AI News: These Google Updates Are Dividing People

Here's the AI news you might have missed this week. Quick spoiler alert, we're going to talk a lot about Google today. This week was Google I/O, and I did get to attend in person and see the keynote live, demo some of the things, talk to some of the people that actually built these tools, and really kind of had a very Google-centric week. But again, there was a lot. In fact, so much that Google actually released an article this week called "100 Things We Announced at Google I/O 2026." Now I'm not gonna talk about all 100. I'm gonna sort of narrow it down to the things that I think are the most impactful and the most people are gonna care about. But let's start with their brand new Gemini models. There was actually a couple of new Gemini models announced during I/O. A lot of people thought they were gonna announce some sort of giant leap with a new Gemini model and well, they sort of did. They announced the Gemini 3.5 family of models, but the one that we actually got during Google I/O is actually 3.5.5. Flash, which is their sort of smaller, faster model. So it's not the most intelligent model in the world, but it's faster and cheaper than their full-fledged Gemini 3.5 Pro, which they said is coming later. If you are somebody that's interested in the benchmarks, here's how the new 3.5 Flash stacks up. We can see it falls kind of in line when it comes to coding with GPT-5.5 and Opus 4.7, sort of falling between the two models on TerminalBench and falling just short of Opus and 5.5 on SuiBench. Bench Pro. Where it really sort of stands out is on these various agentic benchmarks for using within agents, beating out the Anthropic and OpenAI competition here. Now where this model really stands out is its speed to intelligence ratio. We can see along the bottom axis here, this is speed, so further to the right is faster. And on this axis here we have its intelligence level. So this new 3.5 Flash is pretty much on par with 3.1 Pro and Claude Opus 4.7. In terms of intelligence, but more than double as fast as 3.1 Pro and more than 3 times as fast as GPT-5.5 and Claude Opus. So it's a lot faster model for being similarly as smart. The other area where it really stands out is its pricing. So if you're using it via the API, we can see the input price for 3.5 Flash is $1.50 and an output price of $9. That's per million tokens used. So $1.50 and $9 versus Claude's 4.7 being $5 input and $25 output, and OpenAI's GPT-5.5 being $5 per million tokens and $30 per million output tokens. So a lot faster, a lot cheaper, similar level of capabilities, if not quite as good for things like coding. However, I said they released two new models this week, and the second one is the one that more people are actually talking about and is a bit more impressive, and that's the new Gemini Omni model. This model is designed to create anything from any input. Now, right now you can input video and it will understand and edit that video for you. But in the future, you're going to be able to input audio, input images, and sort of output any of those as well. Now, I actually sat down with the Gemini Omni team and here's how they described it.

You can think of it as Nanobanana for video. So a really good video editing model where you can basically take any video that you shoot in the real world or generate and then edit it using just natural language. Second thing, we're really proud where the model landed in terms of being able to preserve character consistency. So you can take, you know, images of yourself, feed it into the model, and then generate new videos of yourself. That's like really fun, but you can also see how it's useful for marketing, ads, those types of applications. And then the third thing, reason why we're calling it Gemini Omni, is we really think about it as the next step on our journey towards world models that can understand and generate anything. Right now, you can put any modality into the model as input, images, videos, text, text, audio, and generate video. But in the future, we look for the model to support also image outputs and audio outputs as well as a path towards Gemini to be like truly multimodal, which is something that we've been excited about from the start.

Essentially any input to video right now, and in the future, any input to any style output. Now the other thing that's really impressive about this Gemini Omni model is that it's grounded in world knowledge. So you know how if you're using Nano Banana, you can say, generate an image about how quantum computing works, and it will generate an infographic type image, but it'll do the research and then bake in what it found into the image. Well, Omni kind of does that as well, but with videos. So for example, here's a claymation explainer of protein folding. Proteins start as chains of amino acids. They fold into patterns like the alpha helix and flat sections called beta sheets, forming a perfect three-dimensional shape. So the prompt on that was claymation explainer of protein folding. Everything is made out of clay. No hand stop motion accurate. Now it only generates 10 seconds at a time right now, but you can see how an explainer video on a topic you're interested in could be made with just a prompt. Now, in order to use Omni, you do need to be on one of their paid plans, so AI Plus, Pro, or Ultra. But another thing they did announce during Google I/O was some new pricing for these various Google plans. So they now have a new Google AI Ultra tier at $100 a month. This is a brand new plan, and the Google AI Ultra plan that used to be $250 a month was reduced to $200 a month. So they did lower the pricing on some of those. But if you just want to get in on the lowest paid tier, they do have a plan as low as $7.99 now. Now if we head on over to the Gemini app here over at gemini.google.com, they also rolled out a new layout. We can see the design is slightly different here. We now have a new library where we can see all the documents we've created, all of the various media we've created, and it's all sort of just better organized now. There's also new sections for videos, images, a new daily brief feature that they just rolled out, and of course your ability to your chats and to start a new chat. But let's first check out Omni. If I click into the videos tab here, we can see try a template or describe a video in chat, create with Omni. Let's do create a cartoon animation that explains how diffusion models work. It's gonna go ahead and generate our video, but it's gonna ground it in world knowledge. Now you might be curious, how is this different from the Veo models? And I was curious of that as well. So I asked, and this is what they said.

The way to think about it is Veo is primarily kind of a text-to-video model, and Gemini Omni is more of a sort of video editing, interleaving different modalities together, combining and manipulating different kinds of audio, text, image, video to produce a desired output. So it's a significantly more powerful text-to-video model, but also significantly more capable model that can do a lot of other things. Like anything in, anything out. Right now we're really focusing on video, but More will come.

There's this AI startup called Victor that plugs directly into your Slack so it can get stuff done for you and your team and is already connected to everything you already use. I thought this was really interesting because I use Slack literally every day. It's where I manage my whole business from my YouTube channel to the Future Tools newsletter and website. And you know, I love using AI to solve little bottlenecks in my day-to-day work, but I'm honestly using pure AI chatbots a lot less these days and integrated AI agents like Victor a lot more. So instead of typing a generic prompt like, help me find the latest AI news from the past week, Victor can actually go into my documents, look at past outlines and research to see what I'm looking for, search the internet, and generate an outline that I can actually use for my next video. Since Victor is already connected to the platforms you're already using, it can also do stuff like pull data directly from Google Ads, analyze performance, build a report, send it to your team in Slack, and even follow up on it all on its own. I can see this being really useful across teams like marketing, ops, or engineering because they can all use the the same Victor at the same time across different departments, but with the same context and understanding. And that understanding will only get better over time as it learns your workflows, tone, and preferences. Plus, the most important part is that you stay in control. So Victor checks in before anything risky like sending an external email or moving money, so it's not going to go off and do something crazy. If you want to check out Victor for free, head to the link in the description box. And thank you so much to Victor for supporting my channel and sponsoring this portion of today's video. And in the time it took us to watch that clip, our 10 10-second video explainer is ready. Ever wonder how AI makes images? It learns to turn pure static noise back into a masterpiece. I mean, with only 10 seconds, it's kind of hard to get a lot of details in there, but I mean, the animation is semi-correct, right? It does start with noise and then finds the images from within the noise, but you can start to see how this is similar to what Nanobanana is, but for video. But it's also designed to edit video. So for instance, let's take this video here that's only 11 seconds where I'm playing with a little toy lightsaber. Let's pull it in here and let's do— turn this into a pixel art animation. And here's what that generated. Now, it didn't really stay true to my background here. You can see it added like a computer monitor in the background and it sort of changed the angle for some reason right here when the original video didn't change the angle. So it definitely has its issues, but there are some details there that are kind of cool. Like it picked up this little wolf on the back and it picked up my guitars on the wall. It did its best, but it does still, you know, change some things about it. Now, just for fun, I also gave it the prompt, "Add a monkey jumping around and dodging the sword." So instead of changing the style, I'm adding new elements into to the existing video. Let's see how it did. So we can see our monkey in the corner here. He jumps out and then another monkey appears. So it kind of added two monkeys and then one of them disappeared, but it added a new monkey that wasn't originally in the scene. One of the other features they talked about was the avatar feature. So if I come down to this plus here and I go up to more uploads, we can see there's a new avatar feature. So if I click on avatar, we can add an avatar. All we do is take a quick selfie video and it'll bring our avatar and voice to any scene with one tap. So let's try it. So it put a QR code on my screen. I'm blurring it out just because I don't know what happens if somebody else scans it, but I'm gonna go ahead and scan this QR code here. It's asking me to allow camera and microphone. Okay, let's try this. 44, 95, 17, 60, 47, 17. Okay, so now it's uploading my avatar, and now when I jump back to my computer, it looks like I have the ability to use my avatar. So add yourself to any frame you create by typing @me or @MattWolfe into your prompt. All right, so let's go ahead and use avatar. Let's me fight myself at Matt Wolfe battling the man in the video. Let's just see what it does here. All right, so I can already tell it's a bit wonky just from the screenshot, but let's see what we got. Show me what you can do. Okay, well, that didn't quite look how I thought it would. I also thought I would test it with using the avatar plus the world knowledge, so I said, my avatar explains how quantum computing works on a chalkboard, and this is what it gave me. Traditional computers use bits that are either 0 or 1. Quantum computers use qubits, which can be both at once. This phenomenon is called superposition, allowing for immense power. I mean, it actually did a good job. I don't know why it gave me an Arizona Diamondbacks hat, but other than that, it looks like me. It gave a very simplistic version of what quantum computing is, which, I mean, not super in-depth but actually accurate. And overall, this seemed pretty impressive. I gave it prompt, my avatar slam dunking a ball on a beachside basketball court. Let's check it out. Oh, that was a very weird slam dunk, but the likeness, it nailed it. It actually got the correct hat this time as well. Either I'm like 9 feet tall or this basketball hoop is only like 6 feet off the ground. It was also very generous on the muscles there. I do appreciate that. So that's Gemini Omni, which as of right now is really useful for videos, generating videos with knowledge, generating videos with your own avatar in them, and actually editing videos. Now, another big announcement that came out of Google this week was Gemini Spark, and this is Google's answer to like OpenClaw and Hermes, and it's them putting an agent out into the world. Instead of prompting things and getting responses back, it will actually go out and do actions on your behalf, unlike Hermes and OpenClaw, which will run locally on your computer or on a VPS, on a cloud computer somewhere. This one runs entirely on Google's servers, so it will always continue to run even if your own computers aren't online. I had the opportunity to sit down with Korey, the CTO of Google DeepMind, and here's how he described it.

With Spark, we have a full computer that is running in Google Cloud for you. You might say, look, every day in the morning send me an email that is like motivating me. It'll do that. It is very simple. But you can also ask every day, summarize all the news that is going on around and then send me an email or write that in a doc. Or you can say we have a neighborhood party organization going on. Take that organizational job. Like it can actually do things like that where it can communicate with the neighbors and make sure that everyone is following up. So it can actually do actively things for you that you don't need to pay attention every time.

Now, since this has access to your Gmail, your Google Calendar, and anything that's in your sort of Google ecosystem. It can connect all those together and take actions with those tools. So some examples on their website: it'll give you an inbox summary from your email, it'll do things like look across your emails and your Google Calendar and your Google Drive and then pull information together and put it all into like a Google Sheet for you. From their blog post here, with Gemini Spark you can set recurring tasks or triggers, teach it new skills, it can create complete workflows, ask it to synthesize raw meeting notes across emails and chats, create polished Google Docs with findings, and even draft the companion email kicking off a project. It's gonna use MCP connectors to work with things like Canva and OpenTable and Instacart to be able to book things on your behalf. Now, if you've already built out skills for your OpenClaw or Hermes or Claude Code or Codex, I actually asked during one of the demos, are we gonna be able to pull those over and actually use skills that have already been created? And I think I remember they said you can import skills over if you've built them in other—

That's right. Okay. You can upload skills.

So yeah, there you go. It shouldn't be too hard to move over if you are using other agents and wanna try this one instead. Instead. As far as when it's going to be available, it's currently rolling out to trusted testers, and then after that it will be available to Ultra subscribers, and then after that I'm sure they're gonna sort of roll it out to lower and lower tiers. Now if you're worried about an AI agent owned by Google having access to a lot of your private data and making decisions on your behalf, I did actually ask Korey, the CTO of DeepMind, about that as well, and here's what he said.

Everyone will have a different threshold there. So it is more about designing a system where users are feeling in control for the kind of control that they want. And the agent, there are, of course, like there are sort of like commonly accepted, there are some safety thresholds that you don't want the agent crossing without you ever approving things. For example, right now we are fine with the agent sending you email because it's your agent. But like if it needs to send an email to someone external, it needs to ask for permission. Or if it needs to do like anything that touches external world a little bit on your behalf, you should be in control, right? Like you need to say that, okay, you can approve it. You can say that I'm comfortable, do it and do it in the future. But like you should be able to feel like you are in control. Those kinds of things are quite important. Exactly like in what cases people are going to use these, like we are not trusting ask, like prescribing specific use cases, but the kinds of things that either information sharing about you or taking actions on behalf of you, users should feel in control.

So everybody on the Google team was constantly reassuring us that they are designing this to make sure you stay in control and your agent doesn't go rogue and buy things on your behalf or send emails you don't want it to send. Now I'm just sharing clips and I am kind of cutting up these clips to keep this video moving a little bit, but if you want to see my full conversations with the Gemini Omni team as well as with Caray, the CTO of DeepMind. I will be making the entire interviews available to subscribers of the Future Tools newsletter. So if you're not subscribed to the Future Tools newsletter, head on over to futuretools.io and get subscribed and I'll make sure that the full conversations are available for free to all subscribers. All right. Moving on to the next thing Google announced, which was somewhat of a more controversial update, and that is how they're handling Google.com. From and the actual search engine. It's starting to sound like this search engine is going to get more and more replaced by like an AI engine. So the first thing we'll notice with the new Google Search is that it's turning into a longer prompt box instead of just a one-line search. So you can type full-in prompts straight into Google Search instead of going to AI mode. So you're essentially going to be in AI mode by default in the near future. They also demoed what they call search agents, which allows you to search for things and then continually sort of alert you on those things. It'll keep you updated on those things on a regular basis. This almost feels like an evolution of like Google Alerts where it will constantly do searches and then alert you when there is new information based on the thing that you set it up to keep on top of for you. They also showed off that Google will write code directly in the search response. In this example, they say, how do black holes affect spacetime? And you can see it actually codes an interactive visual where they drag the slider around to show the different visual of how things are impacted. Now, this is similar to features we've recently seen roll out in both Anthropic's Claude and OpenAI's ChatGPT models, where it will actually generate visuals for you as well. And now it looks like something that could happen directly in search when you give it prompts where a visual will help better describe the explanation. Now, the reason this is sort of a controversial update is that it's sort of making the AI response sometimes even more prominent than the actual search results for what you were looking for. So if I do a search for something like best video games of 2026, you have your AI overview right here, but it takes up a relatively small amount of space on Google. I can click show more and sort of get directly into AI mode here, but without doing that, I have some organic rankings of people who actually wrote blog posts on the best video games of 2026 so far. These are websites that rely on, you know, people actually clicking on the link and going to their website. They're ad-supported or subscription-supported, or they're trying to grow a mailing list or sell a product on their site to support their business. Now, if Google essentially scrapes the responses from all of these blog posts, puts them in an AI overview so that nobody ever ends up clicking into the blog post, well, that's going to mean a drastic reduction in traffic from Google to these websites. And the responses that Google is giving is based on the content that they got from these websites. So if people lose their incentive to create content for Google and they stop creating it, then where does the content come from that Google is then pulling from? And that's kind of the issue people are having with this new update. If it's going to work the way we think it is, if I do the same search inside of AI mode here, it just gives me the response. I don't need to click click into any blog posts. It's just telling me the answers to my prompt here. And the links to external websites are pushed over here to the right, but I'm less likely to click on 'em because it just kind of told me what they say anyway already. And they're also rolling out this same feature to YouTube. They're rolling out an Ask YouTube feature where you can ask questions on YouTube. It's gonna search through the videos and sort of provide you with the answer potentially without you needing to even click into the video. So you can see how this could really negatively impact content creators, making this probably one of the more controversial decisions. And we'll see if Google really pursues this hard or if they eventually kind of roll back this idea, because at the end of the day, they need people to create content so that they can actually have a place to pull the answers from that they're going to present in their AI overviews. So I don't know how smart of a decision it is to take away that incentive for people to create content. Again, this is unfolding still. It doesn't appear that this new mode is rolled out quite yet, but the real question is going to be how is Google balancing the AI response with actually organic links and incentivizing people to actually click away from Google? Because that's going to be the balance they're going to have to figure out if they still want people to like favor Google. Quite honestly, this feels like an opportunity where a company like like Bing or somebody else that's been trying to top Google in the search engine game could sort of swoop in and be like, hey guys, when you give us a prompt, we're just going to show you organic links and we're not going to give you AI responses. And if a company did that, you're probably going to see people who just want the old version of Google migrating to a tool like that. Again, Google's going to figure this out, I think, and they're going to constantly play with that balance of links versus AI responses. But, uh, it'll be interesting to see how this plays out. All right, a few more announcements from Google and then we'll get into the non-Google stuff. I'm going to kind of move more quickly because I think I've already talked about some of the biggest things during Google I/O. Google introduced what they're calling Universal Cart. It's a shopping cart that actually works across merchants and across services. So essentially the idea is you can be browsing the web looking for different products. You find one product on one Shopify store, you add it to your cart, you go to a different store and find a different product that you want and that's completely unrelated on a different Shopify store or on a completely different website, you know, target.com or whatever. You find something you like there, you add it to your cart there, and you just browse the web and you're adding stuff to your universal cart. And when you're done, you check out, and then it just kind of sends the payments to all of the places for you. That's my understanding of what this universal cart is— just one cart that you use for the entire web. They also showed off that this cart has like intelligent features, like it'll show things like this is the lowest price it's seen in 30 days. It might suggest you use your Amex card to get 3% cash back, and it will actually double-check and confirm that, uh, if it thinks you're making a bad choice, that do you really want to make that choice? For example, the CPU and motherboard that are in your cart are not compatible. So the example being they added a CPU and a motherboard and it noticed that they won't work together, thinking this person's probably buying a computer, and tells them, hey, are you sure you want to buy those two together? That could actually be pretty handy, honestly. They also created the agent payments protocol, which is going to allow your AI agents to go and check out in these carts on your behalf. And like Corey already said earlier in our conversation, they are making it so you can kind of adjust the level of trust you have for your agent to go buy things on your behalf. If you're a coder or a vibe coder, Google introduced Antigravity 2.0. This is their new IDE, their integrated development environment, where you can have it go and write code for you on your behalf. Here's a screenshot of what the new Antigravity looks like. If we're being honest, their old version was kind of like a VS Code fork because they brought on the Windsurf team and built it based on that. And this new version, it looks an awful lot like OpenAI's Codex product. Like, I mean, pretty identical. They released some new features in Pomeli, which is like their brand building kit. It sort of figures out your color schemes and your fonts, and it'll build little ads for you and things like that that are on brand. Well, they added a few new features like brand books to generate comprehensive guides featuring your brand's custom images, fonts, and colors, and website You can design and stand up a complete website in just a few clicks using your entire brand kit that it designed for you. They also showed off their Vibe Design website tool called Stitch, where you give it a prompt for a website that you want and you can actually watch it sort of build that website out for you in real time. So it sounds like the website building is in Pomeli, but now also in Stitch, but you can also build websites websites directly from Gemini, and you can also build websites directly from Antigravity. And pretty much all of the Google products will build websites for you now, it seems like. They also announced Gemini for Science. This is a collection of science tools and experiments designed to expand the scale and precision of scientific exploration. So it's like a suite of tools to do scientific research and things like that. And they also launched Science Skills, which is a specialized bundle that integrates insights from over 30 major life science databases and tools, including AlphaFold, AlphaGenome, and InterPro. And this is one that I actually thought was pretty dang cool. I've always thought Project Genie was fun. It's where you can sort of prompt almost like a video game world that you can interact with for like 60 seconds at a time. I've played with it in some previous videos. Well, this new one, you can actually do the same thing but put it in real locations using Street View from Google. So if we head over to Project Genie here, let's go ahead and create our own. And we can see it says choose a location from Google Maps. So let's go ahead and say select that, and here's a location. But just for fun, I'm gonna go ahead and put Petco Park in San Diego. Okay, so here's like the Petco Park baseball stadium here. Let's zoom in on our map here and maybe drop it over on this side of the park. So now we're on the streets outside of Petco Park, and let's go ahead and use this location. And then for our character, let's do a monkey on roller skates. I haven't done that in a while. And look at this, we've got our monkey on roller skating around near Petco Park in San Diego, and it's actually grounded by real-world map data. I just— this is so wild to me because I remember 3 years ago or so when I was playing with Modelscope and Xeroxcope, I would generate a monkey on roller skates and you could barely tell what it even was. You're kind of like, if I squint, I see a blurry, fuzzy sort of monkey there, and it only lasted like a couple seconds. And here we are, fast forward just a couple years, now I have a monkey on roller skates skating around in downtown San Diego, and I can actually control him and make him jump around, and it's in the real-world environment. Like, this to me is just absolutely mind-blowing. Like, it's obviously not the best, most insane graphics you've ever seen, but just how far we've come in the amount of time that we've come is just so mind-blowing. So yeah, this is like one of the coolest features that rolled out. Now, let's be fair, I have no idea what you're gonna use this for. I mean, maybe you're gonna make some fun little videos. It's kind of like a novelty right now. Like a fun little demo that you could play with, but it's really impressive that we can do it now. Anyway, that's the new update to Genie, the new like Street View grounding, which I thought was pretty cool. One of the other big announcements they made was around eyewear. I was at Google I/O not last year, the year before, and they were already teasing AI glasses from Google. And then I was there last year and they demoed AI glasses from Google. And then I was here this year and they demoed AI glasses from Google. We finally have something that's going to get into our hands and we kind of know when it's going to get in our hands. Google is partnering with Gentle Monster and Warby Parker to put their version of AI glasses out, and there's going to be two versions. There's going to be ones that have the camera and the audio, very similar to the Meta Ray-Bans, and then there's going to be the ones that have the display. The audio version are going to launch first, and they're coming in fall. We still don't know when the ones with the display are coming. You can do a lot of this stuff that you can do with the Meta Ray-Bans on these glasses. But they almost sound like they're more designed to be on all of the time and sort of giving you things in your ear. Like when you get a new text message, you don't have to pull your phone out and look at it. It'll just say, "Hey, you have a new text message," and sort of tell you what the message is straight to your ear if you want it to. But you can also ask about what you see. It'll help you navigate places. Like I mentioned, you can manage calls, send texts, have Gemini summarize missed messages, things like that straight into your ears. You can snap photos and videos, translate speech and writing, order an Uber, you know, get help with various tasks. Again, a lot of the same stuff you could do with the Meta Ray-Bans sound like they're coming to this first iteration of the Google Glasses. And they announced that they're going to make it easier to understand how content was created and edited. Their SynthID embeds imperceptible signals into AI-generated content, but they actually announced that other companies are going to adopt this standard, including competitor OpenAI, ElevenLabs, Kakao, and more companies to come. So more and more companies are going to adopt this SynthID standard, which means you'll be able to drag an image in or drag an audio file in or drag a video into Gemini, and it will actually tell you if it was generated with AI or not, because the AI can read the SynthID watermark, but it is imperceptible to the human eye. So cool to see that even competing companies are getting on board because they all believe that it is very important to prevent, you know, fake information and deepfakes and that kind of thing. And again, there was over 100 announcements made during Google I/O, but Those are the ones that I found most interesting and probably most important for the most amount of people. But I'll link to everything that I came across in the description so you can go and learn more about those. If I was to sum up Google I/O and what they were trying to accomplish this year, it felt less like they were trying to benchmark Max and try to get the best absolute model on all of the benchmarks this time around, and more like they're hunting for how do we actually make this useful? How do we bake this into more products? Products and find more ways for people to actually use these AI tools without having to think about it. How can we be more proactive? How can we provide like daily briefings to people automatically? How can we create agents so things just happen on people's behalf? How do we make it easier and more intuitive to generate videos? How do we make it easier to browse and shop the web? It just feels like for a long time we've seen companies just push the benchmarks, push the benchmarks, push the benchmarks. Benchmarks and try to get these models to be better in areas like science and math and things that like most daily everyday people aren't really using these models for. And it felt like with this event, Google was like, okay, let's go to the general consumers, let's go to the small businesses, let's go to the like individuals and let's try to figure out ways to make all of this AI stuff we're all releasing actually more useful and integrated into the ecosystem they're already using. I have a feeling a lot of the stuff that we saw at Google may not exist a couple of years from now. Some of the things will probably get ramped up. It feels like they're throwing a lot of things against the wall, trying to figure out what users are actually going to use when they bake AI into it. Those things they're going to probably double down on. And then the things that people push back on, well, some of these things they may go, okay, we went in the wrong direction a little too far with that. Let's pull back on some of these things. That's kind of what it feels like Google's doing. They're in that mode right now because right now It's a lot and it's overwhelming and sometimes confusing. Like there's 6 different Google tools right now that have AI baked into them that I can go to to build my website. When it comes to videos, you've got, you know, Google's Flow and you've got Veo and you've got Gemini Omni. When it comes to images, Google had Imagine and then they had Nanobanana. When it comes to coding, you can generate code over in AI Studio or you can use Antigravity or it'll now generate code straight in AI Search or you could do do the code over on the Gemini app if you want. Like, there's so many different platforms and there's a lot of like overlapping use cases on a lot of these platforms. And I think where to go to do certain things is sort of getting convoluted and confusing. And I think that's what Google's gonna have to figure out. And I think they will. I think they're going to sort of dial it in. And right now, again, it's throwing a bunch of spaghetti at the wall and seeing what sticks. But that's not all that happened. Before I jump into the rapid fire, there was one other sort of bigger announcement that happened at the end of last week. But because it happened on Friday, it didn't make it into this video. OpenAI rolled out a new personal finance experience inside of ChatGPT. So this actually allows you to connect your bank details using Plaid straight into ChatGPT, and then you can actually use ChatGPT to learn more about your personal finances and actually get guidance on your personal finances directly in ChatGPT. But it's grounded in your actual bank account information. So you can ask it things like, how do I save on subscription costs? Or how do I balance my investment portfolio? Things like that, and it will look at all of the financial data it knows about you and be able to give you direct advice. Now, do most people want to give ChatGPT financial details? Uh, yet to be seen. It is an interesting concept. I'm just not sure how I feel yet about a company who's trying to build like an ads-based business on top of their normal business having details about my finances, because at what point do some of those financial details get passed along to the potential advertisers in OpenAI? Like, if OpenAI AI knows all of my spending habits, how much money I have available, what I invest in, where I travel, all of these kinds of details. That's starting to get into like creepy territory when it comes to one company knowing a little too much about me, especially if they're trying to put like ads in front of me. And also if there's ever a point where maybe some of the responses ChatGPT gives me are influenced by data that it has on me as well as what advertisers want them to do. I've gone the whole ChatGPT ad rabbit hole. I'm not a fan of that model for chatbots. Hopefully it doesn't play out in the way that I'm imagining it could play out. But if it does play out the way that we all don't want it to play out, it having direct access to all my financial details is one of the last things I feel like I wanted to have. Anyway, let's talk about everything else that happened this week, but I'll do it quickly and we'll do it in a rapid fire. I've talked about the Elon Musk versus Sam Altman trial that's been happening for a few weeks. TL;DR: Elon Musk lost. The court basically ruled the statute of limitations is up and Elon really doesn't have a case anymore. That's kind of how it all ended. This was a story that I don't think anybody saw coming, but it actually happened right before Google I/O kicked off, and that's that Andrej Karpathy joined in Tropic. Now, if you don't know the whole lore behind this, this is interesting because Andrej Karpathy was one of the original co-founders of OpenAI. He helped develop LLMs as we know them. At one point, he was poached away from OpenAI and went to Tesla and helped develop a bunch of the AI over at Tesla. He left Tesla, came back to OpenAI, worked at OpenAI for a couple more years, and then decided to leave OpenAI to focus on his own projects. His own projects being he wanted to figure out how to build like an AI education company. He's been working on that for the past year or so. I don't remember exactly when he left OpenAI, but to hear that he now joined Anthropic is just kind of crazy because of all of his history over at OpenAI. And he's brilliant, by the way. If you haven't checked out Andrej's YouTube channel, he's a must-watch on YouTube. He actually goes into deep dive tutorials on how you can actually build your own LLMs if you want to. And he's, I believe, the number one voice in AI. AI over on X. He's got 2.6 million followers, and pretty much anytime he tweets anything, it goes viral. So it's just wild to think that he's going over to Anthropic now. We actually got a few things out of Spotify that are pretty interesting this week. They put on an event out in New York and made a ton of announcements, but I'm going to focus in on a couple of the AI announcements specifically, starting with the fact that Spotify and Universal Music Group announced a licensing agreement for fan-made covers and remixes. So this is like the music industry getting on board with AI and saying, we're gonna let our artists be remixed with AI. It does say it will open up additional revenue streams, meaning it sounds like if you remix an artist, that artist is actually going to get paid for those streams of songs that were remixed. So Spotify is sort of figuring out how to get musicians paid when music is generated that sounds like that artist. Like, this has been a big gap in the AI sort of music world that people have been trying to figure out, and Spotify seems to be doing it. This introduces a creation model where artists and songwriters can directly share in the value generated through AI-driven licensed covers and remixes on the Spotify platform. Now you're gonna have to be on a Spotify Premium plan to use it, but I like this move a lot. Now a few other announcements that Spotify made this week around AI is about podcasts. You're actually going to be able to use AI to ask questions about podcasts. Users can ask Spotify questions about the podcast they're listening or watching and get answers in real time. Whether it's clarifying a concept, exploring something mentioned in an episode, or getting recommendations related to their favorite creators, they can delve deeper without leaving Spotify. But they're also letting you create your own AI-generated podcast, sort of competing with something like NotebookLM. They're releasing something called Personal Podcasts, which is an experience that lets listeners generate and schedule short private audio episodes tailored to their interests and their voice. In their listening habits. So Spotify allowed people to create custom podcasts using agents like your OpenClaw or Hermes, and when they saw people doing that, they basically said, hey, let's just build that into Spotify. They say all you need to do is write a prompt and Spotify generates personalized private audio based on your input. It draws on world knowledge and your Spotify taste profile to create something relevant and uniquely yours. So things like a daily briefing, a deep dive on topics you're curious about, or a weekly roundup of whatever's on your mind. Some examples: share my daily city updates and tell me about local concerts from artists I love, or help me understand economics in 5 minutes, and Spotify will go and create that custom podcast for you. That's pretty cool. Interestingly, Amazon released something very similar this week. Alexa Plus now generates podcast episodes on demand. A new feature creates AI-generated audio episodes on any topic, ready to listen whenever you are. Alexa can now generate podcast-style episodes on virtually any topic in just minutes. No documents or prep work needed. Just ask and Alexa does the rest. If you're a fan of Cursor, they released a new model that's really impressive this week called Composer 2.5. When it comes to coding, it's pretty on par with state-of-the-art models, performing about as good as Opus 4.7 on TerminalBench. On SuiBench Multilingual, it's about on par with Opus and even better than GPT-5.5. But the biggest deal is the cost here. So we can see Composer 2.5 on this chart is almost as good as Opus 4.7 on Macs, almost as good as GPT-5.5 5.5, but Opus is way up here at like $11. GPT-5.5 is like, I don't know, $4.50 or something. And Composer 2.5 is way down here at like less than a dollar per task. So it's nearly as good as these other models, just way, way cheaper to use. And this is a model you can use directly in Cursor to do most of what you need it to do. And then when you need a much smarter model, you can flip over to GPT-5.5 or Opus, but mostly use Composer most of the time. The company stayed Stability AI, which is a company I haven't really talked about in quite a while, released a new music model this week called Stable Audio 3.0. This new model can generate up to 6-minute songs as well as generate sound effects, and it's open weights. So the sound effects, the small and the medium models are all completely open, and only their largest 2.7 billion parameter model seems to be closed off. The company Krea AI introduced LoRAs for Krea 2. So a LoRA, think of it as like a, a plugin or like an extra effect that you can add on top of images that you generate. So you can have a specific style that you like, you can train a LoRA on that style, and then any image that you generate will match that same style. Here's another example of it where these are probably the images it was trained on, and then when they asked it to create an image with that same character, it made an image right here with that same character. I don't know if I've ever really talked about this company, but I use this tool behind the scenes a lot when I'm on meetings. It's called Granola, and it's a tool that when you're on Zoom meetings or Google Meet meetings or things like that, it will actually listen in and take notes on that meeting for you. And then afterwards it'll just kind of give you a summary or bullet points of what was talked about in that meeting. It's really, really handy. Well, they just rolled out a new feature that actually goes and does research on the people you're talking about before you even get on the call. So their new Briefs feature prepares you for your next meeting. We can see here that if it knows you have a meeting the next day, it will look up that person on LinkedIn and find out details about them. So when you do jump on a call you have a little bit of extra context to help with the conversation you're about to be on. And finally, as I often do, let's end with a robot. This week, Boston Dynamics showed their robot that can pick up and carry a fridge around. So instead of asking the robot to go to the fridge and grab you a drink, this robot goes to the fridge and will bring you the whole dang fridge. It'll just walk it right over to you. Why just grab a drink when you can have the robot bring you the whole fridge and then you can pick the drink yourself? The one thing I find funny about this is well as if I pause it right here. Yeah, that robot's done a little bit of damage to that fridge, but it's still impressive that it can go and carry this fridge around for you. Also, I don't know if that drink's gonna be cold. This fridge isn't plugged in, clearly. Anyway, that's what I got for you this week. I warned you it was a lot of Google. Google kind of owned the week with all the announcements they put out at Google I/O. Google did not sponsor this video, but they did fly me out and put me up in a hotel room at Google so that I could attend Google I/O. So thank you to them for that. But they had absolutely no control of what I said or talked about in this video. Again, I record these on Thursdays, put them out on Fridays. Any news that came out late Thursday or on Friday will make it into next week's video. My goal with these videos is to sift through all the AI news all week long, drink from the fire hose, read about all the hype and all the news and all of the announcements and all of the launches, and find just the signal and turn around and deliver it to you each week. Find the things that just are important to the most amount of people, because AI is moving so fast. There's so many announcements, there's so much hype, there's so many new tools being launched every day now that anybody can vibe code anything. I think it's important to try to filter through all that and find what's important to people and what's just kind of background noise that you can ignore. If that's something that you're interested in and you want to make sure you are always tapped in without needing to keep up with the news every single day, consider giving this video a like and subscribing to this channel, and I will do my best to give you that one breakdown video every week that's going to keep you fully looped in and on the cutting edge of what's happening in AI and the world of sort of future technology. That's my goal. Again, I'd love the like and subscribe. I'm trying to get to a million subscribers this year, and your support really, really helps me out. But again, thanks so much for tuning in, hanging out with me, nerding out with me. It's been a crazy week in the world of AI. Google just sort of unloaded on all of us, and, uh, yeah, I'm, I'm here for it. I'm here to keep up with it and do my best to break it all down for you. I really, really appreciate you, and well, hopefully I'll see you in the next one. Thanks so Much. Bye-bye. No!