Video eLP3ag0YpyA

[00:00] In just one day, this year's Google I.O. keynote has accumulated more views than any other previous Google I.O. keynote ever. But the question is, were millions of us tuning in out of excitement, or was it out of existential dread?

[00:16] See, only 0.2% of the people who watched smashed that like button. And to put that number in context, this is one of the most hated videos that we've ever uploaded to our channel. It still managed over 27 times the positive engagement.

[00:32] However, none of that means that the tech that was on display wasn't objectively very impressive. Google showed off everything from AI-assisted search that can create bespoke animations to help explain complex subjects

[00:45] to a new intelligent shopping cart that can monitor for product availability and deals across the known universe and even go as far as to just buy things for you within the parameters you set. I think the one that blew my mind the most, though, was when they used Antidravity 2.0, backed by their new Gemini 3.5 Flash model, to break down and tackle the incredibly complex task of creating an operating system that they then used to run Doom live on stage.

[01:13] Oh, and then there's Omni, which is Google's new super multimodal world model that can ingest basically any input, text, pictures, videos, whatever, and then turn it into basically any output you want according to your prompt.

[01:27] Google's vision for the future is that at every touchpoint, search, Android, even YouTube, where AI search is rolling out right now, they're going to combine their AI's general knowledge and context with all the data they've collected about you personally and your life and then turn pretty much everything they serve you into a personalized experience.

[01:49] I mean, why search when you can just talk to your AI? Why download an app when you can just have your AI create and deploy one for you? Why Segway when the AI can tell you about our sponsor?

[02:04] A man obsessed with computers, yet overlooking the most comprehensive vehicle combat games that can be played with them. Over 2,500 tanks, planes, helicopters, ships, and 95 million others playing on consoles and mobile too.

[02:20] The question that remains is if he or you will join them for free in War Thunder.

[02:36] At its core, Google's business is still built on search. So why don't we start there? Google says that their new Gemini-infused paradigm for search is meant to be less like individual queries and more like a conversation.

[02:50] So instead of searching troubleshooting graphics drivers and then reading through an article about how to do said thing, you might just say, my GPU doesn't work. Can you help me figure out why? And then a back and forth ensues involving text instructions, YouTube video links, or maybe even custom visualizations.

[03:08] Not going to lie, that actually sounds pretty cool, at least on the surface. And, according to Google, its contextual awareness of your preferences are only going to grow over time. But, like, does this kind of just feel like the AI version of hitting the I'm feeling lucky button?

[03:25] Where, instead of looking through all of the various sources of information to find credible, authoritative voices, we're just kind of trusting whatever the AI throws at us? Google isn't even hiding all of the shopping partnerships they have that create, in my mind, a pretty big conflict of interest.

[03:42] I mean, can I really trust that the activities that my Gemini-powered search recommends for this weekend aren't sponsored? And on the subject of where I'm going, where is Maps going to send me?

[03:54] Because, of course, this feature is coming to Maps as well. I think one of the creepiest moments in the entire presentation was when Google CEO Sundar Pichai showed what was apparently a real prompt that somebody put into Google Maps asking,

[04:08] My kid just fell into the duck pond and the wedding starts in 30 minutes. Where can I walk and buy her a new dress? Again, objectively impressive technology. But does Sundar just have a folder on his phone book with funniest prompts from real users that he can just browse in plain text?

[04:24] Well if he does guess what You can contribute to it today because Ask Maps with more natural language interaction is starting to roll out now And the same goes for Ask YouTube which I gotta say honestly looks pretty cool

[04:38] Instead of relying on the almighty algorithm to surface what you might want to watch, Google is overhauling search, finally, to surface context-aware videos that will focus on exactly what you want to do when you want to do it.

[04:51] The example they gave was teaching a kid to ride a bike with an emphasis on the exact stage that they were at in the training to try to surface the most relevant videos. I mean, I don't know, I managed to teach my kids to ride a bike without AI,

[05:03] but if we're being real, we've probably all seen some parents who could probably be replaced by AI. An AI that's going to be more powerful than ever if Gemini Spark has anything to say about it. Spark was a major undercurrent of the whole keynote

[05:17] with integration across multiple Google products and maybe finally the promise of actually becoming the assistant or agent that Google has been trying to build for years.

[05:29] But before I can talk about what Spark does, let's do a quick primer on what is an AI agent or agentic AI. Now, I could have written this, but it seemed on theme to just ask Gemini, so let's see how close it got.

[05:43] Identic AI represents a major shift from reactive, prompts-driven catbots to proactive systems that possess agency, the ability to independently reason, create multi-step plans, and execute complex digital workflows to achieve high-level goals.

[06:00] Rather than just generating text, these autonomous systems continuously perceive their environment, break massive objectives down into smaller sub-tasks, interact with the external world using APIs and software tools,

[06:12] and dynamically adjust their strategy when they encounter errors. Whether operating as single agents or collaborating in specialized multi-agent teams, Adentic AI shifts the human role from tedious execution to high-level supervision,

[06:28] transforming industries by independently handling intricate pipelines like software engineering, customer service resolution, and business operations. Thanks, Gemini. Very cool. I see you've learned to glaze yourself as hard as you glaze the user.

[06:41] That whole last sentence was probably unnecessary when I asked for a simple definition, but hey, always be closing, right? Also, hey, since we're chatting, I noticed you've gotten a facelift. You're right, Linus.

[06:53] It's called Neural Expressive. Sorry, I'm going to stop doing that now. It sure is called Neural Expressive, and actually it looks pretty fancy with some new animations and typography, which has already rolled out. But the facelift is not the impressive part.

[07:05] The impressive part is that Gemini Spark is integrated into, like, everything Google. And if the demos are anything to go by, this is by far the most polished and useful

[07:17] agentic AI that we've seen that can not only browse the web and look stuff up for you, but check your calendar and consolidate information from your emails and check your photos and your documents and bring all of that together to tackle complex and get everyday tasks.

[07:33] I think the most useful demo that they probably did for my daily life was having Sparks go through all of the separate email chains for a block party and then consolidate the guest list, our SVPs, and even what dish everyone is bringing into a Google Sheet. Like, what a time saver.

[07:47] To be clear, all of this can be done today with tools like OpenClaw, but Google's innovation is to make it easy and to integrate it into the tools that you were already using. That is something that is going to be hard for the other big players to compete with, especially when you consider Google's

[08:02] partnership with Apple. Now let's talk about the big updates to video and image generation, courtesy of Gemini Omni, not to be confused with Omni-Man. Omni boasts better than ever multimodal capabilities and an awareness of the world that should result in improvements

[08:17] to everything from spatial consistency to physics simulations and even gravity. And once again, the demos were powerful. See, this is where I would normally use superlatives like cool and amazing, but by this point in the presentation, the pit in my stomach had gone from

[08:34] solid pit to, like, sharp pit. Anyway, they showed that just plonking any media into Omni and then asking it to do stuff would result in that stuff done. Like, here, okay, look at this. This is a

[08:48] single shot Okay Omni give me 12 angles Also while you at it make it at night Like what Then there this one Here a riff Can you make it sloppier Okay that one wasn that impressive but I think the coolest part was how well it handled complex multi instructions

[09:08] So you could just say, here's a video, and I want the background to look kind of like this image, and also, can you clean up the audio? The hope here is that this means no more writing 10 prompts and then, like, mashing clips together, together, allowing you to do everything in one place and maybe even all from one initial

[09:23] prompt and then with some refinement prompts. Pretty neat. Moving on to Docs Live, we've got another tool that's both useful looking and depressing.

[09:35] Instead of opening up a Google Doc and adding some notes, you can just tell Gemini to summarize your thoughts in one place and then add whatever you want while you're at it. Here, let's try it. Hey Gemini, can you plan all of LTT's 2026 content for me?

[09:49] Please include a table of our best performing videos so far this year, and why not add a few presentation slides telling me that I'm the best? Can you do that for me, please? No, I'm joking right now, of course, but in the future,

[10:02] I paused the video so that I could actually read and look at some of the quality of these outputs that they were doing live on stage, and if they were honest about that, I guarantee you that as we move forward, many of your favorite creators will be doing just that.

[10:17] I mean, it's amazing how much AI they're building into YouTube. This wasn't part of I.O. today, but YouTube's relatively new Catch Me Up On This Video tool spits out shockingly relevant suggestions for, like, follow-up content, for instance.

[10:30] Moving on from that, though, let's talk about shopping! Google's intelligent shopping cart envisions a future where you can just browse any store, anywhere on the web and just add an item to like a universal shopping cart.

[10:44] And while they're at it, why bother using a store? Why not just highlight an image of something and ask for it to buy it and then throw it in your cart from there? Taking things to another level. Are you tired of bots beating you to those floor-level tickets to your favorite events?

[10:59] Well, Google says, if you can't beat them, join them. Because now you could say, hey, Gemini, can you buy me tickets to the upcoming Angeen de Patrine concert? as soon as they go on sale, and they'll scan the web in the background

[11:11] and let you know as soon as it's ready to pull the trigger. They also boast that it can save you time and money by looking for deals proactively or analyzing your cart for incorrect items like this CPU and motherboard, and then even placing orders on your behalf.

[11:26] Will I give my credit card to Google when it rolls out then? Absolutely. Moving on, it's time to talk about your Uncle Space Time's life. And to gravity. Okay, maybe that joke didn't land. But it makes sense, because there was no gravity.

[11:39] Okay, it doesn't matter. The point is, considering that I dipped my own toes into Vibe coding not that long ago, I was really impressed by the Anti-Gravity 2.0 demo, where they used it and Gemini 3.5 Flash to code a fully armed and operational operating system.

[11:56] Now, to be clear, we have no idea if it actually just, like, copied the Linux kernel, or if it actually wrote any new kernel-level code. But what it did do was, well, nothing when they tried to launch Doom.

[12:09] That is, until they asked it to fix the issue, and then a while later, graphics and keyboard drivers had been obtained, and the problem was solved, and the demons were ready to die. Impressive. Most impressive.

[12:22] But, like, if there's anything that I learned from my vibe coding adventure, it's that I'm not actually a developer, and I have no idea how to QA code or maintain code in the longer term. So, to put this very lightly, as cool as that was, I am a little worried about how we're going to verify all this Vibe coding that Google has seemingly fully leaned into, even in production applications.

[12:47] I mean, they boasted on stage that they created this Mac application that got 100 features in 100 days. You are not going to convince me you did that without Vibe coding.

[12:59] But wait, I know I've got the solution to move you to AI to do the coding and do the QA work for us. Then we can replace both the developers and the QA staff. Then we can use all that money to buy more AI coaches.

[13:12] This is genius. Jokes aside, as someone with aging relatives, there was at least one thing in this two-hour-long slog of a keynote that gave me a warm and pleasant feeling. And that was the talk about SynthID and C2PA According to Google even now people can only detect AI videos about a quarter of the time That is a terrifying number and that only going to shrink as it gets better

[13:36] But now that the SyncID Watermark feature is expanding in partnership with OpenAI, 11 Labs, and Kakao, hopefully things will become a little more verifiable. As for C2PA, they're adding verification to those content credentials

[13:51] so that next time your grandmother sends you a video of crabs wearing sunglasses at an underground rave, you can tell her, hey, that was either fully made with AI or it was edited with AI. That is, as long as the content was generated using participating tools.

[14:07] On the subject of crabs with glasses, something that I have extreme experience about is mixed reality glasses. I've been playing around with the display model from Meta, and objectively, they are very powerful.

[14:20] that word again. And also hardcore dystopian. Coming back to Google I.O. though, it's kind of funny because for Google's new product, they didn't really lean into what are some of the more obvious benefits to me,

[14:32] like at all. Like, what about helping with visual impairment? In theory, you could have these on your head, they could be powered on, and you could just ask them stuff and get AI-powered answers straight into your ear. I mean, I don't know about you,

[14:44] but I think someone who can't see asking if it takes across the street or if that's orange juice here in the grocery store, is infinitely more important than asking to take a picture from the stage

[14:56] and then sloppify it live. But hey, it's a Google Keynote, not an Apple one, so they've got to stay on brand, I guess. Oh, another update was the improvements to Google's TPU, or Tensor Processing Unit.

[15:09] Or should I say units? They're going with a dual-chip approach with this one. The AT is going to be specialized for training, And the other, the 8i, is specialized for watching you everywhere.

[15:21] Sorry, the i is actually for inference. It's good for inference. Both of these chips got some decent generational performance and efficiency improvements, which is neat. But the real jaw-dropper was their talk about seamlessly distributing training across multiple data center sites,

[15:36] helping to overcome the power and performance limitations of individual data centers. This gives Google the largest Shonen cluster in the world using technologies like Jaxx and Pathways, bringing us, yes, one small step closer to the singularity.

[15:51] Not for us, of course. Just for the machines. And if you weren't scared already, I've got good news, everyone! All of this stuff is either already out, launching this summer, or coming later this year.

[16:04] And one more thing. If you're already subscribed to Gemini Pro or Ultra, those are both getting price cuts, So you're getting more eyes for less. That is at least for now, because we've all seen how this works.

[16:16] You get the massive hurt, then you dig those claws in, and you keep them in your ecosystem forever, and you jack up your prices, just like how I'm jacking up this thing like this. So I thought... Portrait of a tech YouTuber.

[16:31] Surrounded by powerful hardware, but the rest of entertainment. Regrettably oblivious to his own need for the tactile sense of joyous destruction that comes with the grip of a jet's joystick in his palm.

[16:43] Yet, transplant his mind into a simulation like War Thunder, the most comprehensive vehicle combat game, and allow him control over 2,500 of humanity's most deadly machines from the past and the present,

[16:56] and he may rapidly become aware of the yearning in his core for destruction. This game charges no price of admission. The only sacrifice it asks of its players,

[17:08] the time they'd previously spent on more superficial experiences. Simple health bars and hitboxes are no place in War Thunder where high-fidelity graphics and authentic sounds are a gateway to the heat of battle.

[17:22] But the limitations of players are just on the other side of a door to be unlocked with a key in his own imagination, one that may imagine playing on PC, consoles, and mobile, and find it to indeed be true, as 95 million others have.

[17:36] A bonus pack of vehicles, boosters, and more awaiting players and those returning after six months as they await you if you use the link below and join them in War Thunder.

[17:49] If you guys enjoyed this video, why not check out the Android event video from just last week. It covered some of the legitimately cool stuff, gotta admit, that's happening on the Android side as well.