Gemini 3 Rumors Are CONFIRMED, It's VERY GOOD

Well, the new Google Gemini model is here and I was lucky enough to be given early access and they told me, matt, don't you dare share this video with the world yet. Which is why I waited until the very moment they announced to put this video live. So in this video I'm going to deep dive into Gemini 3 and check out everything this new LLM is capable of, including their all new agent mode. I'll share with you the good, the bad and the visually unpleasant. Like yo mom. So let's dig in. As you've probably heard by now, Google released a brand new flagship state of the model called Gemini 3. A new deep Think mode that's basically extra brain power for hard problems and a Gemini agent that can actually do stuff for you like read your email, look at your calendar, browse the web and put everything together. Now the deep Think portion of it, they did announce it today, but that part is still coming soon. Gemini 3 and Gemini agent, well those are both available today. All right, so what is Gemini 3? Gemini 3 is Google's new top end thinking model. This is their next generation flagship, the one that's meant to sit at the very top of their lineup and compete with the best models from ever everyone else. So compared to Gemini 2.5, they're actually claiming some pretty big jumps in four areas. One, reasoning things like multi step logic, complex problem solving, planning, you know, stuff like that. Coding, writing and refactoring code. Understanding big code bases using tools, multimodal so handling text, images, charts, documents and long videos all in one go. And four, long context so staying coherent over bigger prompts, longer conversations and even longer videos. This is meant to be the model you reach for when you really want it to think through something and not just spit out a quick paragraph. They also shared some pretty wild benchmark numbers to back that up. Now I'm not usually a huge fan of benchmarks because they're not normally geared towards things that most general consumers would typically care about. But I do feel like they're worth sharing quickly, especially when we have a new state of the art leader. So on humanity's last exam, which is a benchmark that tests multi step reasoning and expert level humanities knowledge, Gemini 3 Pro scores 37.5% without tool use. And that's a very hard benchmark. And the number puts it ahead of their previous best. It also makes it the new leading model on the benchmark, with the second best model being GPT5 Pro, scoring a 31.64 on the GPQA diamond benchmark, which is basically PhD level science and reasoning questions. It hits 91.9%, making it the new leader there as well, beating out GPT 5.1 which scored an 87.6%. And pretty much all the other benchmarks tell the same story. Gemini 3 is the new top dog among large language models. In plain English on paper, this is not not a tiny incremental upgrade. It's a pretty serious jump in can it reason through hard stuff. At the moment we have a new state of the art winner according to the benchmarks, but we'll see how it holds up in tests in just a minute. Now as for availability, here's what they announced. They're launching Gemini 3 directly into Google search. So if you're in the US on the Google AI Pro or Google AI Ultra subscriptions, you'll see Gemini 3 powering the AI mode experience, but again, only if you're on one of their paid plans. It's also coming into the Gemini app itself. So if you're using Gemini on the web, you'll get Gemini 3 as one of the models behind the scenes. And Gemini 3 is also available to start building with inside of AI Studio and via the Gemini Command line interface or cli. And they specifically called out that this is for all tiers of users inside of AI Studio. So I believe that means anyone can play around with this new model for free inside of AI Studio. They also mentioned Google Anti Gravity, their new dev environment, coming in public preview and rolling out on Mac, Windows and Linux. That's more on the developer side. But the important bit is Gemini 3 is the model underneath a lot of that experience as well. So on day one it's going into search, the Gemini app, AI Studio and the cli, and then into developer tools like Anti Gravity. Now the other big announcements were Gemini3DeepThink and Gemini Agent. Gemini3Deep Think is an enhanced reasoning mode. Think of it like take a little bit longer, think a little bit deeper. Right now they're only giving Deep Think to safety testers and then it'll be rolling out to Google AI Ultra subscribers next in the coming weeks. But that's like the $250 a month plan. It's basically meant for the hardest problems. You know, long chain of thought, multi step reasoning, that sort of thing. And then there's also new Gemini Agent mode inside the Gemini web app, another feature I've had access to for a little while now. This lets Gemini not just answer questions, but actually scan recent emails, find relevant docs in drive, build slide decks, book things online for you. It's web only. Right now. It's still experimental and they repeatedly said you need to supervise it. Some sites block it, can occasionally move ahead without asking for confirmation, and you're responsible for what it does. So for right now, if you are going to use the agent mode, use it with caution. While it's still in its early phases, we're going to go ahead and play with that one on camera as well in a bit and see how far it can go before things start to get weird. All right, I think I covered all the announcements and things people would care about, so let's go play with the thing As a content creator, AI has really helped me with productivity and getting the small things out of the way. But it wasn't always that way. At a certain point, having too many different AI tools to juggle was becoming more work than just doing it manually. That's why I love what MAKE is building. It's basically a visual control room for all your AI tools where you can actually see how everything connects and works together instead of hopping between 10 different tabs and apps. I wanted to build an Audience Growth Agent to help me track my audience growth using real data, and here's how it works. A data agent pools analytics from my YouTube studio, a trend agent analyzes what content themes are performing best, a planning agent generates next week's content plan and emails it to me and my producer Dave every Monday and a reporting agent emails me every Friday with a summary of last week's wins and insights. This provides such real world value for my channel and because of Make Grid, I can actually see the entire process in motion and adjust anything when needed. And MAKE is continuously improving to meet you where you are. Recently at Waves 25 they introduced Maya by Make, a conversational builder that lets you create automations by just talking to it. They also rolled out new ultra adjustable AI agents and made Make Grid even easier to use with full analytics and control over your automations. So whether you're a content creator, a marketing agency, or building your own startup, this is how you actually operationalize AI. Not with a bunch of disjointed apps, but with an entire connected ecosystem system. Right now, MAKE is giving our community one month free of the Pro plan when you sign up using the link below. So if you're ready to stop juggling and start leading your AI powered workflow, check out the link in the description, grab your free month and start building smarter with make. Thanks to make.com for sponsoring this portion of today's video. Now Before I get into testing, Gemini 3 should be available inside of Google AI studio@aistudio.google.com if we come over here to our models, we should be able to select Gemini 3 from the model list. However, I am recording this before it's officially live and they gave me demo access inside of the Gemini web app. So I don't actually have access inside of Google AI Studio. So all of my demos will be in the Gemini web app. But again, it should be available here as well. So for my first test, I'm going to give it something that I think is going to be fairly easy for it, but it will require some logic and thinking through things. So you are my operations Planner. I have four videos I want to publish over the next 10 days. Video A sponsor must be alive by day four. Video B needs two days of editing after filming. Video C can't be published on a weekend. Video D depends on A being published first. Here's some constraints. I only film on Monday, Wednesday, Friday. I only want to publish max three videos per week. I need at least one buffer day before any sponsored video goes live. Create a detailed 10 day calendar with what I'm doing each day. When each video is filmed and published, a short justification for why this schedule satisfies all constraints. Then propose an alternative schedule that also works and explain the trade offs. And then we can see see that it actually gave us our Filming breakdown. Film November 17, Edit 18th, Film 19th, Publish video A, edit Video B, film Video C, edit Video B, et cetera. You can see it sort of broke it all down. And then here's the justification for why this works. And then it also provided an alternative schedule that I can follow if I wanted, along with the trade offs of the various schedules. This passed the test and it actually met all of the constraints that I provided here. So that's a win. All right, let's get a little bit more complicated with the logic here and, and throw it some puzzle math here. Solve this probability puzzle step by step. A game show has five doors. Behind one door is a car. Behind four doors are goats. You pick one door. The host, who knows where the car is, opens two different goat doors and offers you a choice. Stick with your original door. Switch to one of the remaining two closed doors. So here's what we're asking the model for. Compute the probability of winning the car if you always stay. Compute the probability if you always switch. But choose randomly between the two remaining doors. Generalize your reason for the case with endors where the host always opens N minus three goats and you choose between staying or switching to one of the two remaining doors. Show your work clearly. I actually think the showing your work part is the most interesting part to me. On this prompt we can see here, it does the math. Probability of winning if you stick 20%. We can see the math here. Probability of winning if you switch. It breaks down some scenarios, but then it does a pretty good job of visualizing the math that it used to figure this all out. And so I thought this was pretty cool to see. Like, like visually this is impressive to me the way it actually showed its work. All right, now things are getting really fun. Here's the next prompt that I gave it. Go to this archive paper and I gave it a URL, so we're making sure that it can go and grab information from other URLs. And then step one, summarize the core idea of the paper in 10 bullet points for a non technical audience. Step two, turn those 10 bullets into a two minute explainer script for a YouTube video. Step three, this is where it gets really fun. Design a simple animated visual sequence using HTML CSS SVG that could run alongside the script to explain attention to beginners, including the actual HTML CSS SVG code for the animation. So the paper in question that we're looking at is the famous attention is all you need paper. The paper that introduced the transformer architecture that all LLMs are built on today for the most part. So at first it's it, you know, did what any other LLM would do. Here's the summary script and visualization code for the paper. Attention is all you need. And you know, pretty much any large language model can do this just fine. It's summarizing a paper. And then step two, a two minute explainer script for YouTube. It wrote the script so we can see that it's doing these multi step things. Give a summary, write a script, then write code. Do step one, do step two, do step three all within the same prompt. So we've got a script here. Any LLM can also write us a script. But still impressive that it's kind of doing this step by step by step for us. But then step three is what I really got excited by. This is a self contained HTML file. You can save it as index.HTML and open it in any browser. I didn't want to copy and paste this to a text editor and then open it in any browser. So after it gave me this code here, I said let's open this code in Canvas and it said I've extracted the code for the self attention visualization into the canvas so you could run it and edit it. Let's see what it generated inside the canvas. It generated this visual for attention is all you need. That actually does a really good job of visualizing what the attention mechanism is. So with a sentence like the animal didn't cross the street because it was too tired. The big breakthrough of the transformer architecture was that it was able to take context from other words earlier in whatever sort of document you gave it. A traditional sort of like next word prediction may not necessarily understand what it is referring to because the word right before it was because. And the attention mechanism sort of made it so the model can have a better idea of the words that came before it. So it could understand more context around things like what it meant. So we can see these colorful lines going from it realizing that it was referring to animal. The animal didn't cross the street because it was too tired. Same with tired. What was tired it. Well, what was it? The animal. Awesome visualization. And I also love the color choices. And this really impressed me. I know it looks simple, but I thought it was really cool that I was able to take a PDF that was a very complicated explanation of what this is and make this visualization completely simplify the concept. Here's another prompt I tested. Here's a block of messy unstructured notes and some just random notes. AI models getting bigger every year, Energy usage increasing, New chips from Nvidia, AMD and Google. Open source versus closed source debate. Like not a lot of details or context, just some random like bullets. And then I say turn this into a clear three act narrative structure for a YouTube video. A storyboard with eight shots each described visually. A simple text based motion graphics sequence using HTML CSS keyframes to open the video. So again, testing multiple steps off of a single prompt. And what we see here, here's our three act structure. Act one, act two, act three. It pulled all of this with very little context, just some random bullets. And then it created an eight shot storyboard. Shots one through with a visual description and an audio voiceover cue. Feel free to pause the video if you want to actually check out what the storyboard did. And then my favorite part, because I love visuals. HTML CSS motion graphic intro. Once again it just wrote the HTML right in line with everything else. And I said, hey, let's open this in canvas. So let's see what it made inside of canvas. I actually haven't looked at this one yet. We're seeing it for the first time. All right, so it created this centralized, decentralized animation here. It's pretty good. I mean, I can actually, but it's got some cool HTML animation. Not quite as impressed with this one as I was with the last one. But again, this was the third step in like a multi step process. It wrote a 3x structure, it wrote a storyboard, and then it made this animation as the third step. But you know what? We haven't seen anything yet. Let's go even deeper and see what other cool stuff we can get this to do. Inside Canvas. I got smarter and started telling it to use the Canvas. Build a minimal Minecraft like Voxel world using only HTML, CSS and JavaScript. No external libraries. Requirements. A 2D or pseudo 3D grid of blocks that the player can move around in. Keyboard controls to move the player. Ability to place and remove blocks. A simple lighting or shading effect so it doesn't look completely flat. All code must be runnable as is in a browser. After you write the code, explain the architecture and how each file fits together. All right, so it thought through and then down here gave us an architecture explanation of what it actually built for us. And then if I open the canvas here, my God, this is what it generated. Let's see, did it meet our requirements? A 2D or pseudo 3D grid of blocks? Yes. Keyboard controls, well, they show right here. I haven't tested them yet. Ability to place and remove blocks. A simple lighting or shading effect, so it doesn't look completely flat. All code must be runnable as is in browser. All right, so we know that's working. It's running in the browser, so does it actually work? Let me click back into this. I can move around with the mouse, but my WASD keys are not doing anything. My spacebar is not doing anything. My shift isn't doing anything. Left click and right click are actually working, so I actually can add blocks and click to remove blocks so that functionality is actually working. But I. Oh wait, what the heck? Now moving is working. The first time I tried moving wasn't working. It's almost like I had to place a block and then as soon as I placed a block, I was able to move this one Shotted. This is a one shot right here using no external libraries. This is crazy. Oh my gosh. I am insanely impressed with this right now. Let me see. Can I go full screen? All right, so I'm not seeing options to go full screen with the Canvas anywhere. Keyboard controls to move player. Yep. That's working. It didn't work at first, I don't know why, but it is working now. And I can place and remove blocks. I can't really like change the style of blocks or anything, but you know, couple more prompts. We just vibe coded Minecraft. I. I do want to test one other thing here around the same Minecraft concept. In this one I said using only HTML, CSS and JavaScript and no external libraries. I'm curious what happens if I just remove this portion and just let it do it however it wants, like take away the constraints. So I gave the exact same prompt but just removed that. And now if we open this one in the canvas. Oh my gosh. All right, so this is what this one looks like. I'm actually starting like floating above the landscape instead of down on it. But if I go shift, I can actually go down. Do I go through? Yep, I actually go through the environment. Not a big deal. I mean it was a single prompt, so wasd keys, work space make takes me up, shift brings me down. Left click breaks blocks, right click adds blocks back in. And yep, this is so insane to me. All right, one through five, select color. So let's see, let me move to a flat area here. Let's do one block with color number one, one block with color number two, color number three or five. And yeah, we've got different color blocks here. This is pretty wild. So kids, if you're watching this and your parents won't buy you Minecraft, just make your own. Let's try another game. Create a simple turn based strategy game inspired by advanced wars using only HTML CSS and JavaScript. Requirements. 10 by 10 grid, two players, red and blue unit types. Infantry moves two tiles and Tank moves three tiles for a higher attack. Each unit has HP and attack damage. Click to select a unit, then click a valid tile to move or attack. Very simple AI for the opponent. On its turn, it should move each unit toward the nearest enemy and attack it if in range. No external libraries. And here's the many wars that it created. So I'm red, so it's red's turn. So let's go ahead and move some players. So it's telling me the slots that are available for me to move so I'll move everyone closer to the enemies. Go ahead and do this. Get all of my players up close to the enemy and then let's end our turn and let's see how its AI is working. All right, so blue AI is thinking and it's moving all of the players. Closer to my players. Let's go ahead and move closer again. See what we can do here. And then I'll end my turn. And then I think, let's see, if I select this one, I attack. I don't know how to attack. I don't know why it won't let me. Oh, it did. So it attacked that time. So let's do this. Tractor on tractor. I don't know why it's not letting me attack. It's kind of acting a little bit funky, but it's pretty dang close. I was able to attack this guy. Why can't I attack? Yeah, I don't understand. There's a few issues. Let me see here. I basically prompted how do I play this game to see if I'm doing something wrong. But it does say after moving, if an enemy is adjacent. Okay, so maybe I have to move first. All right. So if I click here, it should attack this guy. Yep. So now I attack that guy. Okay. I don't know what I'm doing here. So if I select this guy and I move him adjacent, now I can attack. I select this guy, move him adjacent. Now I can attack. And we'll move this guy here and attack there. All right. All right. Now it's clicking. Then I can end my turn. And they're going to attack my players. And it actually works. I just didn't know what I was doing. One of the things I've tested in past videos is getting it to generate one of my favorite games ever. Create a Vampire Survivors clone that runs in the browser in Canvas so we can see it did that. When I open this one up, no problem. We got a Vampire Survivors clone. I just leveled up. Let's do level up. Our magic wand. Let's do garlic. And then it adds this aura. Let's make a bigger aura. It's insanely fast. I mean, holy crap, look how fast everything's moving. But I mean, one shot. I just keep on leveling up. So a little wonky. I think the level up system needs to be slowed down a little bit. I'm already level 25. But I mean, again, one more prompt and we're probably there. Let's give it some feedback. This game is insanely fast. All NPCs and characters are moving too fast. And I'm leveling up way too quickly. Let's just see if it solves it for us. All right. I've rebalanced the game to provide more strategic pace. Let's go ahead and try this again. All right, now we're talking. Things are moving at a little bit more normal pace. Let's go collect our xp. It's giving us the same two options, but whatever. It did a pretty dang good job. Now I'm moving faster and faster because I'm drinking coffee. I mean, the speed escalates really, really quickly, but. And I died. Pretty close, though. If you've ever played Vampire Survivors, it's. It's close. I wanted to test if it can make music, so I gave it the prompt. Make the happy birthday song and then make it so I can play it in the browser. So let's see. Enter a name and press play. Let's go. Gemini is the name.

Happy birthday, dear Gemini.

I don't know why it would ask for the name. It's not like singing it out or anything. But all I did was make the happy birthday song and then make it so I can play it in the browser. And it actually generated all of these animations in the background with the balloons and the confetti and the page and the sort of sing along karaoke. It did all of that on its own without me prompting it to. It just sort of inferred that, hey, let's just make this cool for the user. And then my curiosity took me to what would happen if I asked it to generate a unique song. So I said create a completely original song and then make it so I can play it in the browser and let's go ahead and initialize audio and let's play it.

It's even got lyrics to it. Stars are glitching in the sky. So I guess I'm supposed to, like, sing along with this.

And I didn't ask it to make this visualizer. I didn't ask it to put this color scheme or a play or a stop button or even give it lyrics. I gave it none of that. I gave it the prompt. Create a completely original song and then make it so I can play it in the browser. Since I can't send you an mp3 file directly, I've built a custom web player using Tone JS. The single file acts as a synthesizer, sequencer and visualizer all in one. It generates the audio in real time within your browser using code, so you get to hear the song exactly as I composed it. Yeah, I mean, I'm actually pretty blown away by what Gemini 3 is doing, and I've only scratched the surface. And I know we're already getting a little bit longer on this video, but we haven't even tried the agent mode yet. So let's go test a couple things in there. So inside of our Gemini here for agent mode I need to make sure it's on thinking and under tools I need to select agent. Now this is a Labs feature so it may not be turned on by default. But if you do have one of the paid plans you should be able to turn this on inside of Google Labs. Start with a pretty self serving prompt here. Research three of the most important AI announcements from the last seven days. For each give me a three sentence summary link to the primary source, one potential downside or criticism. Then turn this into a ten bullet show rundown for a YouTube news video and a short intro and outro script for the video. Let's go ahead and run this with our agent mode on we can see it's processing the task information. I believe it's basically building out like a step by step checklist that it's going to then go and follow. It's identifying verifying AI news so it's doing its own Google searches on the latest AI news, analyzing product releases, verifying funding announcements, confirming release data, verifying GPT 5.1's existence, assessing competing announcements, finalizing announcements, prioritization. Now right now this isn't actually spinning up a sort of cloud based browser and running in the browser it's able to do all these agentic tasks directly inside of Gemini, but it will spin up a browser with the right prompts. It's now doing its deep dive analysis and we can see it opened up our camp canvas here. Top three AI announcements November 10th through the 17th so OpenAI launched GPT 5.1 Future House unveils Cosmos AI scientist Google Notebook LM ads deep research agent and then here's our YouTube show rundown. Our cold open intro story one GPT 5.1 critique users complaining about safety refusals Cosmos Hallucinated Stats issue Notebook LM some honorable mentions Nvidia Blackwell AI Climate Institute and that gives me an intro script and outro script. And it did that all with just a single prompt. It was able to do all of that. All right, let's test it as an actual assistant. Act as my executive assistant. Look at my upcoming calendar events for the next three days. Check recent emails for anything that looks like a task or deadline. Combine that into a single control panel for me with today's top three priorities, Time block suggestions and a list of things I should probably say no to or delegate. Presented as a simple dashboard I could copy into a Doc or not Notes app. Now there is one problem with this prompt that's a me problem. You probably won't run into this. The Google account that I'm using here isn't my normal everyday Google account. It's not the Google account that my normal calendar's attached to or my normal Gmail is attached to. This is my Google account that, well, Google hooks me up with access to stuff, but it's not the actual Google account with my Gmail and calendar. So I actually don't know what it's going to find. But let's see. All right, so it did this task for me. There's probably quite a bit I I need to blur on this page unfortunately. But as I expected, it says my schedule is clear, no calendar events found. So we can see. It gave me today's top three priorities. Attend, slash, verify this briefing meeting that I had activate. I have a early access to a specific tool that I can't talk about so I have it blurred and then plan my weekly schedule review. It does say my calendar is currently open. It's not. It's just looking at the wrong calendar. And then when I get down here, time blocking suggestions for the next three days. Since your schedule is clear, here's a proposed high impact impact structure. So Monday. So it broke down what I could be doing with the information above but again it didn't actually look at my main email or calendar. So it's not super helpful. But it would be helpful to you probably. And I want to try to get it to fire up its browser because certain tasks it will actually open a cloud based browser and browse on your behalf. So let's try this one. Book a dinner reservation for two people for this Friday night around 7:30pm Find a well reviewed Italian restaurant in San Francisco with available outdoor seating. Let's go ahead and submit that. All right, so it's doing some research for us. All right now it does say we're going to leverage the computer use agent. Aha. See now it's popping over open a browser. So it says using a browser now. It's not actually using my web browser. This is a browser in the cloud. You can see hands up. It just went to open table for me. I'm not touching anything. It is actually doing all the browsing for me. This is my mouse over here. So the mouse you're seeing inside the browser is the AI mouse doing its thing. You could see it is looking at the calendar for me. It searched November 21, 7pm Two people searching San Francisco and it found a handful of restaurants. I think it accidentally clicked back for Some reason. All right, found those restaurants again. It's doing a search for a specific restaurant that it found in its earlier research called social or Social. We could see the booking times down in the bottom right of the browser that it's running for us. Available seating options, standard outdoor. Let's see if it select outdoor for us. And if we look on the left side over here, we can actually see its thought process. I've decided to return to the San Francisco search results page to begin to search for Roma Entica. I'll use the breadcrumb navigation to get back to the list restaurants. I'm ready to input the new restaurant name and check for availability. Back on the main search, input the restaurant name here. Found it. You get the idea. It can actually go and browse on your behalf. Let me just fast forward to the end result. And it took a little bit of time, but it eventually got there. But what's really cool is I can actually click through the browser here and see as it, it clicks step by step every action it took within the browser. This is actually pretty cool. I don't have to watch a whole video. I can just look at step by step screenshots of it finally getting there. Now, the browser did pause cause I wasn't logged into this resi system where it made the reservations. So it paused, asked me to take control so I could log in and finish the booking. But it got me all the way up to the point where I can finally book it. And that's the Gemini 3 and the Gemini agent announcement that they made today. Today. And man, it is impressive. Like, I was genuinely blown away when I was using Gemini 3, especially when I was asking it to make graphics and visual videos and games that you can actually jump into and play all with a single prompt. Now, if you compare where we are right now to where we were one year ago or two years ago, like, it is insane. Like, this is blowing my mind how good this stuff is getting. And out of all the model providers that are out there, in my opinion, Google's kind of one of the most generous because they give us access to AI Studio, where we can kind of test and play with all these models for free, even if we don't have a Pro or Ultra account. And now they have the best model on the market according to all the benchmarks and even according to all of my tests. These are some of the best results I've ever seen. So Gemini 3 is really, really impressive. It is available, it's out now. I need to thank Google for the early access to it and letting me play with it like a week and in advance. I feel like I'm just scratching the surface of what this model is capable of. I'm going to test it with the API on my own a little bit and see what else I can do and really push the boundaries. But I think we're going to be hearing a lot about Gemini 3 over the coming weeks as people learn how actually really freaking impressive this model is. It is really cool, but that's what I got for you today. I wanted to break down all the announcements and really put this model to to the test. I'm sure some of my other buddies like Matt Berman and Matt Vid Pro and the other guys that make AI videos are going to do some other testing and go even deeper on it and try some more complex props. So make sure you check out those guys videos as well. But yeah, I'm impressed and I'm assuming you probably are too. Thank you so much for hanging out with me. Nerding out with me. If you want to stay looped in on all of the latest AI news and tools and tutorials on how to actually implement implement this stuff. Make sure you like this video and subscribe to this channel and I will make sure that you stay clued in and looped in with all of the latest advancements in AI. I am keeping up with it every single day so I can make videos like this for you again. I really, really appreciate you. Hopefully I'll see you in the next video. Bye bye. Like your mama? I don't know. Thank you so much for nerding out with me today. If you like videos like this, make sure to give it a thumbs up and subscribe to this channel. I'll make sure more videos like this show up in your YouTube feed. And if you haven't already, check out FutureTools IE where I share all the coolest AI tools and all the latest AI news. And there's an awesome free newsletter. Thanks again. Really appreciate you. See you in the next one.