is. This is really what I wanted to share with you today. Sonnet 4. 5 big kind of update across the board on capabilities, but ultimately the thing that I think matters most for most consumers is this extension. So check if you have access yet. It's only for the max plans, but they will be rolling this out over the next weeks and months. And yeah, leave a comment below with what you think you could use this for. And with that being said, I can't wait to like retry every prompt I've ever done and see if it could use my browser in an unexpected way to generate more value for me or save me time or just make my life more efficient and better cuz that's what these tools do. And that's what the progress in the space that is available to consumers like us makes possible. Okay, I hope this was interesting. I hope you enjoy these examples. I'm going to follow up with more. This is just kind of my first and second impression after testing this for a bit. And yeah, it's really exciting. You know, if you follow this channel, I don't say that often. So, I hope you have a wonderful day. My name is Eigor and I will see you very soon. Bye-bye.