29 Biggest AI Stories You Missed This Week

There have been a ton of interesting developments in the world of AI over the past couple of weeks, and today I'm going to get you all caught up. We've got some impressive new video models, smart glasses that seem to blow everybody's mind, some big AI updates out of YouTube, and a new app that'll actually pay you to let AI spy on your phone calls. All that and more in today's AI News roundup. There's a lot to cover and I don't want to waste your time, so let's dig right in. Get in my belly. Starting with the newest feature inside of ChatGPT. It's called ChatGPT Pulse, and it's basically a daily news digest that ChatGPT will send to your phone every morning based on what it knows about you. Pulse proactively does research to deliver personalized updates based on your chats, feedback and connected apps like Calendar. You can curate what ChatGPT researches by letting it know what's useful and what isn't. So when I open up the ChatGPT phone on my app here, we can see a new screen. I'm here to surface what's helpful to you. Once a day, every day, you decide what shows up. If you tell me what to focus on, I'll curate it for you tomorrow and we can see some of the stuff that it curated for me. And I'm guessing this is just based on previous conversations I've had with ChatGPT and what it knows about me, like the AI toothbrush gimmick. I definitely need to know more about what that is. Everyday objects with attitude, coffee makers that nag, lamps that boast. Interesting. And then we can see it says you decide what shows up here, what's on your mind lately. I'll remember it for tomorrow's update. So if I put I'm curious about and then put the latest AI news updates, tomorrow's curation should have information about that. This, to me seems very much like Perplexity's like Discover feature. Although this feels even more tailored to you, where Perplexity seems like it's a little bit more general news, but this could come in pretty handy. And if I'm being honest, this is kind of scary for the newsletter business, which, if you know me, I have a very large newsletter, the Future Tools Newsletter, that curates all the latest AI tools and news. Something like this might just curate it to your phone without you needing to read emails. I hate to say it, but it's kind of reality. We got access to a ton of New video models this week as well, including Luma's Ray 3, which is the world's first reasoning video model. They also claim to be the first to generate studio grade HDR or high dynamic range video. There's also a new draft mode where you can have it generate the video pretty quickly, but in sort of a lower resolution, lower quality, and. And then if you like the way the draft version came out, you can have it generated in a higher quality version. It also apparently has much better physics and seems pretty solid. I actually got a demo in person in New York a few days ago, but let's go ahead and try it out. I'll jump over to lumalabs aidreammachine. Once inside Dream Machine, let's click the plus here and then down in the bottom, we should be able to change the model. So right now we've got Ray 2 and I can select Ray 3 or array 3. Reasoning. Let's go use the reasoning model and let's set the dynamic range to hdr. It should give us a sort of larger variation in colors. We'll try the prompt. A man standing on his driveway clearing snow with a flamethrower. And we'll turn on draft mode so that it'll generate something a little bit faster. We can actually see it says analyzing the scene of a man clearing snow on his driveway using a flamethrower. So it's actually thinking through this a little bit. It generated what seems to be like the starting frame here we can see it's examining the snowy driveway scene to enhance the man's determined stance and the cold cinematic atmosphere. It actually annotated some stuff. So see these circles and drawing here We've seen. That's a really good method for telling these images what to do when they become videos. Well, this just drew those on there for me. I didn't have to actually annotate anything. It's using AI to do the annotation. It's still going. But you know what's crazy about it is it made this video of this guy with the flamethrower, but the flamethrower looked off, like the flames weren't coming out of the front of it. And it seemed to have noticed that and then remade the video fixing it so that the flames actually come out of the front of the flamethrower. Like, that's wild. Like, this is to me like a truly agent dick video model. It checks its work and then goes back and makes its work better. Okay, so it's done. And it did take A few minutes. So it was a little bit longer than I thought. But we can see it generated four drafts. And you can see in this first one, the flames aren't really coming out of the flamethrower, they're coming out of the top. And it doesn't really look like snow is melting or anything. We jump to the second draft. Same kind of issues. Then we jump to the third draft, and look, now it's actually coming out of the flamethrower, and we're seeing the snow on the ground melt. And then for whatever reason, the final version is fairly jacked up, where a second guy just sort of pops in and there's like a weird jump cut. But this third one here is pretty good. And we can see that's a draft. So let's go ahead and upscale this one into HDR and turn it into not a draft. And now we have our higher definition version in 720p. And if I wanted to, I could modify it, extend it, or make more like this. But we'll leave it there for now so you can go and play around with it and see what you get out of it. The video model cling 2.5, which a lot of people say is pretty Dang Comparable to VO3, has been made available in various platforms. If you use Kreia AI, you can now use cling 2.5 directly inside of Krea. Or if you're more of a fan of Leonardo AI, you can generate videos with cling 2.5 directly inside of Leonardo as well. Let's go ahead and do it inside of Leonardo. Create with cling 2.5. We'll make sure we have video video selected. And then over on model, it's still selected as VO3. So let's go ahead and click that and swap it over to cling 2.5 turbo and we'll give it the prompt. A fashion show where everyone walking the Runway is wearing clothing made of spaghetti. I will say that cling 2.5 is pretty slow at generating. It's been generating for over seven minutes now, and I still haven't gotten my video. Now, it took kind of forever to generate, well over 15 minutes. But this was the output that we got from cling 2.5. I mean, yeah, that looks like spaghetti noodles on those models. Followed the prompt. I don't know if I just saw this video. If I would realize that supposed to be spaghetti. It just kind of more looks like strings. But this is cling 2.5. For the sake of comparison, this is what VO3 created given the exact same prompt. Something that in my opinion is much more clearly looking like spaghetti. And that tie is kind of awesome. The platform Higsfield AI also added cling 2.5, so if you want another platform to go use it in, it's there as well. But they also added 1.2.5, which is pretty widely regarded as the best open source AI video model. So it's open source. If you have a strong enough computer, you can actually run this one locally. And it generates sound and is pretty comparable to VO3. But let's give 1.2.5 a try inside of Higgs Field. So I'll come up to the top or video. We'll select 1.2.5 here. And since this one generates audio as well, let's give it a prompt. A traffic jam where every car is shaped like a rubber duck. And when the driver honks the horns, it sounds like a squeaky toy. And I'm just going to generate it at 480p just to kind of speed up the process. But you can generate all the way up to 1080p and you can generate five or 10 seconds. But I'm just trying to make a quick demo. So let's go ahead and generate this. And it does say that a start frame is required. So let's go ahead and take this exact prompt and we'll generate our starting image. Paste that in here. Traffic jam, where every car is blah, blah, blah. So let's go ahead and use the model. They have less models available than I thought. So let's go ahead and use the flux context max model to generate this image. And we got a couple options for our traffic jam. So let's go ahead and use this. This one. We'll click it on video and then once again we'll leave it on 1.2.5. And we still have our prompt in here. So let's go ahead and now generate it with our squeaky toy image. And this is what we got out of it.

I mean, it doesn't really sound like a duck squeaking. And I don't know why they randomly get like horns in their mouth that, like, disappear. But it is another model that will generate both video and audio for you. I just picked a very, very random example to try. So, yeah, that's available over on Higgs Field as of right now. And just as a quick update, in case you're curious, if you have a Krea account, I did notice that wand 2.5 is available inside of Krea now as well. So multiple platforms with Multiple video tools, all sort of available. My personal choice has been Leonardo, but I'm definitely biased. And Kreia and Higsfield have a lot of great models that you can use inside of them as well. The tricky part with AI right isn't finding cool tools, it's getting them to actually work together. You use a chatbot for one thing, then copy and paste it into another tool for the next step, then hop into another platform for images. Everything feels so scattered with no easy way to connect the dots. And that's where make comes in. It basically gives you one place to build and run entire systems of AI agents and automations, all in real time. Let me give you an example of what that looks like. I set up a workflow that I call the Campaign in a Box agent. Picture. You're a marketing agency that just picked up a new client, say a coffee brand. Normally you'd have a team doing research, writing copy, designing ads, and scheduling posts. With make, you can have AI agents handle all of that on one visual platform. Here's the flow. First, an AI research agent scans social media to figure out what people are saying about coffee right now. That feeds into a copywriting agent that drafts ad copy and captions the client's voice. At the same time, a design agent spins up creative concepts, product shots, carousel ads, even mockups with new slogans. Once it's ready, the post gets pushed straight into a scheduler like Meta, Ads Manager or Buffer. And finally, the whole system loops back with analytics inside Make Grid so you can see which ads are working in real time and even shift more budget to the top performing content automatically. This isn't just automation, it's orchestration. Instead of juggling disconnected tools, you've got a living campaign that can adapt, scale and self optimize all from a single dashboard. For agencies, this means less time stuck in production bottlenecks and more time focusing on strategy and clients. And for businesses, it means faster workflows, lower costs, and better results. So if you've been wondering how to actually put AI to work in your business, this is the way. Make lets you build, scale and control AI driven workflows at every level. And right now, they're giving my audience one month free of the Pro plan if you use the link below. So whether you're running an agency or a fintech startup, or just want to automate smarter, now's the time to try it. Hit the link in the description, grab your free month and start orchestrating your AI powered business with make. Thanks so much. To make.com for sponsoring this one. Now let's get back to the rest of the video. Now that we've talked about generating video, let's talk about editing video. There's this platform called Mirage, and they used to be called Captions, but they rebranded as Mirage recently. And it's an AI video editing tool now. It also makes like AI avatars, but their newest feature is the ability to make edits entirely with text. Want to zoom at 3 seconds? Just type that into the box. Need captions added, done in a snap. You can even type a vibe you're going for. And our AI editing tool will apply the right style and aesthetic to bring it to life. Now, this editing feature that they just rolled out is on the mobile. So even if I logged into Mirage Studio on my desktop, we won't see this editing feature. So let's pop it open on the phone here. So this is what the app looks like when you're inside of it. I do not have a paid plan. You can see it is prompting me to get the paid plan up in the top, right? But let's say I want to import some videos and have it edit these videos together. There is one thing that I find slightly annoying about this product, but it's not that big of a deal. But if I press the import button and let's say I pull in a couple videos from my recent trip to New York. Like here's one from the plane right here. Here's one from inside of a taxi. Here's another from inside of a taxi right here. So I'm gonna select these three and let's import three items. And it brings me to this page here. And you can see there's a slider for manual edit and AI edit. If I try to click on AI edit, I get this error at the top. AI edit not submitted. Video must be in 9:16. A, why? Why can't I just do this with 16:9 videos also? But B, if it only works with 9:16 aspect ratio videos, why let me select them on this page? Why not just gray out the ones that aren't usable and make the ones that aren't usable something that I can select just so I don't get partway through the process and then find out I can't actually do what I want. But because of that constraint, I really quickly recorded two vertical videos of me in my office. One is this like selfie angle shot here, and another is what my camera sees, which you can see in this video here. So I'll go ahead and trim both of these, select those videos, import two items, and now let's switch to AI Edit. And well, when I try to switch to AI I Edit. It says project must only have one clip. So why did you let me select two clips? Okay, I guess you could manually edit these and that's why it lets you select it. But it's just I find this kind of annoying. Like I want to know what I can't do before I get to the point where it tells me I can't do it. Anyway, I'm being very kind of nitpicky here. So let's just unselect this one and we'll pick this 11 second video import one item. And now I can select AI edit. Yay. All right, so we've got some edit styles here. Parallel play is just for kids. It works for adults too. Plant care isn't just about keeping them alive. It's about learning how to listen. I'll select this one down here. I have no idea how it's going to work because there's no speaking in my video. It's just purely a video with no audio. So it may have no idea how to edit this, but let's go ahead and select that one and let's generate the edit. Let's just see what it does if I give it a quick 11 second clip and a style. So yeah, it didn't know how to do that either. It says AI Edit not supported, no speakers or no words detected. All right, so I guess I have to to make a video with me actually talking because it can't edit unless they're speaking. All stuff I wish I knew before sort of getting down the road of the process. All right, so I just got done recording a vertical video on my phone here and let's see if I can finally get this AI Edit to do something for us. Import one item. AI Edit style. Neon Generate Edit. Progress. Progress. All right, so now it's done. And it opened it in this editor and I can hear myself, but I can't see the edited video. Hey, Matt Wolf here. Don't forget to subscribe to me on YouTube. There's gotta be some way to see the video here, right? Download. No, won't let me download without paying. All right, so I finally got it to work, but if I'm trying to record my phone screen to show it to you, working the app makes the video black and doesn't let me share it. So I can't even share this outcome with you other than doing something like this. Hey, Matt here. Don't forget to subscribe to me on YouTube like this video. Do all the things, leave me a comment. And if you haven't already, check out future tools. There's lots of cool tools there. So yeah, pretty annoying that the app won't let me actually share what it generates. And then if I do try to export the video to send you what that video looks like, they want to charge me. So I can't even like demo this without paying, which is unfortunate. But we'll go ahead and move along. We got a new update in the Photoshop beta this week. All of those people, myself included, who said, man, Nano Banana might sort of hurt Photoshop. Well, Photoshop's just embracing it and saying, look, you could use Nano Banana inside of Photoshop now. Now, if you do have an Adobe Creative Cloud account, open Creative Cloud, go to apps, click on Beta and then make sure you're using Photoshop Beta. This is where the new features that they roll out go. They go into the beta version and make sure you have the most up to date version. I need to update Photoshop Beta right now to make sure I get Nano Banana in it. Once it's updated, go and open Photoshop. But again, make sure that you're opening the Photoshop beta. I actually have both installed. We want to be using beta because that's what's got Nano Banana. So I opened this image in Photoshop. You can use this handy select subject feature. If I select this and then just click on the guy, it actually finds the outline of just this guy and ignores everything else. And then I can click on a generative fill here. And then over here there's a button that says Choose Model. If I click on that, you can see we have Options for Firefly Image 3, Flux Context Pro and Nano Banana. So let's go ahead and select Nano Banana and then for our prompt, let's say turn him into a green alien holding a futuristic bow and then we'll generate that. I actually just realized the moon was selected as well, but it did a perfect job of what I asked for. And the nice thing is it did these changes on a new layer. So if I hide this layer, you could see the old layer right underneath it. So I could continue to like, like screw with other elements and it's non destructive. So pretty cool feature to have in Photoshop. Now Haylou AI rolled out some sort of new agent, they say. Introducing Haylou AI Agent, your next gen Creative Flow free for one week. Just talk and watch your ideas come alive on the Infinite Canvas. The agent Selects the best models, plans the workflow, and brings your vision to life. Images, videos, audio, all in one place. So if I jump on over to Haylou AI, Video, slash, agent, this is where you can play around with it. And we can see over here, it does say multimodal creation is coming. So images, videos and audio, all created with one agent isn't actually a feature that's fully rolled out yet, I guess. But we can see here, if I click on model, they've got the Haylou O2 video model. And then under Image model, they've got Nano, banana, C, Dream 4.0, Flux, context, Quinn and Colors. And up here, you can see, you can either select your model or leave it on auto and it will decide the best model to pick for you. And if I had to guess, this is the way most UIs are going to go. So when you use those tools like Leonardo or Kreia or Higsfield or whatever, and you tell it what you want out of it, I imagine in the future they'll all have an auto selector that will pick the best model based on your prompt. But let's go ahead and say create a mood board focused on wolves and the color purple. It says, you've done enough. Now it's my turn. Thank you. All right, smart plan. Create a mood board centered around wolves and the color purple. So it's searching for images featuring wolves and the color purple. Oh, look at this. Okay, okay, so these are image searches. These aren't AI generated. You can see image searched. While this is working, I can actually click in an image, click on full screen and see this image bigger. Now, this does say Minimax Haylou AI. So did it generate this or did it find it? Okay, that's interesting. This one has a Dreams Time watermark on it, which is a stock photo website. And then Minimax and Haylou AI put their own watermark over an already watermarked image. So these are images that it found online. It's just putting its own watermark on images that it doesn't own. Now it says, calling Image Generation Tool to create four images with the same prompt for the mood board. So it's actually now generating images and it doesn't seem to be doing a good job at it. We could see here calling the Image Generation Tool to create four images with the same prompt for the mood board. And it says, image generation completed. I don't know which one it generated, but then down here generating more with an unknown error. Now it's trying to generate again, but we can see a Couple times in a row it says image generation unknown error. So it seemed to be able to find some images. Just. It struggles to generate new images. Here's another one that says Adobe Stock on it on one side and then Minimax Halo on the other. Gotta love just like watermarking images that are owned by other people. All right, so it finally created some other images here based on the mood board. This one even still has a Dreams Time watermark right in the middle of it. Made some sort of like montage collage looking image and some images of like close up on like purple fur. All right, then if I select multiple and then try to right click, it just only selects one, but I could move them around on the board together. Pretty interesting. I don't know the exact, like, best use case for this yet. I guess if you're just trying to build out like a giant mood board and you want all of the images on the mood board to either be searched or generated, then there you go. Not something I personally see myself using, but I'm sure there's people watching this going, I know exactly how I'll use that. I just don't know how I would. And since we're talking about boards and mood boards and vision boards and stuff, there's a new tool from Google called mix board. Kind of the same idea as what we were just looking at actually. And it is in public beta in the US Right now. So if you're outside the US I don't believe you have access to this. But if we go to labs, Google Mixboard, you get a page like this. And if we look into one of their examples here, you could see they made a board of birds with googly eyes. One thing I find annoying is I can't like click and move around. A lot of times you can like right click and sort of scroll around the screen, but that doesn't work. I actually have to use like the scroll wheels to move around. So once you have a bunch of images in a mood board like this or in a mix board like this, one thing you can do is sort of start to blend them together. So let's say I want to take this like peacock here and this red bird here, put them side by side and select both of them. You can see it pulled these images down into my prompt box and say, blend these two, two images. And then pretty quickly down here we can see it's generating another image. And here's our two images blended. I can obviously upload my own images. Theoretically I should be able to generate New images. And it all goes on to this like mood board, mix board thing in Google. Very similar to what we were just looking at, but less of the like agentic features and more of like a design board to drop images on and then combine images in interesting ways. If you're a fan of Notebook LM like I am, they rolled out some new features this week. Flashcards and quizzes are officially 100% rolled out to all users in all markets. So NotebookLM is a platform where you can upload a bunch of, you know, PDFs and websites and a ton of information and then you can create audio overviews, video overviews, mind maps, reports, and then also ask questions about those documents. We've covered NotebookLM a lot in past videos. The two new features are flashcards and quizzes. You can create a flashcard on whatever content is loaded in here and you can create quizzes. Since we've been on this Birds Aren't Real train lately, let's see what kind of flashcards it makes for us. We can see here the first flashcard, what is the Birds Aren't Real movement, and then flip it over. A satirical conspiracy theory movement which posits that birds are actually drones operated by the US government. Satirical. I don't know about that. Who is credited with creating the Birds Aren't Real conspiracy theory? That dude. So you could see for like, school and education purposes, if you were to load a bunch of like, educational information in here, you could quiz yourself with flashcards as well as an actual quiz. So here's what a quiz might look like according to the source material. What was the original catalyst for Peter McKendo's Creating the Birds Aren't Real movement? This one 2017 Women's March where he improvised the theory as a joke among counter protesters. So we can just go through and quiz ourself on all the information. So let's imagine you're a college student and you just took in a bunch of lectures. Well, let's say you did an audio recording of the lecture. We'll take that audio recording, put it into NotebookLM or get it transcribed. Put the transcription in Notebook LM have it create quizzes and flashcards for you and you've got a really good educational resource. Since we're talking about Google, they're also rolling out new AI features directly in Chrome. From what I understand, you have to be on one of Google's paid plans. I can't confirm that a hundred percent, but it does say starting today we're rolling out Gemini in Chrome to Mac and Windows desktop users in the U.S. okay, so maybe it is available to everybody. And it does say it'll be available to businesses in the coming weeks. So this current browser that I'm in is an actual business account. So there's no Gemini available here. You don't see it in the browser. However, if I switch over to the browser where I'm logged in with a personal account and not a business account, you can see up at the top, I've got a little Gemini logo now in the very top right corner. I can also essentially get Gemini to fire up by clicking into our chat bar. And you can see there's a button that says ask Google about this page. Or highlight key points or describe the AI startup's impact. These are all questions that are being generated by AI based on the content that I'm looking at on the website. Or I can click the Gemini button up in the top. What is this article about?

This article discusses how Steph Curry's venture capital firm, Penny Jar Capital, invested in burnt an AI startup focused on improving food supply.

Second, who is Steph Curry?

Steph Curry is one of the greatest basketball players of all time. He plays for the Golden State warriors and is known for his incredible shooting skills.

Okay, okay, okay. So what does he have to do with AI then?

Gotcha. So Steph Curry's not directly involved in the AI tech itself. His venture capital firm, Penny Jar Capital, invests in promising startups.

All right, you get the idea. You can fire up the Gemini button and it will basically read articles for you and let you have discussions. You don't have to do it in voice mode. If I click up here, I could switch it to like a normal keyboard mode as well instead of the voice mode. And you can see that it's sharing the article that I'm on. And you have the option to switch between the 2.5 Pro model and the 2.5 Flash model. Now, I am on the Gemini Ultra plan, which is their ultra expensive plan. So I don't know if everybody has 2.5 pro or not. That might just be on the Ultra plan. I can't actually confirm that, but that would be my guess. The Pro is maybe only on the paid plans, I don't know. But it does kind of seem like Google is going after Perplexity Comet a little bit. And you know the browser company which makes these sort of like AI agentic browsers. I think it's only a matter of time before this Gemini button, you press it and you tell it to do things and it will go and browse on your behalf and click around and copy and paste things and do some of those like browser operation things that we've seen tools like ChatGPT and Claude and and Manus and some of these other tools start to do. Probably going to be rolled out into Chrome pretty soon. And I mentioned Perplexity, so I might as well mention this as well. They now have an email assistant with Perplexity and this is one that I had planned on demoing. However, when I started to look at it I realized I needed a Ultra plan and well, the Ultra plan is perplexity's $200 a month plan. And I'm on enough 200amonth plans already between Google, Claude and Gemini. So yeah, I can't pull the trigger on Perplexity just for an email assistant. But as we can see here, it will automatically like organize your emails into like to respond or notification or an FYI or a meeting update. So it automatically tags. So at a glance you can quickly see what stuff's probably important and what's not. It's got an AI composer to automatically help you write emails to people. It can actually see your calendar and help help assist with scheduling. So it says sure looping in my assistant to find a time. The assistant checks the calendar and suggests times. So you can have the assistant help with sort of booking calls and things like that. And you can get like summaries of like daily meetings and emails and things you need to take care of every day if you want. It looks like a pretty cool email helper tool, but sorry, I can't justify that at the $200 a month plan. Once it's available for the like smaller $20 a month plan then I'll probably play with this. But yeah, looks cool. Too rich for my blood at this point. So far we've covered a ton of new tools, but I haven't put out a news video in two weeks. So there's still a lot of stuff I want to tell you about. So let's jump into a rapid fire.

So the week before last was Meta Connect and unfortunately I didn't make it out to Meta Connect this year. But they did have some pretty cool announcements. Probably the most notable being the new Meta Ray Ban display glasses. Now these are just like the Ray Ban metas that everybody's used to that have the little cameras and the little speakers and the stem of the glasses and they also have built in AI so you can talk to, you know, Meta's llama AI. However, these new Versions, well, you can probably tell by the name, have a display in them. So when you put them on, there's a little screen on just one of the lenses. It's not in both eyes, it's just in one. And it. It's got like, basically no light bleed, so nobody around you can even tell there's something going on on the screen. And then the other really unique feature about these glasses is that they come with this wristband that they call the neural wristband, I believe. And this wristband can actually essentially tell what you're trying to do. So you swipe around inside of the apps on your glasses with like, finger gestures, and if you want to turn the volume up or down, you literally kind of act like you're turning a knob. And it's just got some really interesting hand gestures to control the glasses. I haven't actually tested these myself yet. I've not experienced this neural wristband and how it works. But one thing that I find very interesting is I saw Google make very, very similar glasses, and I demoed them almost a year ago now. And yet somehow Meta's glasses came out before Google's glasses and Google's were the same thing. It was one eye. You can actually see color and see images and things like that in the one eye. And the demo worked great a year ago. So just, I don't know, weird to me that Meta was the first to release these. But you can see text mess directly inside of your glasses. Although I believe it's just WhatsApp messages, so not like actual text messages. Yet you pinch your fingers to take a photo when you ask a question of AI, you can actually see the prompt that was given and then a visual display of the response. But probably one of the coolest features is the ability to translate and have a conversation with somebody that speaks a different language. And essentially, while you're talking to them, you would see little subtitles, like, right by their face. So everything they're saying is automatically getting translated back into English to you so you can see it. They're also rolling out a feature in both these and the normal Ray Ban, that's like a focus mode. So when I'm talking to somebody, if I have these glasses on and I'm looking at them, it'll try to isolate just the conversation you're having and sort of help you tune out all the rest of the conversations around you, which seems like a pretty cool use case as well, especially if you're at, like, networking events or parties where you're really trying to Concentrate on one person speaking. I'm super excited to get my hands on these, test this neural wristband, and probably even do a giveaway on this channel for a pair of these Ray Ban displays. So if you're watching this and you're not subscribed to this channel and you're not subscribed to the newsletter, make sure you do that because I do like to do giveaways from time to time when new cool gadgets come out. Now this is just scratching the surface of what these glasses can do. Those are the main sort of standout features. But we'll put them through the test and play around with them more once. I can actually get my hands on them in a couple weeks here. Meta also showed off some cool stuff around the Metaverse. So there's a new feature in the Meta Quest where you can put the headset on, walk around the room and it will actually map out that room. And then after it's mapped out that room, that becomes a digital space that you can explore later, even if you're not in that room later. That seemed pretty cool. But what I thought was even cooler was their new Horizon Studio. This Horizon Studio, you can actually prompt any world into existence. You can see they gave it the prompt ancient ruins and it created this scene, this environment here that they are then able to move around in, add treasure and trinkets. They give it that prompt and it adds a bunch of stuff. Add mythical guardians, it added the guardians. So really, really cool that you could just constantly prompt new things into existence into this like AI world. It's like kinda holodecky now. The reason I couldn't make it to Meta Connect this year was cause I was in New York at the Made on YouTube event, which is an annual event where YouTube announces all of the new things that they're rolling out. And this year it was a lot of AI stuff. Stuff as a YouTuber, I'm really excited about this Ask Studio. If you're not a YouTuber, you probably don't care, but basically it's an AI bot that lives in the back end studio of your YouTube channel, where you can ask questions about your analytics and get suggestions on future videos and things like that. It's just an AI bot that understands your data and your stats and your channel that you can chat with. They're rolling out a B testing to test titles. You've already been able to do it with thumbnails, now you can do it with titles. Not AI related, but pretty cool. If you're a YouTuber so YouTube has been doing auto dubbing for a little while now. On some channels my channel has had it where I speak all of my videos in English. But if you're in India and you watch one of my videos, it will automatically try to translate it for you. By default, if you're in Mexico and you try to watch it down there, it will automatically try to translate it to Spanish. But it doesn't look like I'm speaking, it just looks like it was dubbed over with another voice. Well, now YouTube is rolling out auto dubbing with lip sync. So it will actually in the future look like me and my lips moving and speaking the language that you're hearing with my voice. That's going to be wild. Now this isn't fully rolled out yet, but this is going to be really, really cool. They're also rolling out a likeness detection because, well, AI Deepfakes are getting easier and easier and pretty much anybody can do them now. And well, if you don't want your likeness being used on YouTube, even if it wasn't really you, well, they're building technology to, to better detect that and stop people from stealing your likeness and like deep faking you into videos. A few of the other random AI features they're rolling out, you can turn everyday videos into catchy soundtracks. So if you hear like a funny phrase, a memorable quote or a one of a kind sound, they have a speech to song remixing tool that uses Lyria 2, which is DeepMind's most advanced AI music model. And it will convert whatever sort of sound bites or, or quotes or whatever you give it into a music video that sounds something like this.

Isn't it almost cuffing season?

I know you're looking for that new date spot. Isn't it almost cuffing season? Isn't it almost cuffing season? Isn't it almost cuffing season? So, you know, a new random meme generator and as we predicted and talked about in previous videos, they are rolling VO3 directly into YouTube shorts. So you can see they've partnered with DeepMind to bring a custom version of of VO3 to YouTube and they're allowing users of the platform to use it for free directly in shorts. And right now it's rolling out in the U.S. uK, Canada, Australia and New Zealand and plans for more countries after that. I'm not a huge fan of this feature. I think most of the stuff I made on YouTube I was really excited about and pumped up to try out in the future. This is one of the things that they announced that I went I don't know if that's gonna be good. Is not just going to make it. So our Shorts feed is all sort of AI generated slop. I don't know if you make it that easy for anybody to just press a button, generate AI slop and push it to the feed. I don't know, that seems like we might get flooded with that. We'll see. We'll see. Overall, I was a huge fan of the made on YouTube event and most of the announcements they made. But yeah, the VO3 in shorts, not something I'm super excited about. And along the same lines of AI slop, apparently being an AI musician can lend you a record deal. Oh, somebody kill me, please. So this AI artist named Zhonya Zanye Monet climbs the charts and signs a multimillion dollar record deal. So this is somebody who actually uses Suno to turn their own lyrics. So they write their lyrics themselves, I'm guessing, into music. And the label offered them $3 million to sign this artist. And since we're speaking about leveraging AI to make money, I thought this would be fun to talk about. So there's an app called Neon and you can find it in the App Store, I believe, of both Apple and Android. And it's an app that will pay you to share your phone calls. So basically, here's how it works. So let's say I'm calling my mom and I call her through this app. It will record those conversations and send that conversation into some sort of data center to train on that conversation. And the person who made the call maybe makes about 30 cents per minute and can max out at about $30 a day by doing this. It doesn't record your calls if you're just using the normal phone app. You do have to use this specific app and you have to have consent from the person on the other side and they have to be using the same app as well. If you call somebody with this app and they're not using the app and they haven't given consent, it's only going to record your side of the conversation. If you're both using the app and you both gave consent, it will record for both sides of the conversation. And again, all of that conversation goes in as training data that this data company can then sell to the AI model creators. And you earn money by basically allowing this app to spy on your calls and use it for training data. And you might say, well, who would do that? Who would want to use an app that records and sends all of their conversation data to like big tech companies. Well, as of this recording, it's the number 7th most downloaded app for iPhone Phone if you're on an Android phone, well, you now have Nano Banana editing directly inside of Google Photos. If you're a fan of Adobe's Firefly Boards, well they just rolled out two new AI video models into their platform. You can now use Runway Alif and Moon Valley's Mary or Marais directly inside of the boards app. ElevenLabs rolled out an update this week called Studio 3.0. They claim it's their most advanced AI audio model and it now has video support. I haven't really put this one to the test yet, but I guarantee we'll be using 11 labs quite a bit in some future videos. Also, this is pretty cool. Automatic captioning, speech correction for real life recordings, and multiplayer commenting. So it looks like you can collaboratively edit stuff in 11 labs with other users if you like. Generating Music Suno just rolled out V5, which is their new state of the art music generation model. I went and tried to generate some new songs with v5 and compared them to version4.5 and on honestly had a pretty hard time telling the difference. But we will put Suno through its motions. We've got some big ideas of ways we want to test Suno in some future videos, so make sure you're subscribed if you want to see that. And since we're talking about Suno, they also launched their own daw, which stands for Digital Audio Workstation. It's basically your audio editor. So there's a brand new like Audio Timeline editor directly inside of Suno. Again something we will be playing with a little bit more later. But it sounds like like you can generate songs and then get really, really granular with the editing of those songs. So that sounds pretty cool. Now over the last couple of weeks there was a handful of news in the large language model world. I've sort of backed off on trying to explain the benefits and the differences of these new models because quite honestly it all feels so marginal to me and to the majority of the world, the improvements in these LLMs doesn't really make a difference unless you're, you know, a developer or somebody that's really, really, really putting these models to their limits. But here's what we got over the so we got a couple new models out of Google Gemini 2.5 flash and flashlight. Now these are models that have been available but they made some improvements in the quality and speed to them so that they're a little bit more performant. Xai rolled out Grok4Fast, which is a faster but probably slightly dumber GROK model, but actually performs pretty dang close to Grok 4 in almost all of these benchmarks they tested on. We got a couple new models out of Alibaba. We got Quinn 3 Max, which you can actually play with for free over@chat dot.quinn AI apparently this one's really good at coding and agentic skills. We can see the red bars here as it's compared to Quinn three big long number Ed Claude Opus four and deep seek v3.1. And it pretty much beats all of those in all of them except for the SWE bench here, basically showing that it's pretty good at code. We also got Quinn3 omni out of Alibaba. So this is an omni model that can manage text, image, audio and video all in one model and they apparently open source this one as well. OpenAI is trying to build in some age checks into the system and it sounds like it's going to try to predict their age based on the conversations they're having, not by some sort of like age verification form or something. It says down here even the most advanced systems will sometimes struggle to predict age. If we're not confident about someone's age or have incomplete information, we'll take the safer rate route and default to the under 18 experience. So if you talk like somebody that's under 18, it might guess that you're under 18 and give you the kid friendly chatgpt. But they did also add parental controls where parents can actively take control over what kids see when they're using ChatGPT. And since we're talking about OpenAI and ChatGPT, there is some other features that rolled out. You can now share your projects with other team members. I guess the various connectors that are available are slightly better. Like they pull in more details via the API and they've made some new compliance and security updates. So you know, not major updates. But if you are working with a team, having shared projects sounds pretty handy. The company Factory AI this week reached number one on Terminal Bench, which is the most challenging general software development benchmark. So it actually outperforms tools like Claude Code and codex CLI from OpenAI and they even made it so that you can use the Factory AI platform platform with any interface. So you can use it directly inside of like VS code or Windsurf or Cursor or whatever IDE you're using or from places like Slack. Anyway, this is one that I just wanted to Share Because I'm an investor in Factory and I'm excited about this news. They're making some pretty good progress. I thought this story was really interesting. Microsoft 365 copilot just added anthropic models to it. So now if you're in something like Excel, you can actually choose whether you want to use one of the GPT models or one of the anthropic models.

What's going on in there? What? And the reason this is interesting is because, well, Microsoft owns 49% of OpenAI's not for profit side of their company and with Microsoft owning such a big chunk of OpenAI, you would think they would be all in on OpenAI, but nope. Microsoft is really diversifying and even allowing you to use competitors like Anthropo Topic inside of their platforms. And finally, you will soon be able to attend Zoom meetings as an AI avatar. So yeah, you'll be able to upload an AI avatar of yourself, not be on camera and let your AI avatar do the speaking for you if you don't feel like putting clothes on or something. I'm here live, I'm not a cat. And that's what I got for you today. I know it was a lot. This was a fire hose episode and well, it's because I skipped a couple of weeks and quite honestly that's probably going to happen a little bit more often as well. We've got some really, really, really big ideas for videos that really put these tools into action. And so you'll probably see some different kinds of videos starting to roll out on this channel where we really, really figure out the limits of some of these software. We put them to the test, we put them out there into real world use cases and we try to find new, interesting, useful ways to use a lot of these AI tools. And it's a new style of video we're testing and as you we work to produce more and more of those videos, well, the news videos may be further in between, but I promise I will still keep you looped in on all of the biggest, most important news in the AI world that you're going to want to know and probably cut out maybe some of the fluff that might have made it into some of these news videos. So stick around if you're interested in learning real practical use cases for AI tools and learning how you can actually implement them in your business or your real, real life. We've got exciting stuff coming up so you're going to want to stick around for that. Make sure you like this video you subscribe to this channel that'll make sure more videos like this show up in your YouTube feed. Thank you so much for tuning in and nerding out with me about all of the latest stuff that came out in AI over the last two weeks. Hopefully you feel more looped in and that's my goal here is to keep you looped in and learning how to use this stuff. So thanks again for hanging out. Really really appreciate you. Hopefully I'll see you in the next one. Bye bye. Thank you so much for nerding out with me today. If you like videos like this make sure to give it a thumbs up 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.