Did We Just Get the Nano Banana of Video?

Well, it's been another exciting week in The World of AI from OpenAI's Code Red, Amazon's Re Invent event, some impressive new video models and garlic.

Anyway, let's go on a little journey and explore the fun, the interesting and the weird announcements that came out this week in AI, starting with what was likely the biggest news of the week, OpenAI's code red. Actually, I'm not even going to talk about that one here, because I made an entire breakdown video of exactly what's going on there and how OpenAI actually got to this point. So check out that video video for all the details on the whole OpenAI code red thing. And now for something completely different. What I'm really going to start with is all the crazy new AI video models that we got this week. And let's start with the Chinese AI model maker Cling. They didn't only make one announcement this week, they made five. They did five days of announcements. And the day I'm recording this, day five hasn't even dropped yet. So there's probably even another video model that they shared that didn't make this video. But on day one, they dropped clip cling AI01, which a lot of people are sort of describing as the nano Banana of video, where you can take a video and edit almost anything in the video. You can combine videos and images and text prompts to generate your video. Some really, really cool stuff. Day two, they announced cling image 01, which is pretty much the same thing as Nano Banana. It's like an image model where you can give it a bunch of images and combine them into single images. So, yeah, Nano Banana, basically. Then day three, we got cling AI video 2.6, which was a model that stands out because it now has sound along with the visual. So it's actual native audio. After the late night fire little girl hugged me and said she felt safe. Previously, CLING would generate audio for video. I guess it just wasn't natively done. So it generated the video, then generated the audio, then mixed them, I guess. I'm not 100% sure, but this new video 2.6 has native audio. So same kind of thing you get out of VO3 and VO3.1. And then on day four, which is actually the day I'm recording, they released an avatar feature called cling AI Avatar 2.0, which allows you to make these sort of realistic avatars and have them talking to the camera and promoting your products and stuff. So that's the breakdown of what they released. Let's Go test them. So let's start with cling01, the one that's supposedly like the nano banana of video. Here's how Cling described it. Brand new creative engine for endless possibilities. Input anything, understand everything, generate any vision. With True Multimodal Understanding, CLING01 unifies your input across text, images and videos, making creation faster, smarter, and more effortless. So cool. Let's jump into cling here, and I'm going to select cling01 up in the top left. And just to save me some time, I've actually already generated the videos, but I'll show you the exact inputs and prompts that I gave to generate them. So for my first test, I wanted to see if I can give it one texture and have it apply that texture to a different animation. And the textures sort of change the physics of the original animation. So I gave it this image of like a gooey, cheesy pizza. And I gave it this image of a luxury car. And I gave it the prompt. Animate image 1, which you can see is the car driving fast on a racetrack. The car is made entirely of the material from image 2, which is the pizza. As the car turns corners, the body should stretch and melt slightly due to heat and centrifugal force, dripping cheese onto the asphalt. And here's the video. Now, these ones don't have sound. This is 01. This is not the new model with sound, but here's what it came up with. We can see the car driving around the track. It's got the cheese pizza sort of texture to it, but the cheese isn't actually moving with the car very well. Maybe a little bit down on the bottom, but it kind of miss what I was going for. Peach of a hut. For the next one, I wanted to test giving motion from one video with a static image and see how well it applied the motion to the static image. So for the video, I gave it this video of a drone flying through a canyon here. This was generated with VO to start, and then I also gave it this image of like the internals of a watch. And I wanted to see if we could get a first person view of the drone flying inside the watch. I don't have high hopes, but let's see what happened. Here's my prompt. Bring the image to life. This is the image is the watch using the camera movement and speed of the video, which is the drone. We're flying deep inside the watch mechanism. The gears must rotate in sync with the speed of the camera movement. And here's what that generated.

It actually kept the sound effects from the original video. But I'm gonna go ahead and mute those so we see the watch spinning and turning here. But it didn't take anything from the drone other than the audio of the drone. So, yeah, not quite what I was hoping for. But we got a nice animated looking watch with drone sounds over it. So for the next one, I wanted to take like a real video and then give it like a sci fi element and see if it could implement the sci fi element into the real video. So I gave it this video here of me just sitting back and reading a book. And I gave it this glowing tablet image that I generated with AI and I gave it the prompt, replace the book in the video with the tablet. From the image, adjust the hand grip to hold the thin tablet naturally. The light from the tablet must illuminate the face and palms. And this is what we got out of that. So I actually think it did a pretty good job of putting the hands in the right spot on the tablet. But then it kind of loses it when I fold the book up. I also am not seeing like the glow of the tablet on my face, so it kind of missed that step. But again, the first like five seconds of that video are pretty impressive in my opinion. I'm happy with that. So next up, I wanted to test giving it a 9 by 16 video, which you can see here of this guy running on a treadmill and see if I can expand it and change the scenery of where that person was. So I gave it the prompt. Change the aspect ratio of the video to horizontal 16 9. Replace the gym environment with the surface of Mars. The person is now walking in a spacesuit. The treadmill is gone. They're walking on a dusty red terrain. And the expanded background show futuristic colony in the distance. And we can already tell from our preview here that it ignored the aspect ratio and just left it as the original aspect ratio. So apparently it can't do that, but this is what it gave us. We have our men running on Mars space station in the background. Don't really think it looks like the same guy, but it's like the same running motion. And it obviously didn't get our new aspect ratio, but it transferred the scene over pretty well. So, I mean, it kind of didn't do a great job with most of my prompts. When I compare this to the nano banana of video. Well, it wants to be. Is it as good as that description? I wouldn't really say so. I am fairly impressed with this one right Here where it took me holding a book and changed it to me holding a futuristic tablet like that actually worked fairly decent all the way up until I fold the tablet. Then it gets a little wonky. I also thought this one where where the guy was running was okay. It transferred the style pretty well, but it clearly is a different person in that helmet. And it also refused to change the aspect ratio. So I mean, not really. Nano Banana E I could see using it, but I also could see myself like needing to prompt it 10 times to finally get what I want out of it. But the technology is coming a long way, so I'm excited about the future and that we're getting this kind of stuff because it's only going to get better from here. But again, that was only one of the models they launched. They also launched video 2.6 which is their first model model with native audio. So I wanted to put this one to the test as well. So my first prompt was a close up low angle shot of a young cyberpunk street artist in a neon lit alleyway. They're wrapping an intense fast paced verse directly into the camera. The audio is a gritty fast tempo hip hop beat. Character raps clearly. Lights, flash, systems crash moving in a digital dash. The flow is aggressive and rhythmic visual action. The character gestures with their hands to emphasize the rhyming words flash, crash and dash on the exact beat. This is what it generated.

Light, lights, flash, systems crash. Moving in in a digital had light. Lights, flash, systems crash. Moving in in a digital had dash, light.

I mean it repeated some words so it didn't quite get that right. The fingers look a little bit wonky. Overall, it's pretty decent. I'm really, really nitpicking here, but it didn't get the words right. Lights, flash, systems crash, moving in a digital dash. It kind of went lights, lights, flash, flash and sort of repeated itself. I also don't think the lip sync is perfect.

Lights, flash, systems crash, moving in.

But this is one of the better lip syncs I've seen. If you watch my recent video on Perplexity's Comet, you already know that I've started using it as my own default web browser and many of you were interested in it. So I wanted to show a few more examples of how I've been using Comet recently. So I'm browsing for a Christmas gift for my wife and on this product page I'm going to ask Comet summarize the reviews and compare prices for the same product across the Internet. And in just a few seconds I can pick the best option without having to scroll through thousands of reviews. Now let's try this one. 1. Summarize this YouTube video and jump to the part where they talk about the camera specs. Some YouTube videos can get pretty dense. I would know. So I love this feature. And because Comet can actually see the page I'm on, it understands the context and does everything for me without switching tabs or copying and pasting URLs or anything like that. And there's so much more you can do, like translate and summarize articles and even help me draft email replies. And it's also free to use on both macOS and Windows computers. So if you want to upgrade your browser, download Comet through the link in the description and try it out for yourself. But fair warning, once you start using it, normal browsers are going to start to feel like dial up. And thanks so much to Perplexity for sponsoring this segment of today's video. So for this one I wanted to test kind of like ASMR type audio so continuously tracking shot of a hiker's boots. The camera follows at ground level. Visual action. The hiker takes three steps on crunchy dry autumn leaves, then steps onto a hollow ridden bridge, then onto solid wet pavement. For audio. The soundscape shifts distinctly with each surface. Loud, crisp crunch of dry leaves, followed by the deep, hollow thud of boots on wood and finally the sharp solid clack on wet pavement.

Walking on the leaves sounded pretty good. The rest it didn't really follow the instructions. So for this next one, a cinematic shot that starts inside a chaotic, loud nightclub with strobe lights flashing. The camera pushes rapidly through the crowd and exits through a heavy soundproof door into a silent snowy alleyway. Visual action. The visual chaos instantly turns to a still, peaceful night scene as the door swings shut. Audio thumping, muffled bass and crowd noise that is deafeningly loud, which abruptly cuts to near silence the moment the door closes, replaced only by the subtle sound of wind howling and distant city traffic. So I wanted to test sort of the change in sound here.

I mean, the video didn't really play long enough for us to know, cuz as soon as the door opens, the video kind of ends. But it also didn't really follow my instructions very well because the door that it opens up to is like a gate. All right, So I wanted to test people talking over each other. Medium shot of a couple having a tense dinner conversation in a quiet restaurant. The man is speaking calmly, but the woman suddenly raises her hand and speaks over him. The man stops talking mid sentence and looks shocked. For the audio, the man says, I really think we should consider the before being cut off, the woman interrupts loudly, stop. I don't want to hear it. The audio of the man must cut off abruptly as her voice takes over. I really think we should consider the stop. I don't want to hear it. It so not perfect. Obviously they both say stop at the same time. It's not just her saying stop and then she raises her hand when she's done talking instead of when she's telling him to stop. But it's definitely still got that like uncanny AI lip sync where you can just tell those voices aren't coming out of those mouths. And then finally, for my last test, a bright high definition macro shot of a homemade Rube Goldberg machine in motion. A metal marble rolls down a plastic track, hits a row of dominoes that fall rapidly. The last domino pushes attack into a blown up balloon, causing it to pop. So for audio, the rol rolling metallic whir of the marble, the rapid clack clack clack of the dominoes falling, and a sudden sharp pop sound that aligns perfectly with the visual explosion of the balloon. And here's what we got from this one.

Yeah, I don't know what the heck was up with that one. There was no sound as the ball rolled down. It only hit one domino when it smacked the balloon. Yeah, I don't know what that is. So are we at VO3, VO 3.1 or even Sora levels with this? No white. I'm still going to VEO if I want a video with audio in it. Maybe Sora, but probably vio. There isn't really another model that does that same kind of video editing that we saw earlier in their 01 model, but I feel like you could kind of get that using a combination of Nano, Banana and vio. So I don't know, I'd probably still go to those instead. But another model for you to try if you don't want to use, you know what Google has available for you. Also, if you prefer the platform of Leonardo, they are rolling out cling01 as well as the video 2.6 model. They'll both be available to use inside of Leonardo. So when I do use those models, I'll probably be using them in Leonardo. We also got a tease of another video model this week. So on the Same day that cling showed off 01 Runway showed off gen 4.5, which looks insane. Like it looks like the best video model out there. Probably better than VO 3.1 probably better than Sora.

Anything you can do, I can do better.

But we don't have access to it yet. I don't know if Runway announced this because Kling was announcing a bunch of other stuff and they wanted to also be in the mix this week, or if they had planned on announcing it and launching it later. But it does look really good from what I've seen. So according to these leaderboards, where a bunch of people pick what generated their favorite video, Runway gen 4.5 seemingly blowing everything out of the water. Now, this chart is a little bit misleading because you can see it starts at 1200. So none of these are really that far behind. But because they sort of started at 1200 and zoomed in, it made it look like Runway is just like blowing the competition away. But really it's not that big of a leap. But here's some examples of videos that it made. Like this complex scene where it just keeps on scrolling to the right. We've got like a sheep or a goat towing a polar bear, a cactus popping a balloon. Person. Expressive characters. Lots and lots of demo examples here. Photorealistic, non photorealistic, slice of life, cinematic. And there are a bunch of other examples on this page as well that you can click through and see. So like a Lego gecko and somebody walking through like a fantasy land. I'm not going to click through every video, but I will link this up in the description if you want to explore Runway gen 4. 5. I'm hoping to get my hands on this a little bit early so I can demo it some more. But again, it's not available really for anybody yet. Not even for like early testers. This week Google also introduced their new Google Workspace Studio, which is like an AI agent builder. So here's the example they share on their website. You get a prompt box, describe a task for Gemini. They give it the task. If an email contains a question for label the email as to respond and ping me in chat. We can see it says creating agent. And then we can see it built out a workflow almost like a zapier looking workflow. So step one, this is like the trigger. When I get an email, the actions to decide what to do with it, I guess. And then step three, check if step two decision is true. Add labels, ask Gemini, notify me in chat. So it built like this workflow where the trigger is when I get an email. Do these things. Things. Here's a closer look at step two, the decide phase. Does this email contain a question for me? And then they feed in the email subject and the email body as variables for it to look at. And then if true, it goes on to step three in which it adds the to respond label to it. Step five, ask Gemini, please extract the question from the following email. If there are multiple questions, list them clearly. Once again, they feed back in the email subject and the email body. And then final step, notify me in chat and here's the message that they would get. In chat, you received an email with a question for you. Here's the subject, here's the questions found. And then it pulls that from Gemini and then in their video they show that an email came in, it got the to respond tag. They open up their chat and they can see in the chat they got this message. It looks like it's basically like zapier that you can build using AI, but it's designed more for like the Google ecosystem. Right. So Gmail, Google Drive, chat Calendar, those sorts of things. As far as when we actually get access to it, they say Workspace Studio will be rolling out to business customers over the next few weeks. So soon? Ish. So this week Google also finally rolled out Gemini 3 Deepthink. So this is their absolute smartest model. When they made their Gemini 3 announcements, all of the stuff that was just crushing benchmarks were using this Gemini 3 deepthink, but they hadn't given out access to actual users yet. Well, now they're giving out access to users who are on the Ultra plan. So $250 a month plan. You can now use Gemini 3 Deepthink. And we can see Demis shared some benchmarks from it on Humanity's Last Exam blew away the competition. Opus 4.5 isn't here, but I'd have to imagine that Gemini 3 Deep Think probably even beat that one on the Google Proof question and answers. Pretty much one there on the Arc AGI 2 blew everything away there as well. So definitely like the smartest model we have access to right now. If you're willing to pay 250 bucks a month to use it. Now, the first time I heard this was available, I jumped into my Ultra account. You can see up here, I have an Ultra account. I pay the 250 bucks a month. I'm a sucker. I know they're not confessing, they're breathing. I gave it a video game idea that we're working on and asked it to flesh it out and this was the response. A lot of people are using DeepThink right now and I need a moment to sort through all those Deep thoughts. Please try again in a bit. I can still help without deep think. Just unselect it from your Tools menu and start a new chat. So yeah, that message is a little bit frustrating when you're paying like three grand a year to get access to this. However, I did come back and ask the same exact prompt like maybe 20 minutes later and this time it actually did work for me and it was pretty good. It's helping me flesh out a game idea which will be something we talk about in a future video. I'm working with a pretty cool company to do something really interesting that has a crossover between AI and video games and I'm excited to talk about it, but not allowed to totally share yet and it gave me a pretty fleshed out design. This week was also Amazon's annual Re Invent event out in Vegas. I feel like most of the announcements that they made are more geared towards like Enterprise and B2B, but I'll give you a sort of quick rundown of what I found interesting. So AWS unveiled Frontier Agents, a new class of AI agents that work as an extension of your software development team. The three agents they released the HERO Autonomous Agent, the Frontier Agent for software development, so an agent that's going to help you write code the AWS Security Agent, the Frontier Agent for more secure apps, an agent that helps make sure whatever you're building is more secure and doesn't have holes or privacy issues and the AWS DevOps agent, the frontier agent for operational excellence. The AWS DevOp agent delivers fewer alerts and more sleep for your team through always on incident triage, guided resolution and recommendations for how to continuously improve the reliability and performance of your applications. Amazon also released a new AI chip. The new chip is called Trainium 3. So let's go ahead and ask Comet real quick. What makes the Trainium chip different than an Nvidia GPU or Google tpu? All right, so apparently it's a training accelerator that prioritizes cost efficiency, tight integration with AWS services and a custom dataflow architecture versus Nvidia GPUs which are general purpose accelerators with a huge software ecosystem. And Google TPUs are Google centric AI tightly coupled with TensorFlow. So this one is designed specifically for training and pretty much specifically on AWS servers. On top of that, Amazon announced the launch of their AI factory which gives customers dedicated AI infrastructure on premise. So a lot of people right now when they need AI, inference they might use a cloud gpu, they might use something like Hugging Face, which offers Cloud GPUs you might use Grok, you might use Cerebras. There's a bunch of different companies that offer cloud GPUs that you can use. Nebius is another one. Well now it looks like Amazon's going to be offering this service as well. The AI factory offering sees AWS bring AI infrastructure including Nvidia GPUs, Trainium chips and AWS networking, storage and database to customers own data centers. So this is what I'm gathering from this. In the same way you might use Amazon S3 for storage or Amazon SES for sending emails via Amazon servers. Well, this new AI factory is going to allow you to run AI inference on their servers. So probably both AI training and the inference phase, when somebody enters a prompt and gets a response back, that's now something they're going to be offering as a service. That's what I'm gathering from this at least. The other thing that I thought was kind of fun and interesting that came out of the event was Amazon saying that Alexa plus is going to find a movie scene you're thinking about. The AI feature allows Fire TV users to describe the movie moment they want to watch on prime video without fast forwarding. Hmm. I can't really imagine how people might use this one. Alexa Basic Instinct, you know the scene. We got a handful of new large language model releases this week as well. Deep Seek once again kind of shocked the world a little bit when they released two new models, Deepseek version 3.2 and Deep Seek version 3.2 especiale both are reasoning first models which are built for agents. Here's a quick look at the benchmark comparisons. The light blue being the speciale version, the darker blue with the lines being the thinking version. And then we have GPT, Sonnet and Gemini here as these other three. And these are pretty much on par with the rest of the main big models that are out there from the big guys. The difference is these were trained very cheaply and very efficiently, where these big guy models were trained very inefficiently and very expensively. And that is the reason so many people freak out about these Deep SEQ models is that they're managing to get like state of the art level results similar to what these models can get. But they're doing it so much cheaper, so much faster and so much more efficiently than these companies that are building them. In the US the French lab Mistral released the Mistral 3 family of models this week. These are all somewhat smaller models. You've got a 14 billion parameter, 8 billion and 3 billion parameter model. And looking at the benchmarks, they're pretty much on par with Deepseek 3.1 and Chemi K2 across the board. One of the big differences between this one and a lot of other model makers is all models are released under the Apache 2.0 license, so they're actually open sourcing these models. So these are as good as a lot of these other models, but they are making them openly available. You could run them locally, you can do your own fine tuning on them, you could pretty much do what you want with them. And as part of that whole Code Red story that I sort of teased at the beginning, OpenAI sort of sneakily mentioned that there's a new model that's coming soon that they've codenamed Garlic. I was going to do a whole bit here where I bit into a garlic and then went oh God, why did I commit to that bit? But I ran out of garlic. So sorry, I'm kind of glad I didn't do the bit. We don't know much about this model, though. We they've got it codenamed Garlic internally, and we know it's performing well on the company's evaluations, at least compared to Gemini 3 and Opus 4.5 in tasks involving coding and reasoning, according to a person with knowledge. So Mark Chen from OpenAI said that they're looking to release a version of Garlic as soon as possible, which we think means people shouldn't be surprised to see GPT 5.2 or GPT 5.5 released by early next year. I've even heard rumors it could be coming out next week. I kind of doubt that, but there are some rumors that that might be possible as well. I'd probably put more weight on the Information article here claiming early next year. All right, I know we've covered a lot. There's a few more little things that I want to touch on that I think you'll find interesting. So let's go ahead and jump into rapid fire mode.

Starting with more out of OpenAI in the fact that rumor has it they're about to roll out ads into ChatGPT. So Tibor here says ChatGPT Android app one point, you know, includes new references to an ads feature with bizarre content, Search Ad and Search Ads carousel. So they were looking in the code and saw that there's a bunch of things referencing ads inside of the code. And then Benjamin Decracker here posted this Tweet, I'm in ChatGPT on a paid plus subscription. So the $20 a month plan asking about Windows BitLocker and in the response it showed him ads to shop at Target. So we can see the response here. And this is what the ad looked like that he saw like down below the chat. I mean, not great. However, I much prefer ads looking like this than them trying to put the ads sort of contextually within the content because then we'll get to a point where we never know if AI is responding to sell us something or responding because it's the best answer. So if they started putting ads in the responses, that would really suck. This isn't great either. I don't really want to see ads in ChatGPT, but again, this is preferable to the ads being within the response. Now I don't know if there's an AI bubble right now or not, but if ChatGPT starts putting ads inside the responses of the data chats, well, that's going to burst the bubble. Moving on, the CEO of Epic, Tim Sweeney, says that Steam should drop its made with AI tags. I was actually in Steam this week looking at a new game. I was thinking about grabbing and noticed on the game. It did say some functionality of this game was developed using AI, but this is still the creator's artistic vision. And this was a big name game. Like one of the most popular, most downloaded games on Steam right now had this AI disclaimer. And most likely AI wasn't used very much in the development of this game, but maybe the coders used it to like check some of their code or something and they had to put this on their Steam account. The point that Tim Sweeney is making is that in the future every game is going to leverage AI, whether just to some small degree using AI to double check code or AI to sort of edit an image or something like that. But pretty much every game in the future is going to leverage AI in some way, shape or form. So it's kind of pointless to put it on every single game. Korea AI rolled out a new feature this week called the Node App Builder. So basically if you build like a node based workflow where you know, you give it a prompt and it generates an image and then generates this video and you have this whole like workflow. Well, you can turn that workflow into an app and then share that app with people and then people can leverage that same workflow. Pretty smart idea to get people sharing their workflows and driving more people into Korea. Apple, a company I was not expecting to put something like this out, demoed a new video generative model called Starflow V or Starflow 5. I don't know which one, but this is a new way of generating AI videos. The existing way that you've pretty much used anytime you've generated a video so far used diffusion models. This one is the first normalizing flow based casual video generator demonstrating that normalizing flows can match video diffusion models in visual quality while offering end to end training, exact likelihood estimation and native multitask support across text to video, image to video, video to video generation. I'm sure you followed along. Long story short though it is a new way of generating videos that is a different method. It's no longer using a diffusion method like before. Here's some example videos of what it Gener Using text to video Pretty on par with the types of video generations we're getting now. Here's some image to video generations where you can see the input image and the output video that it generated some video to video images and there's a handful of other demos that you can look at. If you really want to dive deep into what this Starflow V is capable of, I will make sure it's linked up in the description so you can go dive in yourself if you like. And finally, the last thing I want to share is my new favorite trend I've been seeing out of Nano Banana and it's these little like game looking images that have the weather. Use the prompt below on Nana Banana Pro to make a cute image of a location with live weather conditions and it actually looks up the live weather and everything. And here's a version of London that really reminds me of like a SimCity kind of image. Here's the actual prompt. I'm not going to read the whole thing. Go ahead and pause the screen if you want to use this, but I'll also link up the tweet if you do want to copy and paste it. Here's some other examples. Tokyo, Paris, New York, Sydney. Here's one that Gavin from AI for Humans made for Mordor. I think I'm pronouncing that right. We got Austin, Texas, Mumbai. I'm not sure how to say that Relie. Not quite sure, but I love that trend. I just think these are so like aesthetic and cool looking. But that's what I got for you today. There's no dull weeks in the world of AI, which is why I'm making these AI news videos every week to make sure people like you can stay completely looped in on all of the important information and things happening in the world of AI AI is moving so fast. It's coming whether we like it or not. A lot of us don't like it, but it's coming. And your best course of action is to stay looped in and to know what's happening and know what's going on. Even if you don't like it. It's coming whether you like it or if you want to stay looped in. This is the channel to do it. So make sure you like this video. Subscribe to this channel. I will make sure you always have your finger on the pulse. I am keeping up with this every single day so I can make videos like this and break it down for you. So again, like and subscribe and press the buttons and do the things that most people suggest you do on channels to make sure you see more of it. Because, yeah, I do that for my channel. All right, thanks so much for tuning in, nerding out with me. Hopefully I'll see you in the next one. Bye. Bye. Lights, flash, systems crash, moving in a digital dash Lights flash, systems crash, moving in a digital dash. 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 Future Tools IO, 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.