0:02 So Google Genie is here and it literally
0:03 just broke the stock market. So let's
0:07 talk about it.
0:08 So Google Genie is Google's latest
0:10 playable world model that lets you
0:12 literally design an environment and a
0:14 character and then you can literally
0:16 create your sketch, create your world,
0:17 and you can just immediately jump in and
0:19 play whatever game you want. Now, this
0:21 hasn't broken the internet, and it did
0:22 break the stock market, which we'll talk
0:24 about later. But this is definitely a
0:26 really cool piece of technology, and a
0:28 lot of people are looking at like, you
0:29 know, what is this? What does this do?
0:30 what what do we want to do with this?
0:32 And I mean, I don't think you guys
0:33 should, you know, dive that deep into
0:34 it. I think this is a really cool piece
0:37 of technology and the many use cases
0:39 will become evident in the near future.
0:40 World models are something that are
0:42 remarkably useful because they allow,
0:44 you know, AI agents to potentially train
0:46 in virtual worlds and that's something
0:48 we're going to touch on later. But
0:50 essentially, this thing just allows you
0:52 to use any image where you can literally
0:54 just interact with your environment, not
0:55 your environment, in the environment or
0:57 whatever character you do want to do.
0:59 And this is super interesting because
1:01 you have a vast array of different, you
1:03 know, environments that are possible.
1:05 And it's super super intriguing to see
1:07 what individuals cook up with with all
1:08 of their custom images, with all of
1:10 their, you know, very, very thoughtful
1:12 ideas. And this is something that has,
1:14 you know, previously we thought Google
1:16 would never release a product like this.
1:18 I mean, currently now, currently now,
1:19 this is something that is actually
1:21 available for ultra users. So, if you do
1:23 have the highest subscription tier on
1:25 the Google AI plan, you are actually
1:27 able to use this. and it doesn't even
1:29 take any AI credit. So, this is
1:30 something that once again we didn't
1:32 think that we would get so soon, but it
1:34 does seem that this has become, you
1:37 know, a remarkable way from the Genie 1
1:40 days. Now this thing is really really
1:42 interesting because I don't know how
1:45 much compute it does cost to run this
1:46 but I cannot imagine that this would be
1:49 cheap to run because it is a real-time
1:50 world model and we already know that
1:53 video models actually do cost a very
1:55 large amount of compute in order to use
1:58 them. So I can't imagine that this is
2:00 not going to be any somehow different.
2:02 And the interesting thing about this is
2:04 that like you can only play this for 60
2:06 seconds. So, if you do want to actually,
2:08 you know, use this and try this out, you
2:10 do have to be in the United States of 18
2:11 years of age. And as I already said, on
2:13 the Ultra plan, but like I said, when
2:15 you're playing with this, they only give
2:17 you 60 seconds. So, they do cap your
2:19 compute time. And I am guessing that you
2:21 probably aren't able to play this all
2:23 day. And there are probably usage
2:25 limits. So, as this does go more viral,
2:26 I'm guessing that maybe there probably
2:28 might be even more usage limits, you
2:30 know, implemented into this entire
2:32 thing. Now, I really do wish I could try
2:33 this. I actually did release a video
2:35 yesterday on how you can actually, you
2:37 know, dive into this. I've looked at all
2:39 the flaws and everything that this
2:40 software offers. So, if you do want to
2:42 watch that video, you do have access.
2:43 You are a Google AI Ultra subscriber.
2:46 You can watch my previous video. But, we
2:47 need to dive into just some of the flaws
2:49 of this. Not to say that this is bad at
2:50 all, but just so you guys understand
2:52 what's going on. And when I dive into
2:54 the flaws, I'm going to show you guys
2:57 why it was so crazy that this thing
2:59 managed to crack the stock market. Now,
3:00 of course, before I dive into all of
3:01 that, I want to say congratulations to
3:03 Google because this is a tremendous
3:05 feat. I haven't seen any other company
3:07 manage to do something like this. And
3:08 when I'm talking about companies, I'm
3:09 talking about, you know, some of the big
3:12 four like OpenAI, Anthropic, X.A.I.,
3:14 Google's the only one that seems to be
3:16 pushing world models as forward as they
3:17 are. Now, there are some other companies
3:19 that are doing world models, but they
3:21 aren't actually on the level of Google,
3:23 so I won't talk about them just yet. So,
3:26 Genie 3 is super super intriguing. Now,
3:27 if you're wondering, is this some
3:29 completely incredible simulator that's
3:32 completely, you know, perfect? No. I was
3:36 reading this uh tweet from S SWYX, and
3:37 he said, I finally paid for a Gemini
3:39 Ultra subscription, and here's an
3:41 unsolicited review. It has obvious
3:42 flaws, but it's a real-time playable
3:44 video world model. And he says, "The
3:46 flaws are that it clips through a lot of
3:47 terrain, which is particularly worse
3:49 than a game engine. If you know game
3:50 engines, they usually have hard physics.
3:52 So when you walk up to a wall, that wall
3:54 is usually completely solid. So you
3:56 weren't able to be passing through that.
3:57 And usually if you do clip through a
3:58 wall, those are like glitches in the
4:00 game. And you do ideally want to remove
4:02 those glitches. And he also says, "Only
4:03 too late do I discover that you can use
4:05 space bar and jump on things." And he
4:07 says it sometimes does error out. You
4:08 can only live in the world for 60
4:09 seconds, which is like what I've
4:10 discussed. And he says it feels like
4:12 some Black Mirror storyline. And he says
4:14 that nothing else moves, ruining your
4:15 immersion. Why does every model lack
4:18 physics? And that is somewhat true. I've
4:19 seen some worlds where some cars move
4:21 and some worlds where things are
4:22 completely static. I do think it is
4:24 dependent on your complete prompt. We
4:25 will have to see how that is from time
4:27 to time. And he does say prompt to
4:29 prompt editing can have bad side effects
4:31 like adding a thing removes a previous
4:33 thing you like. So yeah, this this is
4:34 the problem with generative AI is that
4:36 generative is not you know binary. It's
4:38 not zeros and ones. I mean it is but
4:40 it's not like you know complete
4:41 deterministic piece of software. It's
4:42 completely generative meaning that you
4:44 don't know what you're going to get.
4:46 It's very probabilistic. So, of course,
4:48 I think that is a good thing in some
4:49 instances, but of course, it's a bad
4:51 thing in some instances because it means
4:54 the replayability is pretty rough. And
4:55 yeah, I mean, those are the only flaws,
4:56 but I got to be honest, I do think that
4:58 they probably will fix the majority of
5:00 these. It will be interesting to see how
5:02 this software evolves. Now, here's where
5:03 we get on to the crazy stuff because
5:05 this was what was going on on Twitter.
5:08 We had Mini say, "Holy moly, Genie3 just
5:10 created this mock 3D game mode from
5:12 Breath of the Wild. How I did it and
5:13 prompts in the comments." Now, I wanted
5:14 to show you guys this, you know,
5:16 screenshot, but I do think that this was
5:18 removed from copyright by Nintendo. I'm
5:20 not sure if that is true, but I know
5:21 that Nintendo copyright stuff all the
5:23 time, and I did see something on his
5:24 account about something getting
5:26 copyrighted, but the long story short is
5:28 this gameplay right here. It looked
5:30 really, really impressive. Okay, now I
5:31 want to show you guys this. Okay, so
5:34 this is uh Genie 3 and this looks like
5:36 uh I think it looked like Dark Souls.
5:38 Okay, and this is Genie 3 and it does
5:40 look like gameplay. Okay, now here's the
5:43 problem. Okay, that happened and it was
5:45 just crazy. So, the problem is is that
5:47 these gameplay videos were gaining quick
5:50 virality. Now, as you know, some people
5:52 when virality does happen in AI and on
5:54 social media, some people just hop on
5:56 the bandwagon with lies. One of those
5:58 lies was this. So, so there was this
6:01 video which was fake news. So, someone
6:03 tweeted out, you know, this Zuma guy
6:05 tweeted out, "Oh my god, this video game
6:06 was made in minutes with Google AI.
6:09 Short all gaming studios. It's over."
6:11 Now, this gameplay video here isn't
6:14 actually Genie 3. This was an actual
6:17 footage of GTA 5. And then what the
6:18 person did was they used some software.
6:21 I think it's runway video to video where
6:23 they converted GTA V gameplay into
6:25 realistic gameplay. This happened almost
6:26 a year ago now. And you can, you know,
6:29 look up GTA V gameplay on YouTube. Um,
6:30 and type in realistic Call of Duty
6:32 gameplay, realistic GTA V gameplay,
6:34 you'll see that this gameplay does exist
6:35 on that website. And so, you might be
6:37 thinking, okay, well, a piece of fake
6:39 news went viral. The problem is, okay,
6:41 the problem is is that this fake news
6:44 went viral to the point that it did not
6:45 crash the stock market, but crashed a
6:48 bunch of gaming related studios. Now,
6:49 it's crazy because they said short all
6:51 gaming studios, it's over. And then you
6:53 can literally see here video game stocks
6:54 are suddenly crashing today with the
6:56 launch of Google's Project Genie as
6:58 investors think games will start be
7:00 being made with AI. You can see that uh
7:02 major major stocks, I think it was Take
7:04 to Interactive and a bunch of other
7:07 gaming studios, you can see Roblox um
7:09 It's it's crazy. Like genuinely, this is
7:11 crazy. Okay, now this is a textbook
7:14 example of emotional panic-driven market
7:16 behavior because think about it.
7:17 Investors are seeing that, okay, AI can
7:19 make video games and they're immediately
7:20 just hitting the sell button and that's
7:23 a, you know, 8 to 20% drop across the
7:25 board. But this ignores the obvious
7:27 realities of the, you know, gaming
7:30 industry. Okay, Nintendo, take two.
7:31 These are companies that are not sitting
7:33 around helpless. These, they have the
7:35 resources and if anything, AI tools are
7:36 probably going to make these companies
7:38 more valuable. And remember that IP and
7:40 communities matter. People aren't going
7:42 to stop playing Mario, GTA, and Roblox
7:45 just because someone can now generate a
7:47 basic game with AI for 60 seconds. You
7:48 have to understand established
7:50 franchises have decades of brand equity,
7:51 polish experiences, and lower
7:54 communities. I don't believe that even
7:56 if G3 gets marginally better, people are
7:58 going to be playing this over GTA 6 when
8:00 it releases. So that kind of stock price
8:03 decline doesn't make sense. Like
8:04 genuinely, if you've ever played a video
8:06 game, I know some of you guys might not
8:08 be gamers, but it doesn't make sense.
8:09 Okay? And the worst thing about this is
8:11 that this is such a crazy timeline dis
8:14 disconnect that even if Project Genie
8:16 does manage to disrupt gaming in a big
8:17 way, and that's a big big, you know, if
8:19 if it does that, this isn't going to
8:20 happen tomorrow. This is going to take a
8:22 very long time for that to happen. And
8:24 you have to understand just how complex
8:26 the gaming industry is. There are so
8:28 many good games out nowadays. To be able
8:29 to actually get a game that people like
8:31 and enjoy, I mean, this is, you know,
8:33 pretty crazy. So, I think once again,
8:35 this is just human human fear. I'm not
8:36 trying to give you guys any sort of
8:38 financial advice, but I just think it's
8:40 so interesting that a that Gen3 being
8:42 released by Google just erased millions
8:44 or even billions of dollars from some
8:46 stocks, okay? Just to show you guys how
8:49 crazy this, you know, AI hype is. Now, I
8:51 think the main use case of this, okay,
8:52 and most people probably won't think
8:54 this is the main use case. This is my
8:57 honest opinion, is that Gen3 is probably
8:59 huge for robotics. So most people don't
9:01 realize that you know it allows
9:03 simulated training at scale. Robotics
9:04 has always faced a data problem which is
9:06 one of the biggest problems they they're
9:08 facing and it's data collection. Okay.
9:10 So data collection for robotics is is
9:12 expensive. It's time consuming and it's
9:13 risky to train robots in the real world
9:15 because they can make mistakes, they can
9:17 break all that stuff. But Genie 3 can
9:19 generate unlimited diverse environments
9:21 from check prompts. If you want to train
9:22 a warehouse robot, you can generate
9:24 10,000 variations of warehouse layouts,
9:26 lighting conditions, and obstacles. And
9:29 unlike static 3D environments, Genie3
9:31 does simulate physics and interactions
9:33 in real time. So the robot can
9:35 experience how objects move, how
9:36 surfaces respond, contact to gravity,
9:38 all of that stuff. And that's critical
9:40 for manipulation tasks. So if you want
9:42 to see if your robot's going to fail in
9:43 certain environments and in certain edge
9:46 cases, this is basically the perfect,
9:48 you know, kind of thing. So I wouldn't
9:49 be surprised if they kind of maybe
9:51 licensed this software, some kind of
9:53 robotics testing round rather than an AI
9:55 generated video game software. So, for
9:58 me, I think this is probably going to be
9:59 the most interesting application of
10:01 this. I don't know. I could be wrong. If
10:03 you're a robotics researcher down there,
10:04 of course, let me know down below. Leave
10:06 a comment in the free community, but I
10:08 do think that this is one of the, you
10:09 know, most interesting things I've seen
10:11 so far because when you look at the cool
10:13 examples, you'll realize why this does
10:16 feel like a reality simulator in 2026.
10:18 And that's super crazy. So, let's take a
10:19 look at some of the cool examples cuz I
10:21 was browsing Twitter and I saw some
10:22 really cool examples. For example, let's
10:24 take a look at Riley Goodides
10:26 controlling a discarded pack of
10:28 cigarettes. Here you can see he's on the
10:30 underground metro controlling a pack of
10:31 cigarettes and looking around and
10:33 changing the camera angles. You can see
10:35 surprisingly all of the physic remain
10:37 intact. And this is a perspective that
10:39 you wouldn't usually be able to get. Or
10:40 let's take a look at Matt Schumer's
10:42 skiing adventure. Here you can see he's
10:44 controlling a skier as they ski down the
10:46 mountain. And this also includes the
10:48 realistic physics as well as being able
10:50 to jump and control the skier as they
10:52 ski down the mountain. This is really
10:53 cool as it does manage to simulate
10:56 gravity really well as well as colliding
10:57 with the snowy objects in the terrain.
10:59 And here is one of my favorites. You can
11:01 simulate body cam footage with Genie
11:03 Free. Take a look at this footage by
11:05 Chris First. Once you get a little bit,
11:07 you can control the entire environment
11:09 from this super specific camera angle
11:11 that does look super realistic. This is
11:13 a very simple prompt and due to the
11:14 nature of body cam footage, this looks
11:17 incredibly realistic. It really does
11:18 just does just depend on what your
