1 00:00:00,020 --> 00:00:04,380 Ejaaz: Welcome back to This Week in AI. It's been a pretty big week for the robots taking over. 2 00:00:04,660 --> 00:00:08,320 Ejaaz: Tesla are retiring two of their most iconic cars, the Model S and the Model 3 00:00:08,320 --> 00:00:14,020 Ejaaz: X, to make room to scale out their new Optimus Gen 3 robot, which is targeting 4 00:00:14,020 --> 00:00:17,880 Ejaaz: 1 million units this year. Very ambitious, let's see if they make it. 5 00:00:18,220 --> 00:00:21,760 Ejaaz: Figure, another robotic startup, also launched a new humanoid robot, 6 00:00:21,900 --> 00:00:24,040 Ejaaz: which did something that no other robot has done before. 7 00:00:24,220 --> 00:00:28,160 Ejaaz: Clean the dishes, which sounds like a really simple task, but actually requires 8 00:00:28,160 --> 00:00:30,880 Ejaaz: 61 very precise locomotive operations. 9 00:00:31,040 --> 00:00:34,200 Ejaaz: But the best part is that it did it completely autonomously. 10 00:00:34,420 --> 00:00:38,700 Ejaaz: And in final news, the Chinese and Google both released two AI models. 11 00:00:38,960 --> 00:00:42,160 Ejaaz: One can convert video recordings into production-ready apps, 12 00:00:42,200 --> 00:00:47,220 Ejaaz: and the other might just make the first major scientific discovery since we discovered DNA. 13 00:00:47,440 --> 00:00:53,280 Ejaaz: Google also introduced their leading Gemini model to 3.8 billion people in the 14 00:00:53,280 --> 00:00:58,620 Ejaaz: form of a personal AI agent in their browser. an exciting week. Let's get into it. 15 00:00:58,820 --> 00:01:01,360 Josh: Let's start with the Tesla earnings report because that was the highlight of 16 00:01:01,360 --> 00:01:03,420 Josh: the day for me, at least. I've been following these reports forever. 17 00:01:03,540 --> 00:01:06,940 Josh: In fact, it's funny, six years ago, I was watching a video that popped up six 18 00:01:06,940 --> 00:01:09,040 Josh: years ago this day talking about Tesla earnings. 19 00:01:09,160 --> 00:01:12,800 Josh: They were a blowout quarter in terms of forward-looking guidance, 20 00:01:12,800 --> 00:01:16,180 Josh: a little bit less so on the actual numbers. The stock's down for today, 21 00:01:16,260 --> 00:01:17,160 Josh: but that's not the story. 22 00:01:17,420 --> 00:01:22,180 Josh: The story is the forward-looking future of what this company looks like, 23 00:01:22,280 --> 00:01:26,200 Josh: which is very much centered around autonomy and robots and manufacturing. 24 00:01:26,400 --> 00:01:30,940 Josh: In fact, there's a world in which five years from now, you can't even really buy a Tesla. 25 00:01:31,120 --> 00:01:35,640 Josh: And that's signified slightly and early from the Model S and X depreciation. But I think 26 00:01:35,990 --> 00:01:39,730 Josh: You'll continue to see more of that as we go because Tesla is fully leaning into autonomy. 27 00:01:39,950 --> 00:01:43,630 Josh: So much so, in fact, that they're doubling their capital expenditure this year. 28 00:01:43,710 --> 00:01:45,690 Josh: I believe last year it was $8 billion. 29 00:01:46,150 --> 00:01:50,030 Josh: This year they're going up to $20 billion in CapEx spend just to scale these 30 00:01:50,030 --> 00:01:55,890 Josh: production lines for things like the CyberCav and for things like the Optimus humanoid robot. 31 00:01:56,090 --> 00:01:59,350 Josh: Which is really exciting because I think this year, paired with the Helix News, 32 00:01:59,490 --> 00:02:04,650 Josh: we're seeing a lot of humanoid robots, a lot of automation actually coming to the physical worlds. 33 00:02:04,650 --> 00:02:07,790 Josh: Like entering meat space, we're going to be able to walk down a street and not 34 00:02:07,790 --> 00:02:12,290 Josh: only have a self-driving car roll right by you, but also see a humanoid robot 35 00:02:12,290 --> 00:02:15,350 Josh: maybe walking down the street or doing something actually useful this year. 36 00:02:15,530 --> 00:02:19,010 Josh: And Tesla has a plan to scale these at a rate that no one else really is, 37 00:02:19,270 --> 00:02:23,450 Josh: starting with a million units this year and then going to 10 million, 100 million. 38 00:02:23,710 --> 00:02:28,250 Josh: And then they project up to a billion of these. The idea is that these will 39 00:02:28,250 --> 00:02:34,490 Josh: be able to offload enough of the human capital workforce that it'll unlock significant 40 00:02:34,490 --> 00:02:36,070 Josh: amounts of GDP for the actual country. 41 00:02:36,150 --> 00:02:38,450 Josh: So this is a gigantic project that's getting underway. 42 00:02:38,550 --> 00:02:42,470 Josh: And what we're seeing now is the early seeds being planted through these huge 43 00:02:42,470 --> 00:02:46,110 Josh: capital expenditures happening across the globe at all of the gigafactories. 44 00:02:46,110 --> 00:02:53,410 Ejaaz: This feels like a very prescient moment because Elon's clearly saying that Tesla 45 00:02:53,410 --> 00:02:57,030 Ejaaz: in five years or 10 years is not going to be known to be a car company. 46 00:02:57,110 --> 00:02:59,670 Ejaaz: It's going to be an autonomous robot company. 47 00:02:59,770 --> 00:03:04,110 Ejaaz: And we've kind of seen the signs of that over the last year where Elon's kind 48 00:03:04,110 --> 00:03:08,310 Ejaaz: of just spent all his time on full self-driving and improving that neural network 49 00:03:08,310 --> 00:03:12,810 Ejaaz: and focusing on his kind of Cybertruck's autonomously working in SpaceX and 50 00:03:12,810 --> 00:03:16,470 Ejaaz: then this new Optimus Gen 3 robot. that to me is like 51 00:03:17,060 --> 00:03:20,880 Ejaaz: kind of like crazy switch, I guess, to still kind of like materialize in my head. 52 00:03:21,200 --> 00:03:24,940 Ejaaz: And I don't think it's coincidental because this year was also, 53 00:03:25,100 --> 00:03:29,400 Ejaaz: or rather 2025 was the first year that Tesla's annual revenue went down. 54 00:03:29,480 --> 00:03:31,300 Ejaaz: They only put out 1.64 million cars. 55 00:03:31,400 --> 00:03:34,280 Ejaaz: I think that's down like around 9% from the previous year. 56 00:03:34,420 --> 00:03:37,540 Ejaaz: And the fact that he's shutting down two of his most early models, 57 00:03:37,760 --> 00:03:41,940 Ejaaz: I think it was like back in 2012 and 2015 that he released both the S and the X. 58 00:03:42,720 --> 00:03:45,980 Ejaaz: He's now going all in on humanoid robots. I feel like, Josh, 59 00:03:46,080 --> 00:03:50,100 Ejaaz: that this is the same moment where Elon was like, no, FSD is going to be the 60 00:03:50,100 --> 00:03:51,400 Ejaaz: biggest thing for Tesla cars. 61 00:03:51,520 --> 00:03:53,620 Ejaaz: He's doing the same thing for robots now. 62 00:03:53,780 --> 00:03:57,060 Ejaaz: So people might kind of take it as this is kind of crazy right now. 63 00:03:57,200 --> 00:04:02,520 Ejaaz: This humanoid robot isn't really useful for me or can't automate a bunch of manual labor. 64 00:04:02,880 --> 00:04:05,780 Ejaaz: But I think Elon sees something that we don't. Do you agree with this? 65 00:04:05,880 --> 00:04:09,020 Ejaaz: Do you feel kind of like solemn that we're going from cars to robots now? 66 00:04:09,100 --> 00:04:10,460 Ejaaz: Or has this always made sense to you? 67 00:04:10,460 --> 00:04:14,060 Josh: Yeah, well, this has always made sense because they've telegraphed this years in advance. 68 00:04:14,460 --> 00:04:17,760 Josh: Tesla famously has released these master plans, or Elon has released the master 69 00:04:17,760 --> 00:04:21,900 Josh: plan, once every three to five years for the last 15 years. 70 00:04:22,080 --> 00:04:25,080 Josh: And it just very clearly states exactly their plan and what they're going to do. 71 00:04:25,280 --> 00:04:28,480 Josh: And going back and looking at the old content from me, the old content from 72 00:04:28,480 --> 00:04:30,980 Josh: the master plans, it's very clear that this has always been the plan. 73 00:04:31,060 --> 00:04:33,680 Josh: They're just accelerating the advent of sustainable transport. 74 00:04:33,940 --> 00:04:37,120 Josh: And now they're kind of moving to a sustainable abundance through the Optimus 75 00:04:37,120 --> 00:04:42,540 Josh: humanoid robot. But the idea is always just to move the world towards an autonomous future. 76 00:04:42,860 --> 00:04:46,320 Josh: And cars were the gateway drug to doing that. It started with the Roadster. 77 00:04:46,500 --> 00:04:49,620 Josh: The Roadster was the very high-priced model that wealthy people were able to 78 00:04:49,620 --> 00:04:52,800 Josh: buy to fund the production and development of the Model S and the X, 79 00:04:52,920 --> 00:04:56,220 Josh: which funded the production of the scale vehicles, which is the three and the Y, 80 00:04:56,460 --> 00:05:00,240 Josh: which is funding the production of full self-driving, as well as serving as 81 00:05:00,240 --> 00:05:02,860 Josh: the actual network in which they could distribute the software on. 82 00:05:03,100 --> 00:05:06,500 Josh: So it makes sense to depreciate the S and X. I'd like to take a moment of silence 83 00:05:06,500 --> 00:05:09,440 Josh: for these unbelievable cars that changed the world. 84 00:05:09,600 --> 00:05:12,220 Josh: I mean, I would argue the Model S is one of the most. 85 00:05:12,780 --> 00:05:16,800 Josh: One of the most important cars that was ever built and ever shipped because 86 00:05:16,800 --> 00:05:19,820 Josh: it came out in 2012, just five years after the iPhone. 87 00:05:20,000 --> 00:05:24,740 Josh: And it proved that electric cars were not only possible, but they can be great 88 00:05:24,740 --> 00:05:27,620 Josh: and they can be better than gas vehicles across the board. 89 00:05:27,860 --> 00:05:31,440 Josh: And not only that, but they could be capable of autonomous transport in a world 90 00:05:31,440 --> 00:05:35,040 Josh: where all of those things were complete contradictions and believed to be impossible. 91 00:05:35,280 --> 00:05:39,140 Josh: And the Model S did that. And it went on an amazing 15-year, 92 00:05:39,360 --> 00:05:42,480 Josh: 13-year run where they sold almost a million of these vehicles. 93 00:05:42,680 --> 00:05:46,160 Josh: It proved that it was possible. It funded the development of where we are today. 94 00:05:46,280 --> 00:05:47,780 Josh: And they're just unbelievable vehicles. 95 00:05:47,960 --> 00:05:50,920 Josh: A lot of people don't understand that the Model S and the X actually only account 96 00:05:50,920 --> 00:05:54,760 Josh: for about 3% of Tesla's vehicle sales now. It's a very niche audience. 97 00:05:54,900 --> 00:05:55,620 Josh: They're fairly expensive. 98 00:05:55,840 --> 00:05:59,120 Josh: They're over-engineered to hell. They have the craziest features, 99 00:05:59,260 --> 00:06:03,620 Josh: particularly the Model X. So it is very sad to see them go, but it is very much 100 00:06:03,620 --> 00:06:07,800 Josh: expected and very much exciting in the sense that that production line starts 101 00:06:07,800 --> 00:06:09,480 Josh: turning out humanoid robots this year. 102 00:06:10,100 --> 00:06:13,960 Ejaaz: Well, it's also important to remember that robots, as it were, 103 00:06:14,100 --> 00:06:16,740 Ejaaz: isn't just going to materialize as humanoid robots. 104 00:06:16,900 --> 00:06:19,840 Ejaaz: It's also going to look like the cars themselves, right? You mentioned like 105 00:06:19,840 --> 00:06:20,900 Ejaaz: they only account for 3%. 106 00:06:21,680 --> 00:06:25,600 Ejaaz: The rest of the vehicles that he's going to be producing are mainly going to 107 00:06:25,600 --> 00:06:29,240 Ejaaz: be targeted for just completely autonomous vehicles. 108 00:06:29,480 --> 00:06:30,620 Josh: So they won't even have a steering wheel. 109 00:06:30,720 --> 00:06:33,440 Ejaaz: They won't have a steering wheel. Like, it's a good point, right? Which is 110 00:06:33,440 --> 00:06:36,700 Ejaaz: these vehicles are going to become thought of 111 00:06:36,700 --> 00:06:39,640 Ejaaz: as less of a car where you sit in the front and 112 00:06:39,640 --> 00:06:42,360 Ejaaz: steer something to kind of like a third space to kind 113 00:06:42,360 --> 00:06:45,300 Ejaaz: of like a living room where you can be productive or enjoy 114 00:06:45,300 --> 00:06:48,140 Ejaaz: and consume entertainment when you get from 115 00:06:48,140 --> 00:06:52,960 Ejaaz: as you go from point a to point b and we're seeing this with his plans to roll 116 00:06:52,960 --> 00:06:57,080 Ejaaz: out the robotaxi service so um you know at the end of last year he was all kind 117 00:06:57,080 --> 00:07:01,640 Ejaaz: of like there were numerous sightings of him trying robotaxi um he's going to 118 00:07:01,640 --> 00:07:06,500 Ejaaz: be rolling out or planning to roll out the robotaxi service to nine more cities this year, 119 00:07:06,620 --> 00:07:09,320 Ejaaz: which is super exciting to see, nine cities in total. 120 00:07:10,140 --> 00:07:14,200 Ejaaz: And I know that there are several reports at this point where it is good enough 121 00:07:14,200 --> 00:07:19,040 Ejaaz: to drive much better with lower incidental damage than humans. 122 00:07:19,140 --> 00:07:23,380 Ejaaz: I don't know if you saw this, Josh, but Lemonade Insurance, which is a very 123 00:07:23,380 --> 00:07:29,240 Ejaaz: popular insurance startup here in the US, are offering drivers 50% off their 124 00:07:29,240 --> 00:07:33,540 Ejaaz: driver's insurance if they enable FSD on their Tesla cars. 125 00:07:34,260 --> 00:07:37,200 Ejaaz: And I kind of thought of this like fun idea where like it would be funny if 126 00:07:37,200 --> 00:07:42,520 Ejaaz: Elon just kind of like packages his new FSD subscription that we spoke about 127 00:07:42,520 --> 00:07:46,960 Ejaaz: in a previous episode last week with this kind of like insurance discount 128 00:07:47,460 --> 00:07:50,600 Ejaaz: to kind of like increase FSD subscribers and stuff like that. 129 00:07:50,660 --> 00:07:53,760 Ejaaz: I know those are ways that you can mix and match this. I'm excited to see more 130 00:07:53,760 --> 00:07:54,820 Ejaaz: robots roll out this year. 131 00:07:55,210 --> 00:07:58,770 Josh: Yeah, for the people who are listening, the new cities that you will be able 132 00:07:58,770 --> 00:08:02,810 Josh: to get a robotaxi are going to be Austin, Dallas, Houston, Phoenix, 133 00:08:02,970 --> 00:08:04,690 Josh: Miami, Orlando, Tampa, Las Vegas. 134 00:08:04,850 --> 00:08:06,890 Josh: If you live in any of those places by the end of this quarter, 135 00:08:07,050 --> 00:08:10,170 Josh: you will be able to try this for yourself, which I think is so cool. 136 00:08:10,270 --> 00:08:12,290 Josh: And as it rolls out, I would encourage everyone to try. 137 00:08:12,410 --> 00:08:15,030 Josh: Like I mentioned last week, I was in LA and using Waymos. 138 00:08:15,210 --> 00:08:18,970 Josh: And it's unbelievable if you've never experienced it before to see and feel 139 00:08:18,970 --> 00:08:21,170 Josh: what it's like to have a computer driving you around. 140 00:08:21,310 --> 00:08:25,490 Josh: So far, they have 600,000 autonomous miles driven on the robo taxi network that 141 00:08:25,490 --> 00:08:29,150 Josh: number is going to increase it is safer than a human driver as reflected with 142 00:08:29,150 --> 00:08:32,570 Josh: the insurance tesla actually also offers their own insurance which incentivizes 143 00:08:32,570 --> 00:08:36,430 Josh: people to drive safely so really cool updates from tesla across the board very 144 00:08:36,430 --> 00:08:39,990 Josh: excited about that and this isn't the only news from the elon, 145 00:08:40,770 --> 00:08:43,490 Josh: atmosphere i guess this week uh there's also some more 146 00:08:43,490 --> 00:08:46,850 Josh: spacex news particularly as it relates to the ipo which 147 00:08:46,850 --> 00:08:49,970 Josh: seems as if they're going to be raising even more money 148 00:08:49,970 --> 00:08:55,010 Josh: than previously thought 50 billion dollars which is going to be i think the 149 00:08:55,010 --> 00:08:59,470 Josh: largest amount of capital ever raised for an ipo and the largest market cap 150 00:08:59,470 --> 00:09:03,310 Josh: of a company to ever go public all in one this is going to be huge they're targeting 151 00:09:03,310 --> 00:09:08,230 Josh: june as the announcement and release date for this ipo to go public for people 152 00:09:08,230 --> 00:09:09,750 Josh: to actually invest in SpaceX. 153 00:09:10,330 --> 00:09:14,750 Josh: And SpaceX, to me, I mean, this is a huge opportunity. This is a huge company. 154 00:09:14,910 --> 00:09:18,590 Josh: I'm so excited for this to get into the open market and see how the public reacts. 155 00:09:19,010 --> 00:09:23,690 Ejaaz: It's also one of the few IPOs that are raising at such a large valuation that 156 00:09:23,690 --> 00:09:26,850 Ejaaz: I think justifies the amount that they're raising. 157 00:09:26,950 --> 00:09:32,010 Ejaaz: We're going to talk about OpenAI raising $100 billion later on this episode. 158 00:09:32,150 --> 00:09:36,170 Ejaaz: But I would say $50 billion is probably too low in an IPO. 159 00:09:36,310 --> 00:09:38,770 Ejaaz: And I know he's raising other rounds, but like he probably needs more money 160 00:09:38,770 --> 00:09:42,530 Ejaaz: to achieve his mission of getting to Mars, the moon and setting up, 161 00:09:42,630 --> 00:09:48,710 Ejaaz: you know, these interplanetary energy factories and stuff like that that he's spoken about on X. 162 00:09:48,970 --> 00:09:53,670 Ejaaz: So this is really cool and exciting. I think 2026 is gonna be the year of big IPOs, Josh. 163 00:09:54,050 --> 00:09:57,450 Ejaaz: We were talking about this before we started recording. We've got what, the Anthropic IPO. 164 00:09:57,710 --> 00:10:03,270 Ejaaz: You've got OpenAI raising an $850 billion round now and they're gonna be presumably 165 00:10:03,270 --> 00:10:07,950 Ejaaz: IPO-ing more than that. like 1.4 is what some more before trillion dollars is 166 00:10:07,950 --> 00:10:11,210 Ejaaz: what some of the rumors are saying um yeah this is just crazy to see 167 00:10:11,210 --> 00:10:15,850 Josh: Yeah we might get striped this year too a lot of ipos and part of the value 168 00:10:15,850 --> 00:10:20,430 Josh: in which enabled starlink and spacex to go public is the starlink network is 169 00:10:20,430 --> 00:10:23,910 Josh: the internet network and as they improve this network through starship launches 170 00:10:23,910 --> 00:10:26,790 Josh: which allows them to put their new version 3 satellites into space, 171 00:10:27,550 --> 00:10:31,330 Josh: And for reference, those version three satellite launches are 20 to 40 times 172 00:10:31,330 --> 00:10:35,930 Josh: more effective than the current versions that they're pushing out right now with the version two. 173 00:10:36,110 --> 00:10:39,530 Josh: So it's a huge increase in bandwidth. And that bandwidth is going to make its 174 00:10:39,530 --> 00:10:43,770 Josh: way into cell phones in our pockets because the direct to cell network is growing very quickly. 175 00:10:43,890 --> 00:10:47,750 Josh: If you're a user of T-Mobile, you can try this now. But Apple is now reportedly 176 00:10:47,750 --> 00:10:51,650 Josh: planning to integrate Starlink directly into the iPhone 18 Pro, 177 00:10:51,790 --> 00:10:54,210 Josh: which is coming later this year in September. 178 00:10:54,410 --> 00:10:58,590 Josh: So this is really exciting news coming from Apple. Again, outsourcing the software 179 00:10:58,590 --> 00:11:02,310 Josh: stack to someone else who is more capable and competent in doing this. 180 00:11:02,370 --> 00:11:03,190 Josh: In this case, it's Starlink. 181 00:11:03,330 --> 00:11:07,750 Josh: And it solves one of the huge pain points for the perfect mobile device, 182 00:11:07,950 --> 00:11:12,550 Josh: where my hypothetical perfect mobile device has unlimited battery, 183 00:11:12,990 --> 00:11:14,890 Josh: unlimited connectivity, no matter where you are. 184 00:11:15,190 --> 00:11:17,710 Josh: And now this solves the connectivity problem, where there's nowhere you can 185 00:11:17,710 --> 00:11:21,490 Josh: go where Starlink will not be. And that's a really exciting partnership that I'm looking forward to. 186 00:11:21,490 --> 00:11:27,530 Ejaaz: In other robotics news, I saw one of the coolest, but maybe unsexy to some, 187 00:11:27,890 --> 00:11:29,510 Ejaaz: demo of a humanoid robot. 188 00:11:29,870 --> 00:11:32,430 Ejaaz: What you're watching on the screen right now, and if you're listening to this, 189 00:11:32,510 --> 00:11:36,830 Ejaaz: I'll describe it to you, is a humanoid robot, which looks very much like a human, 190 00:11:36,850 --> 00:11:41,050 Ejaaz: I have to say, he's just missing some clothes, unloading a dishwasher, 191 00:11:41,470 --> 00:11:43,790 Ejaaz: walking across a full-sized kitchen, 192 00:11:44,250 --> 00:11:47,110 Ejaaz: placing all these dishes and crockery into the cabinet. 193 00:11:47,110 --> 00:11:49,550 Ejaaz: Very delicate, precise movements are needed for this, by the way, 194 00:11:49,910 --> 00:11:53,310 Ejaaz: walking back and refilling the dishwasher with all the dirty dishes. 195 00:11:53,470 --> 00:11:56,910 Ejaaz: These include wine glasses, by the way. Okay. So for all of you haters out there 196 00:11:56,910 --> 00:12:00,750 Ejaaz: that are saying they're using plastic stuff, no, he's using delicate glasses as well. 197 00:12:00,970 --> 00:12:05,850 Ejaaz: Now, what you're looking at is the new robot from FIGURE, a very hot robotic 198 00:12:05,850 --> 00:12:09,290 Ejaaz: startup here in the US called Helix O2. 199 00:12:09,670 --> 00:12:13,550 Ejaaz: And this humanoid robot is different from the other humanoid robots in a few different ways. 200 00:12:13,670 --> 00:12:20,610 Ejaaz: Number one, it is the first robot to perform a four-minute fully autonomous task of this level. 201 00:12:20,920 --> 00:12:24,100 Ejaaz: Just get that into your head. That may not sound very impressive right now, 202 00:12:24,180 --> 00:12:28,800 Ejaaz: but I remember a time where AI model LLMs could only work autonomously for 10 203 00:12:28,800 --> 00:12:31,680 Ejaaz: minutes at a time. And now they're doing weeks at a time. 204 00:12:31,840 --> 00:12:34,960 Ejaaz: This is the same kind of moment that we see for robotics. 205 00:12:35,200 --> 00:12:38,040 Ejaaz: And I think that Helix is kind of like nailing that. 206 00:12:38,360 --> 00:12:40,880 Ejaaz: To emphasize, it is completely autonomous. There are no humans. 207 00:12:41,080 --> 00:12:45,720 Ejaaz: There's no teleoperation. You know who dares leverage teleoperation? 1KX. 208 00:12:46,620 --> 00:12:50,260 Ejaaz: Also Tesla Optimus, dare I say, although the new generation should be fully 209 00:12:50,260 --> 00:12:52,200 Ejaaz: autonomous in some way, shape, or form. 210 00:12:52,460 --> 00:12:57,100 Ejaaz: This is a really impressive feat because the technicalities and the mechanisms 211 00:12:57,100 --> 00:13:00,960 Ejaaz: that you need to achieve this is nothing short of just insane. 212 00:13:01,140 --> 00:13:06,540 Ejaaz: So 61 locomotive manipulations are required for this robot to do this, 213 00:13:06,680 --> 00:13:08,580 Ejaaz: but that's not even the most impressive part. 214 00:13:08,900 --> 00:13:15,000 Ejaaz: Typically, these robots are pre-programmed to execute on a set number of different moves. 215 00:13:15,080 --> 00:13:18,240 Ejaaz: So those 61 moves I just mentioned, you need to hard code it. 216 00:13:18,360 --> 00:13:21,900 Ejaaz: And when it faces a scenario, it'll then activate that movement. 217 00:13:22,060 --> 00:13:23,340 Ejaaz: This is completely different. 218 00:13:23,520 --> 00:13:27,960 Ejaaz: So what Felix did was they launched something called System Zero, 219 00:13:28,160 --> 00:13:32,440 Ejaaz: I believe, which is their new neural net, which is only 10 million parameters. 220 00:13:32,440 --> 00:13:36,900 Ejaaz: But it basically ingests what it sees around it through its palm senses, 221 00:13:37,080 --> 00:13:39,640 Ejaaz: through its eyes, and reacts in real time. 222 00:13:39,820 --> 00:13:42,280 Ejaaz: This is different from how the model previously used to work, which is 223 00:13:42,810 --> 00:13:45,930 Ejaaz: Take in some kind of input, process it through its hard code, 224 00:13:46,110 --> 00:13:47,350 Ejaaz: and then output an action. 225 00:13:47,610 --> 00:13:53,510 Ejaaz: And hey, presto, the kind of like trophy win here is that it's fully autonomous, 226 00:13:53,590 --> 00:13:55,990 Ejaaz: and it's the first humanoid to be able to do this. 227 00:13:56,390 --> 00:14:00,350 Ejaaz: Now, to remove all that code and transform it into neural net is cool enough. 228 00:14:00,430 --> 00:14:03,890 Ejaaz: But what I also like about it is that they have this new form of tactile sensors. 229 00:14:04,050 --> 00:14:07,330 Ejaaz: And I think we spoke about something about this on our previous run, 230 00:14:07,330 --> 00:14:09,210 Ejaaz: or maybe it was even last week's. 231 00:14:09,430 --> 00:14:12,850 Ejaaz: But because of its palm sensors and its new tactile sensors, 232 00:14:13,130 --> 00:14:16,970 Ejaaz: it's able to know when to apply pressure and when to be delicate. 233 00:14:17,170 --> 00:14:19,670 Ejaaz: For example, lifting a wine glass and placing it into a cabinet. 234 00:14:19,850 --> 00:14:23,530 Ejaaz: So I personally like took this kind of boring task and thought, 235 00:14:23,670 --> 00:14:26,610 Ejaaz: huh, this is actually something that I would actually have in my home because 236 00:14:26,610 --> 00:14:28,870 Ejaaz: I can trust it. Also, look at the speed that it's moving. 237 00:14:29,110 --> 00:14:33,030 Ejaaz: It's definitely quicker than my sister that I've seen her loading the dishwasher at home at least. 238 00:14:33,390 --> 00:14:37,510 Josh: Do you think this is ready to be in your home? Is this something you'd be interested in buying? 239 00:14:37,910 --> 00:14:41,150 Ejaaz: Well, I would need it to do something more than just dishwasher stuff. 240 00:14:41,310 --> 00:14:44,570 Ejaaz: But presumably, if it can do this, it can also fold my laundry, 241 00:14:44,730 --> 00:14:47,250 Ejaaz: Josh, and load the laundry basket as well. 242 00:14:47,390 --> 00:14:50,830 Ejaaz: So depending on what they're charging, if it's a subscription or if I have to 243 00:14:50,830 --> 00:14:54,470 Ejaaz: buy it outright, would influence my decision, right? If I could have a live-in housekeeper, 244 00:14:55,250 --> 00:14:57,110 Ejaaz: sounds pretty good i'm a pretty messy guy 245 00:14:57,110 --> 00:14:59,870 Josh: Yeah when evaluating robots i think the things 246 00:14:59,870 --> 00:15:03,270 Josh: that are most important to look at are first the hands the dexterity the 247 00:15:03,270 --> 00:15:06,130 Josh: the sensors within the hands how capable the hands 248 00:15:06,130 --> 00:15:09,010 Josh: are because that's by far the most complicated thing the second thing is 249 00:15:09,010 --> 00:15:12,990 Josh: the intelligence how general purpose is it we we've seen this example over and 250 00:15:12,990 --> 00:15:17,170 Josh: over again of loading a dishwasher or doing simple tasks like sorting clothes 251 00:15:17,170 --> 00:15:21,230 Josh: how complex are those tasks able to get and how quickly are they able to roll 252 00:15:21,230 --> 00:15:25,630 Josh: that out is the second thing and then the third is how quickly and how effectively 253 00:15:25,630 --> 00:15:27,650 Josh: are they able to actually manufacture this at scale? 254 00:15:27,850 --> 00:15:31,930 Josh: Are they going to be able to create enough of these at a price point that makes 255 00:15:31,930 --> 00:15:36,310 Josh: enough sense with a feature set that makes enough sense to encourage people 256 00:15:36,310 --> 00:15:37,550 Josh: like me and you to want to buy one? 257 00:15:37,690 --> 00:15:42,330 Josh: It seems like there is a very long and challenging path for figure to get there. 258 00:15:42,570 --> 00:15:46,390 Josh: I'm really looking forward to Tesla's Optimus 3, which is coming this quarter 259 00:15:46,390 --> 00:15:50,070 Josh: and will be rolling out at scale this year. And I... 260 00:15:50,740 --> 00:15:53,360 Josh: I'm hopeful that figure can keep up because competition is good. 261 00:15:53,460 --> 00:15:55,980 Josh: And I hope they don't turn into like the Rivian of electric cars. 262 00:15:56,120 --> 00:15:57,100 Josh: I want them to stay on the forefront. 263 00:15:57,320 --> 00:16:00,760 Josh: And I hope Brett and the whole figure team continue to keep their foot on the gas. 264 00:16:01,180 --> 00:16:05,260 Ejaaz: Yeah, I mean, I think this demo is cool, right? But it's pitched towards... It's like, it's cute. 265 00:16:05,880 --> 00:16:11,140 Ejaaz: It's cute. But I think the real application, the reality is these robots aren't 266 00:16:11,140 --> 00:16:13,520 Ejaaz: going to be in your homes first. They're going to be in factories. 267 00:16:13,780 --> 00:16:19,120 Ejaaz: In fact, their Helix O1 robot has been working at BMW controlling their operating 268 00:16:19,120 --> 00:16:23,660 Ejaaz: system or one of their manufacturing plants for, what is it now? Four months now? 269 00:16:24,140 --> 00:16:27,820 Ejaaz: 24-7? There's like a live stream of this. Well, hopefully, maybe we can link 270 00:16:27,820 --> 00:16:29,080 Ejaaz: to this and people can go watch it. 271 00:16:29,160 --> 00:16:32,700 Ejaaz: But Brett keeps talking about it every single week. 272 00:16:32,900 --> 00:16:36,520 Ejaaz: So I don't know. I'm bullish on it. I think Optimus is also going to be in factories 273 00:16:36,520 --> 00:16:37,740 Ejaaz: first before people's homes. 274 00:16:38,480 --> 00:16:40,020 Ejaaz: But we shall see. 275 00:16:40,860 --> 00:16:44,820 Ejaaz: We have some big news from Google. Josh, you were showing me some pretty cool 276 00:16:44,820 --> 00:16:46,420 Ejaaz: demos. What's going on here? 277 00:16:46,680 --> 00:16:50,460 Josh: Yeah, big news. If you remember from a little while ago, we had the CEO of Perplexity 278 00:16:50,460 --> 00:16:53,260 Josh: on the show, and he was talking about the new pivot to the AI browser. 279 00:16:53,500 --> 00:16:56,120 Josh: Since then, OpenAI has released their own AI browser. 280 00:16:56,820 --> 00:17:01,600 Josh: Claude has released a plugin for Chrome, but now Chrome, where everyone actually 281 00:17:01,600 --> 00:17:04,280 Josh: uses their browsing, has rolled out their own... 282 00:17:05,020 --> 00:17:08,740 Josh: Gemini agentic controller inside of the browser. And it's pretty incredible. 283 00:17:09,140 --> 00:17:14,600 Josh: I mean, like you mentioned at the intro, 3.8 billion people on earth use the Chrome browser. 284 00:17:14,760 --> 00:17:18,520 Josh: And now all those people have just been given a magical update overnight that 285 00:17:18,520 --> 00:17:22,840 Josh: allows this agent in and allows it to actually control and do things on behalf 286 00:17:22,840 --> 00:17:24,500 Josh: of the user within this browser. 287 00:17:24,700 --> 00:17:28,800 Josh: So we're watching on screen a introduction demo video that shows some examples. 288 00:17:29,020 --> 00:17:31,100 Josh: And I thought they were really amazing. 289 00:17:31,340 --> 00:17:35,920 Josh: The first one that they did was with NanoBanana Pro, which is the image generating model. 290 00:17:36,060 --> 00:17:39,680 Josh: And they were just looking at a home and they wanted to add furniture to it, 291 00:17:39,760 --> 00:17:44,040 Josh: like looking at an empty apartment. And you could just ask NanoBanana to stage the apartment. 292 00:17:44,300 --> 00:17:47,440 Josh: Yeah, here's the demo. Where it's empty, you could ask it to stage the apartment 293 00:17:47,440 --> 00:17:50,580 Josh: and it shows you what it looks like with furniture on the fly as you're browsing. 294 00:17:50,800 --> 00:17:56,360 Josh: The follow-up feature to this is they were looking for an apartment that accepted 295 00:17:56,360 --> 00:18:00,620 Josh: dogs, that was in X proximity to this place, that had a whole list to parameters 296 00:18:00,620 --> 00:18:03,980 Josh: that were pretty complicated that would have taken the average person a long time to do. 297 00:18:04,200 --> 00:18:08,240 Josh: And they typed in that prompt and it actually clicked through the browser and 298 00:18:08,240 --> 00:18:09,460 Josh: navigated through this for them. 299 00:18:10,050 --> 00:18:13,050 Josh: Putting in all the filters and figuring out which results were best for them. 300 00:18:13,390 --> 00:18:17,110 Josh: And I think this is such a great use case because it's just so prevalent in 301 00:18:17,110 --> 00:18:21,330 Josh: what you do every day as a user of Chrome and as someone who uses a Chrome browser. 302 00:18:21,470 --> 00:18:24,910 Josh: And I'm excited to see it integrated where people are, where you don't actually 303 00:18:24,910 --> 00:18:28,310 Josh: need to download a separate piece of software, move everything over. 304 00:18:28,450 --> 00:18:31,190 Josh: It's just meeting people where they are and building on top of this existing 305 00:18:31,190 --> 00:18:34,170 Josh: knowledge base that Google has about you through your Gmail account. 306 00:18:34,170 --> 00:18:39,810 Ejaaz: What I like about this is they're not over-promising, or they didn't over-promise 307 00:18:40,570 --> 00:18:44,370 Ejaaz: Gemini 3 being pretty much wherever you are on the internet. 308 00:18:44,550 --> 00:18:48,750 Ejaaz: Like, I've used Chrome browser for years now. I've used Gmail for longer. 309 00:18:49,070 --> 00:18:54,150 Ejaaz: And the fact that, you know, these demos show kind of places that I visit every 310 00:18:54,150 --> 00:18:56,970 Ejaaz: day, because like, let's face it, like most of my online life, 311 00:18:57,070 --> 00:19:00,130 Ejaaz: or rather, most of my computer life doesn't live on my desktop, right? 312 00:19:00,210 --> 00:19:03,630 Ejaaz: Which is my, which was separately my point around cloudbot i'm like it's not 313 00:19:03,630 --> 00:19:07,650 Ejaaz: useful for me because i'm always on the internet but google is bringing gemini 314 00:19:07,650 --> 00:19:12,310 Ejaaz: and ai cutting edge ai to me on the internet on the browser where i am every day 315 00:19:12,940 --> 00:19:16,340 Ejaaz: I might not be someone that edits images all the time and says, 316 00:19:16,440 --> 00:19:20,300 Ejaaz: hey, show me in this outfit, but I am someone that writes a lot of emails. 317 00:19:20,300 --> 00:19:22,160 Ejaaz: I am someone that spends a lot of time in documents. 318 00:19:22,340 --> 00:19:27,120 Ejaaz: And in some of these demos, Josh, it doesn't just get me to prompt it. 319 00:19:27,240 --> 00:19:33,620 Ejaaz: It prompts me and says, hey, I see that you have updated some of these goals for the next month. 320 00:19:33,760 --> 00:19:36,200 Ejaaz: Do you want me to email so-and-so on your behalf? 321 00:19:36,480 --> 00:19:39,680 Ejaaz: I've generated you a prescript if you want me to do that for you. 322 00:19:39,820 --> 00:19:43,740 Ejaaz: All I need to do is look over to it on my task panel on the right, and click send. 323 00:19:43,960 --> 00:19:46,200 Ejaaz: And that's just like such a beautiful experience. 324 00:19:46,440 --> 00:19:50,020 Ejaaz: It sounds, again, unsexy, but I think this is going to make people way, 325 00:19:50,120 --> 00:19:53,460 Ejaaz: way more productive and most importantly, loyal to Google. 326 00:19:53,680 --> 00:19:57,320 Ejaaz: We mentioned their major release of personal intelligence, which is basically 327 00:19:57,320 --> 00:20:02,220 Ejaaz: a personal AI assistant that is prevalent across every single Google product 328 00:20:02,220 --> 00:20:06,400 Ejaaz: that you use, not just on your browser, but on Google Maps, Android, wherever you are. 329 00:20:06,740 --> 00:20:10,500 Ejaaz: This is it coming into reality just for Chrome. It's frigging awesome. 330 00:20:10,900 --> 00:20:14,100 Josh: Yeah, really nice feature. It's just, it's, it's again, it's another muscle 331 00:20:14,100 --> 00:20:17,740 Josh: that people are going to need to train as they get used to these new tools in 332 00:20:17,740 --> 00:20:21,020 Josh: reaching for the agent in your browser versus trying to do it all yourself, 333 00:20:21,020 --> 00:20:24,680 Josh: where I would encourage anyone as you're browsing and you're doing something, 334 00:20:24,680 --> 00:20:26,460 Josh: you're trying to achieve a goal. 335 00:20:27,180 --> 00:20:30,640 Josh: Try to click that Gemini button and see if it can achieve that goal for you. 336 00:20:30,840 --> 00:20:35,000 Josh: And the process of doing that over and over kind of trains your brain to rely more on this. 337 00:20:35,100 --> 00:20:38,640 Josh: And it just leads to a far better quality of life because it's better at doing 338 00:20:38,640 --> 00:20:40,220 Josh: a lot of these tasks than we are. 339 00:20:40,460 --> 00:20:43,420 Josh: The other thing that I thought was interesting to note is that they updated 340 00:20:43,420 --> 00:20:48,420 Josh: their Gemini AI Pro plans, the Pro plan and the Ultra plan, which are either 341 00:20:48,420 --> 00:20:50,360 Josh: $20 a month and $250 a month. 342 00:20:50,660 --> 00:20:54,540 Josh: If you have those now, the big thing on this long list of things that they include 343 00:20:54,540 --> 00:20:56,100 Josh: is Google Cloud Credits. 344 00:20:56,320 --> 00:21:00,720 Josh: So now you can actually use, you could build your own environments for these AIs. 345 00:21:00,880 --> 00:21:04,660 Josh: You could start querying more of their API to get more, I guess, 346 00:21:04,780 --> 00:21:06,240 Josh: more just requests done, more intelligence. 347 00:21:06,420 --> 00:21:10,240 Josh: They're including a lot more tools in this package for people who are interested 348 00:21:10,240 --> 00:21:12,440 Josh: in just toying around with these products. 349 00:21:12,600 --> 00:21:15,660 Josh: And for me, this is really exciting because I'm not running large scale, 350 00:21:16,310 --> 00:21:18,730 Josh: products. I don't need a server that costs more than $10 a month. 351 00:21:18,830 --> 00:21:22,570 Josh: And essentially for one membership of $20, I get all of these really cool perks. 352 00:21:22,670 --> 00:21:25,030 Josh: You get access to the entire Google AI ecosystem. 353 00:21:25,150 --> 00:21:29,750 Josh: And I just, I liked this pivot that they're moving towards kind of giving people 354 00:21:29,750 --> 00:21:33,250 Josh: the tools to tinker at a very low price point at $20 a month. 355 00:21:33,390 --> 00:21:38,170 Ejaaz: So in other Google news, they released what is probably the most impressive 356 00:21:38,170 --> 00:21:40,570 Ejaaz: scientific AI breakthrough ever. 357 00:21:41,050 --> 00:21:44,930 Ejaaz: And no one kind of noticed it because it got published in Nature, 358 00:21:44,990 --> 00:21:48,290 Ejaaz: which is a scientific journal, but it was 100% open source and super cool. 359 00:21:48,430 --> 00:21:54,310 Ejaaz: So this is, the best way to think about this is this is kind of like a DNA sequencer AI. 360 00:21:54,750 --> 00:22:02,070 Ejaaz: Now, the reason why this is cool is 98% of our DNA is actually not that important. 361 00:22:02,210 --> 00:22:04,750 Ejaaz: It doesn't actually lead to protein production. 362 00:22:04,910 --> 00:22:07,670 Ejaaz: The reason why protein production is important is because it influences pretty 363 00:22:07,670 --> 00:22:12,890 Ejaaz: much how your entire body, mind, and world works, how your perception is, right? 364 00:22:13,530 --> 00:22:18,490 Ejaaz: The issue is we don't understand the 98% of that dark matter because it controls 365 00:22:18,490 --> 00:22:20,670 Ejaaz: some types of genetic expressions. 366 00:22:21,010 --> 00:22:26,230 Ejaaz: So for decades, like humans have basically just kind of like mapped out the 367 00:22:26,230 --> 00:22:30,050 Ejaaz: genome themselves and then like just tried a bunch of different tests to figure 368 00:22:30,050 --> 00:22:34,230 Ejaaz: out what proteins or what genes are expressed in A, B or C type of condition. 369 00:22:34,680 --> 00:22:38,200 Ejaaz: Now there's an AI model that can not just sequence a couple hundred thousand 370 00:22:38,200 --> 00:22:43,020 Ejaaz: of these base pairs for your DNA, but up to 1 million base pairs. 371 00:22:43,200 --> 00:22:46,340 Ejaaz: Now there are four main base pairs in your DNA strand. We're taking it back 372 00:22:46,340 --> 00:22:49,040 Ejaaz: to like, you know, fifth grade biology here. 373 00:22:49,320 --> 00:22:51,740 Ejaaz: And if you connect them together in different types of combinations, 374 00:22:51,740 --> 00:22:55,440 Ejaaz: you can result in different types of proteins, which kind of make up your genetic makeup. 375 00:22:55,960 --> 00:23:01,360 Ejaaz: Now you have an AI model that can just sequence this entirely and predict what 376 00:23:01,360 --> 00:23:05,100 Ejaaz: the expression is going to be. Now, the reason why this is super cool is you 377 00:23:05,100 --> 00:23:06,380 Ejaaz: can apply this to so many different things. 378 00:23:06,440 --> 00:23:09,820 Ejaaz: You could apply it to preventative medicine. You could apply it to personalized 379 00:23:09,820 --> 00:23:12,600 Ejaaz: medicine to create personalized pharmaceuticals for yourself. 380 00:23:12,740 --> 00:23:17,520 Ejaaz: Or you could recognize genetic disorders before it manifests, 381 00:23:17,660 --> 00:23:23,620 Ejaaz: before it becomes an issue and kind of snip that baby out. So it is such a cool thing to see. 382 00:23:23,740 --> 00:23:26,140 Ejaaz: It kind of harkens back to my biology degree where I'm just like, 383 00:23:26,240 --> 00:23:30,080 Ejaaz: oh my God, I can't believe that they've produced a tool that could have saved 384 00:23:30,080 --> 00:23:32,400 Ejaaz: me like months of work when I was doing that degree. 385 00:23:32,520 --> 00:23:35,820 Ejaaz: But I just think it is super awesome. Josh, are you impressed by something like this? 386 00:23:36,080 --> 00:23:41,140 Josh: Yeah, this to me is the coolest news of the week, actually, which is funny enough. 387 00:23:41,480 --> 00:23:46,260 Josh: The biology stuff is so unbelievably impressive. And I think it's easy to overlook 388 00:23:46,260 --> 00:23:49,760 Josh: it because it's kind of complicated and difficult to understand for those. 389 00:23:49,760 --> 00:23:53,060 Josh: I mean, like myself, who just haven't spent much time in the world of biology. 390 00:23:53,260 --> 00:23:58,240 Josh: But in the research and learning about this segment, I discovered that the DNA 391 00:23:58,240 --> 00:24:00,480 Josh: sequence is 3 billion letters long. 392 00:24:00,640 --> 00:24:03,240 Josh: It's this ginormous sequence of letters. 393 00:24:03,790 --> 00:24:07,070 Josh: It essentially acts as an instruction manual for the 394 00:24:07,070 --> 00:24:09,750 Josh: human body for the makeup it tells you everything that's right with you but 395 00:24:09,750 --> 00:24:13,650 Josh: also everything that's wrong with you and previous ai models before this they 396 00:24:13,650 --> 00:24:16,410 Josh: were like it was like reading a manual with either a magnifying glass where 397 00:24:16,410 --> 00:24:20,450 Josh: you can see tiny details but only a few segments at a time versus binoculars 398 00:24:20,450 --> 00:24:23,470 Josh: where you could see like a lot but it's kind of blurry this is able to do both 399 00:24:23,470 --> 00:24:26,870 Josh: simultaneously and like you mentioned it can read one million letters at once 400 00:24:26,870 --> 00:24:29,570 Josh: with perfect clarity so while it's not the three billion. 401 00:24:29,830 --> 00:24:33,330 Josh: It is 1 million, which is a significant improvement because when doctors find 402 00:24:33,330 --> 00:24:36,910 Josh: a weird genetic variant and they ask, is this variant causing a problem? 403 00:24:37,370 --> 00:24:42,070 Josh: Alpha genome has beat the previous models on 25 out of 26 of these tests, 404 00:24:42,250 --> 00:24:45,010 Josh: which means there's a lot of fewer, we don't know the answer, 405 00:24:45,210 --> 00:24:46,790 Josh: answers for patients with rare diseases. 406 00:24:47,050 --> 00:24:51,210 Josh: And it's able to kind of diagnose things that previously were not possible. 407 00:24:51,330 --> 00:24:54,830 Josh: And as we see progress through this, there's going to be a lot of unbelievable 408 00:24:54,830 --> 00:25:00,130 Josh: examples as it relates to like cancer and splicing these DNA sequences and improving things. 409 00:25:00,250 --> 00:25:04,990 Josh: It's just, it's a huge quality of life improvement for the people who understand 410 00:25:04,990 --> 00:25:07,270 Josh: it and for the people who stand to benefit from this, which is, 411 00:25:07,630 --> 00:25:09,790 Josh: I mean, in a way, everyone we mentioned on a previous episode, 412 00:25:09,870 --> 00:25:13,790 Josh: you can upload your DNA sequence to something like Claude and you could ask it questions. 413 00:25:14,070 --> 00:25:18,090 Josh: This is that on steroids. This is going down to the letter and diagnosing what's 414 00:25:18,090 --> 00:25:21,930 Josh: wrong and what's right based on a single variance in a 3 billion letter data 415 00:25:21,930 --> 00:25:23,050 Josh: set. It's really amazing. 416 00:25:23,820 --> 00:25:28,940 Ejaaz: For those of you who haven't spent time in many of the scientific journal world, 417 00:25:29,100 --> 00:25:36,080 Ejaaz: there are tens of thousands to hundreds of thousands of scientific papers published every single year. 418 00:25:36,540 --> 00:25:40,940 Ejaaz: But the common pattern across all of these is that some of these conclusions 419 00:25:40,940 --> 00:25:42,440 Ejaaz: are simply inconclusive. 420 00:25:42,660 --> 00:25:45,980 Ejaaz: They don't have enough data or they just don't know. They haven't got enough 421 00:25:45,980 --> 00:25:48,140 Ejaaz: tools or technology to figure it out. 422 00:25:48,140 --> 00:25:54,060 Ejaaz: Now, a tool like this, to your point, Josh, I don't think it's going to suddenly 423 00:25:54,060 --> 00:25:57,640 Ejaaz: have a crazy discovery this year, but you're going to see the discovery in some 424 00:25:57,640 --> 00:26:00,540 Ejaaz: of these papers that come out from the researchers that are using it. 425 00:26:00,560 --> 00:26:03,960 Ejaaz: In fact, we're probably going to see Alpha Genome be quoted as a main author 426 00:26:03,960 --> 00:26:05,820 Ejaaz: on some of the papers produced in 2026. 427 00:26:06,260 --> 00:26:12,540 Ejaaz: There are already 3,000 researchers using this. I think 1 million API calls every single day. 428 00:26:12,720 --> 00:26:17,180 Ejaaz: So it's really been crushing a bunch of research across 160 countries. 429 00:26:17,180 --> 00:26:21,900 Ejaaz: So I am curious to see what this gets involved in and what discoveries it makes. 430 00:26:22,000 --> 00:26:24,200 Ejaaz: To your point, maybe it's cancer or maybe it's something smaller, 431 00:26:24,200 --> 00:26:28,060 Ejaaz: but one thing's for sure, there's going to be something major this year, would be my prediction. 432 00:26:28,600 --> 00:26:31,400 Josh: Yeah, the velocity in progress of this has just been, it's unbelievable. 433 00:26:31,600 --> 00:26:32,340 Josh: It's really exciting to see. 434 00:26:32,380 --> 00:26:35,400 Josh: I'm looking forward to all the continued breakthroughs that scientists are now 435 00:26:35,400 --> 00:26:38,860 Josh: able to use because this is just open weights, open source, freely available 436 00:26:38,860 --> 00:26:41,540 Josh: for anyone who's involved in research to actually go and use for themselves. 437 00:26:41,800 --> 00:26:45,020 Josh: There's another interesting breakthrough, or at least released this 438 00:26:45,020 --> 00:26:48,280 Josh: week that's slightly less eccentric this 439 00:26:48,280 --> 00:26:51,080 Josh: one's more practical and something you can use today which is Claude 440 00:26:51,080 --> 00:26:54,080 Josh: releasing interactive tools within Claude 441 00:26:54,080 --> 00:26:57,040 Josh: itself so if you are a user of Slack or 442 00:26:57,040 --> 00:27:00,180 Josh: Figma or Asana or any of these programs you 443 00:27:00,180 --> 00:27:04,240 Josh: actually never need to go to those websites ever again they will be integrated 444 00:27:04,240 --> 00:27:08,380 Josh: into your Claude interface on your desktop for you to actually engage with through 445 00:27:08,380 --> 00:27:11,940 Josh: the chat interface to do whatever you want them to do so the example we're seeing 446 00:27:11,940 --> 00:27:15,300 Josh: on screen now is figma if you're a designer this is how you design things this 447 00:27:15,300 --> 00:27:17,500 Josh: is how you create websites create renderings of things. 448 00:27:18,060 --> 00:27:23,100 Josh: And built into Claude now is tooling control that allows you to type in a prompt 449 00:27:23,100 --> 00:27:27,120 Josh: or even manually adjust things yourselves from within the interface in a way 450 00:27:27,120 --> 00:27:31,020 Josh: that's smart, it understands the context, and it writes back to the main server. 451 00:27:31,220 --> 00:27:34,500 Josh: So if you, like for example, we use Asana for production workflows, 452 00:27:34,720 --> 00:27:35,540 Josh: for scheduling episodes. 453 00:27:35,900 --> 00:27:40,320 Josh: I don't need to go to asana.com anymore. I just ask Claude, when do we need to publish this by? 454 00:27:40,460 --> 00:27:43,800 Josh: Who is responsible for publishing this? And it has all the answers built in, 455 00:27:43,880 --> 00:27:47,880 Josh: and it can write the changes to that. And this is on the back of the integrations 456 00:27:47,880 --> 00:27:48,780 Josh: that they've recently added. 457 00:27:48,940 --> 00:27:51,880 Josh: Things like iMessage, which I think is really cool. 458 00:27:52,000 --> 00:27:55,620 Josh: Previously, or early in this week, we released an episode on CloudBot, 459 00:27:55,680 --> 00:27:59,080 Josh: which is this whole desktop controller that allows you access to your whole 460 00:27:59,080 --> 00:28:01,520 Josh: desktop, but it is a bit technical. It does cost a bit of money. 461 00:28:01,820 --> 00:28:07,080 Josh: This is so easy. If you are a Mac user and you have Cloud downloaded on your laptop. 462 00:28:07,780 --> 00:28:12,180 Josh: You can connect Claude to iMessage and then have it send messages on your behalf 463 00:28:12,180 --> 00:28:16,300 Josh: or read messages on your behalf or integrate it into any of these other services. 464 00:28:16,680 --> 00:28:19,780 Josh: And what they're doing is they're just creating, I think what OpenAI was trying 465 00:28:19,780 --> 00:28:23,940 Josh: to do, which is this full stack place in which you can go for any of your needs. 466 00:28:23,940 --> 00:28:26,940 Josh: It's just slowly embedding itself into more and more of your day-to-day workflow. 467 00:28:27,660 --> 00:28:33,280 Ejaaz: To me, what this announcement says is it's confirmation that we are heading 468 00:28:33,280 --> 00:28:37,080 Ejaaz: towards an AI operating system that is ephemeral. 469 00:28:37,080 --> 00:28:41,140 Ejaaz: Now, we've mentioned this many times on the show before, but I believe in the 470 00:28:41,140 --> 00:28:44,400 Ejaaz: future, you're not going to navigate to an app or a website or scroll. 471 00:28:44,660 --> 00:28:49,820 Ejaaz: It's just going to appear ephemerally in front of you, auto-generated by tokens 472 00:28:49,820 --> 00:28:54,720 Ejaaz: generated by an LLM or any kind of AI model that is multimodal, right? 473 00:28:54,720 --> 00:28:57,740 Ejaaz: And the reason that description that you just gave just now, 474 00:28:57,920 --> 00:29:03,720 Ejaaz: actually, Josh, is, you know, it all kind of converges into one synonymous experience 475 00:29:03,720 --> 00:29:09,680 Ejaaz: where you could be on different devices or you could be on your browser or you could be on a chat bot. 476 00:29:09,720 --> 00:29:15,040 Ejaaz: And somehow the same LLM or AI model follows you around and is there to be helpful 477 00:29:15,040 --> 00:29:18,840 Ejaaz: and expressive in many different ways. This is Anthropic confirming it. 478 00:29:19,440 --> 00:29:25,340 Ejaaz: OpenAI actually released this feature themselves in ChatGPT last year. 479 00:29:25,540 --> 00:29:28,920 Ejaaz: Actually, I think it was mid last year. So that was super early with the exact same partners. 480 00:29:29,180 --> 00:29:33,020 Ejaaz: That uptake didn't really take off. I'm curious whether like six months later, 481 00:29:33,180 --> 00:29:38,480 Ejaaz: we now see a similar pattern with Anthropic or whether people enjoy it and find it more useful here. 482 00:29:39,300 --> 00:29:43,720 Ejaaz: Another kind of subtle point that this is using underneath is something called 483 00:29:43,720 --> 00:29:47,820 Ejaaz: MCP, Model Context Protocol, which is something that Um... 484 00:29:48,200 --> 00:29:51,800 Ejaaz: Anthropic, I believe, found it. It's an open protocol that allows you to connect 485 00:29:51,800 --> 00:29:52,800 Ejaaz: to a bunch of different apps. 486 00:29:53,120 --> 00:29:57,160 Ejaaz: Now, the change here is that it's allowing the generative experiences to happen 487 00:29:57,160 --> 00:30:01,160 Ejaaz: within the chatbot, which is a UX decision, which I think is really cool. 488 00:30:01,800 --> 00:30:06,580 Ejaaz: Now, I think this is a very strategic move for Anthropic because they created 489 00:30:06,580 --> 00:30:07,440 Ejaaz: the protocol themselves. 490 00:30:07,720 --> 00:30:12,280 Ejaaz: So Google famously produced Android and open source the entire thing, 491 00:30:12,360 --> 00:30:13,780 Ejaaz: or rather was like, hey, this is for everyone. 492 00:30:13,940 --> 00:30:16,280 Ejaaz: But they ended up becoming the dominant company that controlled that. 493 00:30:16,380 --> 00:30:20,020 Ejaaz: And now they manage like, you know, millions and millions and millions of devices. 494 00:30:20,240 --> 00:30:23,580 Ejaaz: I think this is Anthropic's attempt to do the same thing. Now they're not planning 495 00:30:23,580 --> 00:30:27,900 Ejaaz: to release a new device, but it is yet another step for them to own the entire 496 00:30:27,900 --> 00:30:30,180 Ejaaz: operating stack for whatever AI becomes. 497 00:30:30,880 --> 00:30:33,560 Josh: Yeah. You said, you said this good word. I actually looked it up because I don't 498 00:30:33,560 --> 00:30:34,460 Josh: know what it means. Ephemeral. 499 00:30:34,660 --> 00:30:38,040 Josh: You said ephemeral operating system, which is something that lasts for a very 500 00:30:38,040 --> 00:30:41,040 Josh: short time is fleeting or transient, often describes things that are beautiful, 501 00:30:41,180 --> 00:30:42,580 Josh: but temporary, like a moment of feeling. 502 00:30:42,760 --> 00:30:46,400 Josh: So I guess the idea is that this is, it's dynamically generated, 503 00:30:46,660 --> 00:30:47,960 Josh: right? Like it's kind of built on the fly. 504 00:30:48,160 --> 00:30:53,300 Josh: It doesn't actually exist as a permanent fixture in our workspace, 505 00:30:53,300 --> 00:30:55,540 Josh: which is super interesting. This is cool. 506 00:30:55,740 --> 00:31:00,400 Josh: There is one use case that I particularly loved, which was the iMessage wrapped. 507 00:31:00,580 --> 00:31:04,440 Josh: I think one of my favorite parts of the year is Spotify wrapped when we can 508 00:31:04,440 --> 00:31:06,180 Josh: see all of the most played music. 509 00:31:06,380 --> 00:31:10,680 Josh: Everyone talks about how they're listening age. It's very fun because it has 510 00:31:10,680 --> 00:31:13,640 Josh: access to these more intimate personal data sets. 511 00:31:13,760 --> 00:31:18,180 Josh: It can actually use that data to generate these fun things like iMessage Wrapped, 512 00:31:18,280 --> 00:31:21,920 Josh: which is a code project that this woman did, which I thought was really fun. 513 00:31:21,980 --> 00:31:26,800 Josh: It showed how many messages were sent throughout the year, who the top messages were with. 514 00:31:27,600 --> 00:31:33,240 Josh: It showed heat graphs and charts of when you are most likely to send messages. 515 00:31:33,440 --> 00:31:33,540 Ejaaz: Relationships over time. 516 00:31:33,800 --> 00:31:37,200 Josh: Yeah, you could see relationships over time. When a relationship got hot, 517 00:31:37,280 --> 00:31:41,740 Josh: you were texting a lot. When it faded away, it's this really fun, exciting thing. 518 00:31:42,420 --> 00:31:45,960 Josh: Experiment. It shows grammar, who gets your best English, most formal versus 519 00:31:45,960 --> 00:31:49,080 Josh: most casual, your writing style over time. 520 00:31:49,160 --> 00:31:52,960 Josh: And you can really break it down because it has this deep understanding and 521 00:31:52,960 --> 00:31:57,220 Josh: intelligence, but also because it has this connection to your personal intimate data set. 522 00:31:57,280 --> 00:32:00,500 Josh: And I thought this was so much fun because it shows you even the heat chart 523 00:32:00,500 --> 00:32:03,140 Josh: of when you are most likely to be sending messages. Oh, I like this. 524 00:32:03,340 --> 00:32:03,780 Josh: Your dream dinner party. 525 00:32:03,940 --> 00:32:05,660 Ejaaz: The dream dinner parties is good. 526 00:32:05,880 --> 00:32:08,820 Josh: Very cool. Who's sitting at your ideas table? Who are your ride dies. 527 00:32:08,980 --> 00:32:10,080 Josh: Who's the industry dinner? 528 00:32:10,560 --> 00:32:14,500 Josh: It's a really fun example of a use case that I hope we see a lot more of. 529 00:32:14,860 --> 00:32:18,300 Josh: And again, the hardest problem with these is just figuring out how to extract value from it 530 00:32:18,570 --> 00:32:20,910 Josh: And whenever we get a cool example like this, it's awesome. It's like, 531 00:32:20,990 --> 00:32:22,470 Josh: okay, here, it's open source of my GitHub. 532 00:32:22,610 --> 00:32:26,010 Josh: You can go and copy it for yourself and go have fun. And this is a really awesome 533 00:32:26,010 --> 00:32:27,610 Josh: use case that I see for Cloud in particular. 534 00:32:28,110 --> 00:32:32,630 Ejaaz: As we're filming this episode, Josh, and we've given a bunch of tool demos and 535 00:32:32,630 --> 00:32:37,790 Ejaaz: examples right now, I'm realizing that we have made so much progress over the 536 00:32:37,790 --> 00:32:40,190 Ejaaz: last year when it was incredibly manual. 537 00:32:40,470 --> 00:32:44,090 Ejaaz: And now it's much, much, much less manual. But in my opinion, 538 00:32:44,270 --> 00:32:46,050 Ejaaz: it's still manual enough. 539 00:32:46,190 --> 00:32:49,230 Ejaaz: Do you know what I mean? and i think there's going to be a dissection 540 00:32:49,230 --> 00:32:52,030 Ejaaz: of audiences where you have the more kind of 541 00:32:52,030 --> 00:32:54,910 Ejaaz: i guess you can call them semi-technical folk that will spin up 542 00:32:54,910 --> 00:32:57,890 Ejaaz: a clawed bot or try out a new ai native thing 543 00:32:57,890 --> 00:33:00,590 Ejaaz: and to your earlier point like you should be proactively trying to 544 00:33:00,590 --> 00:33:03,530 Ejaaz: use these tools so you can kind of evolve your brain and thinking 545 00:33:03,530 --> 00:33:06,390 Ejaaz: into using these tools that you're not caught off guard but i 546 00:33:06,390 --> 00:33:09,030 Ejaaz: think eventually uh majority of people are just going 547 00:33:09,030 --> 00:33:12,150 Ejaaz: to use some kind of package system like google is like the perfect placement 548 00:33:12,150 --> 00:33:14,870 Ejaaz: for this right where it's like hey don't worry gemini is just going 549 00:33:14,870 --> 00:33:17,530 Ejaaz: to come to wherever you are and we've got you we're just going to give you 550 00:33:17,530 --> 00:33:20,630 Ejaaz: the features instead of you trying to figure out the features this is an example 551 00:33:20,630 --> 00:33:23,950 Ejaaz: of someone figuring out the feature and i'm actually more of an advocate for 552 00:33:23,950 --> 00:33:27,050 Ejaaz: people to go out and do that but yeah it just occurred to me that i think a 553 00:33:27,050 --> 00:33:29,570 Ejaaz: lot of people are just going to kind of take the lazy option and just wait for 554 00:33:29,570 --> 00:33:32,950 Ejaaz: that pre-packaging which isn't a bad thing because there's less security risk 555 00:33:32,950 --> 00:33:35,130 Ejaaz: it's probably going to be cheaper at that point it's probably going to be a 556 00:33:35,130 --> 00:33:36,230 Ejaaz: better user experience well 557 00:33:36,230 --> 00:33:39,130 Josh: It's also like what is the actual delta between the time that 558 00:33:39,130 --> 00:33:42,010 Josh: early adopters use it versus the time that the general public uses it 559 00:33:42,010 --> 00:33:46,390 Josh: and it's becoming increasingly small it used to be months to years now it's 560 00:33:46,390 --> 00:33:50,350 Josh: a matter of days i mean claude bot was exciting because it had access to your 561 00:33:50,350 --> 00:33:54,110 Josh: iMessage and you could talk with it and now we have a connection and a really 562 00:33:54,110 --> 00:33:57,270 Josh: cool use case that's built right into your claude desktop app without any of 563 00:33:57,270 --> 00:34:00,850 Josh: the technical know-how or installation and that was a matter of days the compression 564 00:34:00,850 --> 00:34:01,930 Josh: is happening very quickly. 565 00:34:01,930 --> 00:34:05,150 Ejaaz: Yeah i i agree with you on the software front i 566 00:34:05,720 --> 00:34:08,940 Ejaaz: maybe don't agree if I look at it from an investment angle. 567 00:34:09,080 --> 00:34:12,680 Ejaaz: Because to your earlier point, you knew Tesla was going to be a robot company, 568 00:34:12,840 --> 00:34:16,640 Ejaaz: a humanoid specific company six years ago with that video that you recorded, right? 569 00:34:16,780 --> 00:34:21,400 Ejaaz: So it's crazy to see how like a lot of these timelines are like collapsing for 570 00:34:21,400 --> 00:34:25,520 Ejaaz: some things, but in other things, they're not. Or maybe it's just lagging. I don't know. 571 00:34:25,720 --> 00:34:28,840 Josh: I think people mostly understand the trajectory. The question is whether they 572 00:34:28,840 --> 00:34:32,600 Josh: can deliver on that or not. And in the case of Tesla, they have actually delivered on it. 573 00:34:34,100 --> 00:34:37,040 Josh: In don't have set timelines they don't have set delivery dates 574 00:34:37,040 --> 00:34:40,200 Josh: it's kind of a fuzzy thing but everyone is generally working towards the same 575 00:34:40,200 --> 00:34:43,900 Josh: goals so what we're seeing is just incremental progress some faster than others 576 00:34:43,900 --> 00:34:47,920 Josh: um and xai right here is another example of incremental progress that perhaps 577 00:34:47,920 --> 00:34:52,260 Josh: leaves a little bit to be desired here the uh the new grok imagine model ijaz 578 00:34:52,260 --> 00:34:54,780 Josh: let's talk about this for a sec okay. 579 00:34:54,780 --> 00:34:59,660 Ejaaz: I'm gonna be honest because i'm a massive elon bull that that is no secret on this show but 580 00:35:00,170 --> 00:35:04,110 Ejaaz: And the video and image models from Grok, Grok Imagine specifically, 581 00:35:04,410 --> 00:35:06,510 Ejaaz: have kind of been subpar. 582 00:35:06,750 --> 00:35:11,210 Ejaaz: Like when I look at like what Mid Journey is putting out, when I look at like 583 00:35:11,210 --> 00:35:14,570 Ejaaz: what VO3 is putting out, when I look at like what Runway is putting out, 584 00:35:14,830 --> 00:35:21,090 Ejaaz: there's just a higher quality and definition that Grok just isn't giving me. 585 00:35:21,250 --> 00:35:25,290 Ejaaz: And that wouldn't even matter because they are, again, like only a two and a half year old startup. 586 00:35:25,490 --> 00:35:28,490 Ejaaz: But it's the fact that Elon keeps on shilling it like it's the next best 587 00:35:28,490 --> 00:35:31,590 Ejaaz: thing and i'm like bro it's not right anyway the 588 00:35:31,590 --> 00:35:34,590 Ejaaz: the news here is we have a new grok imagine 589 00:35:34,590 --> 00:35:37,890 Ejaaz: api uh which is basically their latest um 590 00:35:37,890 --> 00:35:40,750 Ejaaz: image or video model and as you can see it is 591 00:35:40,750 --> 00:35:43,730 Ejaaz: actually quite impressive and i might be being a bit of a brat here because 592 00:35:43,730 --> 00:35:48,470 Ejaaz: this is uh next level compared to like what we had last year and the cool part 593 00:35:48,470 --> 00:35:52,750 Ejaaz: is it's cheaper it's quicker to generate um and it's just simply a better model 594 00:35:52,750 --> 00:35:58,630 Ejaaz: that is available everywhere um whether this inspires me to use it isn't really the point. 595 00:35:58,750 --> 00:36:02,770 Ejaaz: I'm sure there are people that are involved in video production that will find this way more useful. 596 00:36:03,410 --> 00:36:06,270 Ejaaz: Well, actually, let me ask you, Josh, as someone that plays around with a lot 597 00:36:06,270 --> 00:36:08,410 Ejaaz: of this stuff, is this a tool that you would use? Are you impressed by it? 598 00:36:08,830 --> 00:36:12,050 Josh: No, they didn't need to release this. This didn't need to happen. 599 00:36:12,170 --> 00:36:15,490 Josh: They could have just shelved this and kept going with Grok 4.2 and then working on Grok 5. 600 00:36:16,110 --> 00:36:19,230 Josh: Had this been released a year ago, it would have been amazing. It's a great image 601 00:36:19,230 --> 00:36:22,230 Josh: video generation model uh given today's standards 602 00:36:22,230 --> 00:36:26,270 Josh: it's not and while it's an improvement for the xai team it is not an actual 603 00:36:26,270 --> 00:36:30,190 Josh: improvement pushing the frontier of video or photos forward so while i look 604 00:36:30,190 --> 00:36:32,990 Josh: forward to the next iteration of this as they work towards their game development 605 00:36:32,990 --> 00:36:36,310 Josh: engine and building these graphics in real time this is kind of an incremental 606 00:36:36,310 --> 00:36:40,010 Josh: progress update that really didn't need to come out i don't think many people 607 00:36:40,010 --> 00:36:41,410 Josh: are going to be super excited about this. 608 00:36:41,410 --> 00:36:44,830 Ejaaz: Well what if i framed it in a different way josh what if i told you that this 609 00:36:44,830 --> 00:36:46,850 Ejaaz: was actually never meant for video production 610 00:36:47,350 --> 00:36:51,970 Ejaaz: but rather game development. So one of our predictions at the end of last year 611 00:36:51,970 --> 00:36:55,950 Ejaaz: was that XAI is going to release one of the best gaming AI models, 612 00:36:56,110 --> 00:36:59,950 Ejaaz: which will allow for real-time game immersion and gameplay, which is kind of 613 00:36:59,950 --> 00:37:02,290 Ejaaz: like some of the demos that we're seeing on this video right here. 614 00:37:02,850 --> 00:37:05,730 Ejaaz: And I'm wondering whether that might be their angle. Maybe they're just not 615 00:37:05,730 --> 00:37:06,910 Ejaaz: going for video production at all. 616 00:37:07,210 --> 00:37:10,950 Josh: Yeah, and I suspect that's probably why Elon is abnormally excited about this, 617 00:37:11,030 --> 00:37:14,330 Josh: because he understands the vision for this and he understands where 618 00:37:14,330 --> 00:37:17,390 Josh: on the trajectory this incremental like data point falls 619 00:37:17,390 --> 00:37:20,110 Josh: whereas we're seeing it at face value and being like 620 00:37:20,110 --> 00:37:23,130 Josh: this isn't that impressive we've seen this before but gaming 621 00:37:23,130 --> 00:37:26,270 Josh: is certainly where they want to go gaming and animation and graphics and 622 00:37:26,270 --> 00:37:29,310 Josh: building real-time three-dimensional spaces which are going to compete with 623 00:37:29,310 --> 00:37:33,470 Josh: google's genie the world building model so i'm looking forward to that war as 624 00:37:33,470 --> 00:37:37,350 Josh: it comes along because logan kilpatrick from google is actually teasing genie 625 00:37:37,350 --> 00:37:41,010 Josh: 3 this week which hopefully we'll be getting very very soon and we'll have a 626 00:37:41,010 --> 00:37:45,030 Josh: real comparison of a true frontier your world-building model very soon. 627 00:37:45,610 --> 00:37:50,930 Ejaaz: Mm-hmm. In other news, we put out an episode earlier this week. 628 00:37:51,070 --> 00:37:55,830 Ejaaz: I actually think it's our last episode on this new Chinese model, Kimi K 2.5. 629 00:37:56,270 --> 00:38:01,870 Ejaaz: Now, we're not going to get into it on this episode, but the TLDR is it is a very impressive model. 630 00:38:02,090 --> 00:38:07,370 Ejaaz: It's 100% free, open source. You can download it, amend the model weights, and run it locally. 631 00:38:07,390 --> 00:38:11,070 Ejaaz: But it might cost you a few very expensive bits of hardware to be able to do that. 632 00:38:11,270 --> 00:38:14,250 Ejaaz: But you can access it for free on their website. and the coolest part about 633 00:38:14,250 --> 00:38:18,210 Ejaaz: it is you can kind of like convert video recordings into live production ready 634 00:38:18,210 --> 00:38:22,870 Ejaaz: apps but there was an update that we wanted to make based off of this. 635 00:38:23,290 --> 00:38:27,690 Josh: Yeah so we actually said that it would be immensely difficult to run this locally 636 00:38:27,690 --> 00:38:31,290 Josh: on a machine because it would have required a tremendous amount of compute and 637 00:38:31,290 --> 00:38:34,410 Josh: the reality is is that someone actually got this running with far less than we thought. 638 00:38:34,630 --> 00:38:38,650 Josh: You don't need a couple GB200s or H100s or any cutting-edge NVIDIA technology. 639 00:38:38,930 --> 00:38:43,010 Josh: This person actually did it with two Macs on his desktop. 640 00:38:43,250 --> 00:38:47,690 Josh: So he bought two M3 Ultra Mac Studios and strung them together, 641 00:38:47,690 --> 00:38:49,970 Josh: which costs about $20,000 USD. 642 00:38:50,210 --> 00:38:54,890 Josh: And he was able to generate 24 tokens per second using these two computers and 643 00:38:54,890 --> 00:38:56,550 Josh: the new Kimi K 2.5 Pro model. 644 00:38:56,690 --> 00:39:03,870 Josh: So I think that is a testament to how accessible a model this good is and essentially making it free. 645 00:39:03,990 --> 00:39:07,550 Josh: So now, it's funny, tying it back to the Tesla autopilot example, 646 00:39:07,730 --> 00:39:12,290 Josh: where you can buy a license currently for $8,000 and that gives you lifetime 647 00:39:12,290 --> 00:39:16,470 Josh: free miles, or you can do a subscription for $20,000. 648 00:39:16,470 --> 00:39:21,630 Josh: Now you can basically get a lifetime membership to pseudo AGI to one of the 649 00:39:21,630 --> 00:39:26,790 Josh: highest level frontier AI models. And you could run that for free through infinity 650 00:39:26,790 --> 00:39:28,670 Josh: at the cost of these two localized models. 651 00:39:28,770 --> 00:39:31,710 Josh: And I think that's a really impressive and fun breakthrough where, I mean, 652 00:39:31,890 --> 00:39:35,270 Josh: assumedly the server costs to run this would be significantly lower and you 653 00:39:35,270 --> 00:39:38,190 Josh: could just run these open source models now that are very much at the frontier 654 00:39:38,190 --> 00:39:43,270 Josh: for a very reasonable cost relative to what some people are paying to use like 655 00:39:43,270 --> 00:39:48,110 Josh: an API like Opus 4.5 which is charging you $25 per million tokens. 656 00:39:48,730 --> 00:39:53,690 Ejaaz: Yeah, I mean, the funniest part about this model release has got nothing to 657 00:39:53,690 --> 00:39:58,350 Ejaaz: do with the model, but the fact that you can never rest easy because your competition 658 00:39:58,350 --> 00:40:00,070 Ejaaz: is always cooking up something better. 659 00:40:00,270 --> 00:40:04,290 Ejaaz: And literally within hours after recording that episode of Kimi K2, 660 00:40:04,390 --> 00:40:09,130 Ejaaz: Google pretty much announced an identical copy or functionality where they can 661 00:40:09,130 --> 00:40:10,990 Ejaaz: turn, they call it agentic vision, 662 00:40:11,330 --> 00:40:15,650 Ejaaz: but they can turn any screen recording into a production ready application. 663 00:40:15,650 --> 00:40:20,130 Ejaaz: Now, they don't state that specifically on here, but that's technically what it should be able to do. 664 00:40:20,650 --> 00:40:26,490 Ejaaz: And it just is a testament to how quickly and almost synchronously these AI 665 00:40:26,490 --> 00:40:27,550 Ejaaz: labs are working together. 666 00:40:27,690 --> 00:40:32,030 Ejaaz: Now, in my opinion, it's more impressive for the Kimi K2.5 team, 667 00:40:32,170 --> 00:40:35,230 Ejaaz: the Moonshot Labs team, because they have less infrastructure. 668 00:40:35,410 --> 00:40:37,930 Ejaaz: Although someone made the point in our episode in the comments, 669 00:40:38,050 --> 00:40:41,130 Ejaaz: Josh, I don't know if you saw this, that they have a much, much cheaper energy. 670 00:40:41,250 --> 00:40:45,070 Ejaaz: So I don't know how that translates necessarily. Maybe they just run these servers 671 00:40:45,070 --> 00:40:48,530 Ejaaz: for way, way longer and it just costs them less. So it kind of like matches out. 672 00:40:48,650 --> 00:40:52,690 Ejaaz: But it's interesting that the open source teams and the centralized teams are 673 00:40:52,690 --> 00:40:54,450 Ejaaz: kind of like moving at parity right now. 674 00:40:54,570 --> 00:40:57,030 Ejaaz: It's cool that the open source teams have even caught up. But yeah, 675 00:40:57,190 --> 00:41:01,390 Ejaaz: Google's come up with the same thing if you want to try it and use a Western AI lab product. 676 00:41:01,750 --> 00:41:06,870 Josh: It's amazing how quick they were to respond. And also that comment is very correct. 677 00:41:07,710 --> 00:41:10,970 Josh: Chinese electricity costs significantly less than the US electricity does because 678 00:41:10,970 --> 00:41:15,430 Josh: they have so much more of it. So the cost per token, the cost to train these 679 00:41:15,430 --> 00:41:19,850 Josh: models is significantly lower because of that cost per kilowatt as it relates 680 00:41:19,850 --> 00:41:21,950 Josh: to just generating all of this energy. 681 00:41:22,290 --> 00:41:28,090 Josh: Yeah, yeah. But really nice reply from the commenter, but also Google in a matter of seemingly days. 682 00:41:28,230 --> 00:41:31,790 Josh: It feels like they had this ready to go and they were like, oh, 683 00:41:31,930 --> 00:41:35,330 Josh: all right. Well, I guess we should probably just release this now because Kimmy did something similar. 684 00:41:35,490 --> 00:41:37,910 Josh: And now here we are. Agentic vision for all. 685 00:41:38,640 --> 00:41:45,740 Ejaaz: Well, if Moonshot AI, KimiK 2.5, is the model example of spending your resources widely, 686 00:41:46,380 --> 00:41:51,620 Ejaaz: OpenAI probably needs to be on the opposite end of that spectrum with the breaking 687 00:41:51,620 --> 00:41:55,920 Ejaaz: news that they are raising a $100 billion round, 688 00:41:56,240 --> 00:42:01,600 Ejaaz: which is just gargantuan, to value them at around $730 billion. 689 00:42:01,740 --> 00:42:03,840 Ejaaz: I've seen $850 as well. 690 00:42:04,740 --> 00:42:09,780 Ejaaz: I can only imagine that they are raising this amount of money to blow it all 691 00:42:09,780 --> 00:42:15,140 Ejaaz: on training various different models, not just one model, or to kind of build 692 00:42:15,140 --> 00:42:18,400 Ejaaz: out a new social network that we saw on the timeline. 693 00:42:18,600 --> 00:42:21,300 Ejaaz: Apparently, they're building that and their hardware device. 694 00:42:21,420 --> 00:42:24,840 Ejaaz: And what else am I missing, Josh? We've got the Sora TikTok app competitor. 695 00:42:25,180 --> 00:42:28,680 Ejaaz: It just seems like at this point, and I know this might be a negative take, 696 00:42:29,140 --> 00:42:33,180 Ejaaz: they've been raising so much money and endlessly blowing it on myriad different 697 00:42:33,180 --> 00:42:34,900 Ejaaz: things. so they're spreading themselves very thin. 698 00:42:35,240 --> 00:42:38,700 Ejaaz: This doesn't signal confidence to me, actually. It actually signals that they 699 00:42:38,700 --> 00:42:40,920 Ejaaz: are kind of being reckless at this point. 700 00:42:41,300 --> 00:42:44,880 Josh: Yeah, it's probably helpful to look at past fundraising history to get a better 701 00:42:44,880 --> 00:42:49,600 Josh: perspective of how much money this actually is in a way that feels almost existential. 702 00:42:49,820 --> 00:42:52,900 Josh: There's so many vested interests in OpenAI succeeding. 703 00:42:53,060 --> 00:42:56,760 Josh: If they don't, what does that look like for the rest of the industry? 704 00:42:56,900 --> 00:42:59,540 Josh: That's an important question to just kind of sit on for a little bit. 705 00:42:59,660 --> 00:43:03,940 Josh: But if you go back to the funding rounds, even April 2023, they were raising still in the millions. 706 00:43:04,120 --> 00:43:08,320 Josh: They raised $300 million at a $27 billion valuation. Then in October 2024, 707 00:43:08,760 --> 00:43:14,340 Josh: they raised 20 times that, $6.6 billion at $157 billion valuation. 708 00:43:14,620 --> 00:43:21,620 Josh: And now they're raising eight times that at $40 billion at a $300 billion post-money valuation. 709 00:43:21,760 --> 00:43:24,540 Josh: Now, this is interesting because Ejaz, earlier in the episode, 710 00:43:24,760 --> 00:43:30,100 Josh: we mentioned SpaceX was looking to raise $50 billion at about a $1.5 trillion 711 00:43:30,100 --> 00:43:34,000 Josh: valuation. So they're raising 20% more. 712 00:43:34,500 --> 00:43:37,480 Josh: At 500% the market cap. 713 00:43:37,840 --> 00:43:42,020 Josh: And we're starting to see like OpenAI is giving up a lot of equity here. 714 00:43:42,160 --> 00:43:45,860 Josh: They're really giving a lot of vested and controlled interest to a lot of companies 715 00:43:45,860 --> 00:43:47,880 Josh: that may not want that liability. 716 00:43:48,080 --> 00:43:52,800 Josh: And I wonder if this is a strategy in where he's trying to, Sam and the OpenAI 717 00:43:52,800 --> 00:43:56,700 Josh: team are just trying to get as many vested interests as possible in their success 718 00:43:56,700 --> 00:43:58,900 Josh: in a way that they become too big to fail. 719 00:43:59,060 --> 00:44:02,040 Josh: Where so many people have so much money tied up in this thing that they own 720 00:44:02,040 --> 00:44:07,840 Josh: a lot of, that they are forced to give whatever it takes to make this work because 721 00:44:07,840 --> 00:44:11,540 Josh: the downside effect impacts the entire industry in such a large way. 722 00:44:11,780 --> 00:44:14,620 Josh: And I can't help but read these headlines and not think that. 723 00:44:14,940 --> 00:44:17,460 Josh: I mean, this is just outrageous. They have a tremendous amount of debt. 724 00:44:17,580 --> 00:44:20,880 Josh: They haven't made a single profit. They're trying all these things like ads. 725 00:44:21,000 --> 00:44:22,180 Josh: They're trying to take revenue cuts. 726 00:44:22,340 --> 00:44:25,000 Josh: They're trying social media things, seeing what sticks. 727 00:44:25,180 --> 00:44:27,660 Josh: And it's giving it's giving desperate vibes. 728 00:44:27,660 --> 00:44:30,780 Ejaaz: Yeah i mean i have a simple take 729 00:44:30,780 --> 00:44:34,320 Ejaaz: which is uh they are on track to make and 730 00:44:34,320 --> 00:44:37,440 Ejaaz: also burn 20 billion dollars of capex 731 00:44:37,440 --> 00:44:40,480 Ejaaz: this year i think they need someone to foot the bill uh 732 00:44:40,480 --> 00:44:44,660 Ejaaz: to also spin up a bunch of the compute data centers 1.4 trillion dollars worth 733 00:44:44,660 --> 00:44:49,760 Ejaaz: um over the next three years and i think they're panicking that they didn't 734 00:44:49,760 --> 00:44:53,540 Ejaaz: spend enough time and resources on building out a better coding model to compete 735 00:44:53,540 --> 00:44:56,480 Ejaaz: with anthropic which is kind of taking over all of enterprise and eating out 736 00:44:56,480 --> 00:44:57,480 Ejaaz: their market share right now. 737 00:44:58,060 --> 00:45:01,400 Ejaaz: And Neil's is spending it on other stuff like social media apps and stuff like that. 738 00:45:02,170 --> 00:45:05,610 Ejaaz: Uh, my take is, listen, I hope they raise the round. Good for them. 739 00:45:05,870 --> 00:45:10,410 Ejaaz: Uh, but I don't trust Sam and the team to spend it wisely for now. 740 00:45:10,490 --> 00:45:12,430 Ejaaz: And that might be super bearish take. 741 00:45:12,590 --> 00:45:15,750 Ejaaz: I'm willing to be proven wrong, but that's just where my head's at. 742 00:45:16,170 --> 00:45:18,730 Josh: Well, I feel like we have to root for them because we have no choice. 743 00:45:18,890 --> 00:45:24,090 Josh: If OpenAI goes down, so does everything with it. So we're rooting for you, Sam and the OpenAI team. 744 00:45:24,250 --> 00:45:27,410 Josh: I hope everything works. I hope this money goes to good use. 745 00:45:27,630 --> 00:45:32,590 Josh: And I think that probably wraps up our roundup for this week. of all the top news in AI. 746 00:45:32,870 --> 00:45:36,510 Josh: As you enter the weekend, you can now feel properly satiated that you are fully 747 00:45:36,510 --> 00:45:40,210 Josh: up to date and aware of all the hottest news that you needed to this week. 748 00:45:40,350 --> 00:45:43,570 Josh: We released a few pretty great episodes. They're doing really well on Kimmy 749 00:45:43,570 --> 00:45:46,490 Josh: K2 and Claudebot. So if you haven't had a chance to go listen to those, 750 00:45:46,850 --> 00:45:48,090 Josh: please go and check them out. 751 00:45:48,270 --> 00:45:53,370 Josh: The ask for you perhaps this week is to try out the Gemini Chrome extension 752 00:45:53,370 --> 00:45:56,550 Josh: situation and use it to subscribe to our podcast. 753 00:45:56,690 --> 00:45:59,650 Josh: That's like a fun little demo. You can say, hey, subscribe to my favorite new 754 00:45:59,650 --> 00:46:03,250 Josh: AI in Frontier Technology podcast, and it will actually go and click the button for you. 755 00:46:03,510 --> 00:46:07,470 Josh: And then if you go on Claude and you connect your iMessage, you could ask it, 756 00:46:07,890 --> 00:46:11,090 Josh: send the Limitless podcast to 10 of my best friends and figure out which are 757 00:46:11,090 --> 00:46:13,930 Josh: my 10 best friends who would be most interested and share the link with them. 758 00:46:14,050 --> 00:46:17,190 Josh: And I think these are really great use cases that you can try. And then get it to spin. 759 00:46:17,190 --> 00:46:20,710 Ejaaz: Up a number of other Google accounts and do the exact same thing, 760 00:46:20,830 --> 00:46:22,910 Ejaaz: right, Josh? Let's just create an entire platform. 761 00:46:23,350 --> 00:46:26,070 Josh: Exactly. But I mean, this is a bit of a longer episode. So if you made it this 762 00:46:26,070 --> 00:46:28,690 Josh: long, thank you so much. These are the real ones that are still here at the end. 763 00:46:28,970 --> 00:46:31,710 Josh: Thank you for listening. We will be back again next week with plenty of new 764 00:46:31,710 --> 00:46:35,290 Josh: episodes, lots of new news to cover, and a lot of exciting topics that we will 765 00:46:35,290 --> 00:46:38,490 Josh: keep on keeping you up to date with. So thank you for watching, 766 00:46:38,550 --> 00:46:39,490 Josh: and we'll see you guys next week. 767 00:46:40,050 --> 00:46:40,610 Ejaaz: See you guys.