1 00:00:00,020 --> 00:00:04,760 Josh: This week, Meta released MuseSpark, its new flagship AI model built from scratch 2 00:00:04,760 --> 00:00:06,920 Josh: by Alexander Wang's Super Intelligence Lab. 3 00:00:07,160 --> 00:00:10,600 Josh: And noteworthy for the very first time, it's not open source. 4 00:00:10,900 --> 00:00:15,180 Josh: They closed source the entire model, but the model isn't what caught our attention, it's distribution. 5 00:00:15,800 --> 00:00:19,280 Josh: MuseSpark is rolling out across Facebook, Instagram, WhatsApp, 6 00:00:19,560 --> 00:00:22,800 Josh: and the Rayvan Metaglasses to over 3 billion users. 7 00:00:22,960 --> 00:00:26,980 Josh: That's more than every other AI company on Earth has. This is a really big deal. 8 00:00:27,520 --> 00:00:30,080 Josh: Meta's been calling us personal superintelligence. You log in with your Facebook 9 00:00:30,080 --> 00:00:33,640 Josh: or your Instagram account, and then the model pulls from your social graph that 10 00:00:33,640 --> 00:00:37,040 Josh: it has based on the past, what, decade that you've been using these applications. 11 00:00:37,260 --> 00:00:41,420 Josh: So no other model in the world has the type of information that this new meta 12 00:00:41,420 --> 00:00:44,320 Josh: AI model is going to have. Now, here's where it gets interesting. 13 00:00:44,460 --> 00:00:48,500 Josh: Meta is also developing another research project in parallel called Tribe V2, 14 00:00:48,840 --> 00:00:53,500 Josh: something that almost nobody's talking about, but it's trained on brain scans from human beings. 15 00:00:53,500 --> 00:00:57,980 Josh: And it understands how you react to certain imagery in certain videos so that 16 00:00:57,980 --> 00:01:01,280 Josh: it knows when a video or a piece of content is going to light up part of your 17 00:01:01,280 --> 00:01:03,620 Josh: brain that makes you interested and engaged. 18 00:01:03,780 --> 00:01:06,500 Josh: Now, they're claiming to do both of these separately and in parallel, 19 00:01:06,500 --> 00:01:09,780 Josh: but you have to ask the question, what happens when these things merge together? 20 00:01:09,880 --> 00:01:13,160 Josh: What happens when Meta knows everything about you, they know everything that's 21 00:01:13,160 --> 00:01:16,880 Josh: going on in your brain, and they're able to serve the absolute best content possible. 22 00:01:16,960 --> 00:01:20,480 Josh: They become the 21st century drug dealers through this artificial intelligence. 23 00:01:20,720 --> 00:01:21,760 Josh: And maybe that's a doomer take. 24 00:01:22,080 --> 00:01:24,520 Josh: I know, Ejaz, you have a very different take on this, perhaps. 25 00:01:24,880 --> 00:01:27,420 Josh: We're going to talk about that and a lot of other things on today's episode, 26 00:01:27,600 --> 00:01:31,680 Josh: including Anthropik's not-so-real annual recurring revenue. We have some news 27 00:01:31,680 --> 00:01:35,720 Josh: around OpenAI and then some cool robotic arms that we actually had the guest on a few months ago. 28 00:01:35,880 --> 00:01:38,600 Josh: He is back with the final products. There's a lot going on this week, 29 00:01:38,700 --> 00:01:41,580 Josh: but first, Meta, Ejaz. Meta's a pretty big story this week. 30 00:01:41,700 --> 00:01:45,740 Ejaaz: They released MuseSpark, which is the first Frontier AI model that they've released 31 00:01:45,740 --> 00:01:51,720 Ejaaz: in over a year now, which is a crazy amount of time in the AI world. Now, I'm going to 32 00:01:52,090 --> 00:01:57,870 Ejaaz: I'm thinking about the story arc of Meta, and it's hard to be over-enthusiastic about this model. 33 00:01:58,310 --> 00:02:03,570 Ejaaz: Last year, Zuck burned around $75 billion on AI CapEx. 34 00:02:03,730 --> 00:02:08,770 Ejaaz: He then fired 600 of his AI staff, and then hired 150 more staff, 35 00:02:08,970 --> 00:02:11,250 Ejaaz: spending $25 billion to do so. 36 00:02:11,430 --> 00:02:16,310 Ejaaz: $15 billion, of which he used to hire one guy, Alex Wang, who leads Meta Super 37 00:02:16,310 --> 00:02:21,430 Ejaaz: Intelligence Labs, who helped build this model. So I'm expecting big things at this point. 38 00:02:21,830 --> 00:02:28,170 Ejaaz: Unfortunately, the model underperforms Claude Opus 4.6 and GPT 5.4 across the 39 00:02:28,170 --> 00:02:30,830 Ejaaz: most important benchmarks, which is coding and reasoning. 40 00:02:31,030 --> 00:02:34,290 Ejaaz: So it's easy to be bearish about this model, but I dug into it more. 41 00:02:34,390 --> 00:02:38,190 Ejaaz: And what I realized is that's not what Zuck or Meta is going after at all. 42 00:02:38,330 --> 00:02:41,790 Ejaaz: They're going after what I'm dubbing personal AGI. 43 00:02:42,310 --> 00:02:47,110 Ejaaz: Personal AGI is basically super intelligent AI that's trained on your own personal 44 00:02:47,110 --> 00:02:51,550 Ejaaz: data. It'll become an AI assistant that basically lives and breathes you. 45 00:02:51,770 --> 00:02:55,830 Ejaaz: So how is Meta going to be able to pull this off versus all the other competitors 46 00:02:55,830 --> 00:02:56,910 Ejaaz: who are leagues ahead of them? 47 00:02:57,130 --> 00:03:00,410 Ejaaz: Well, they have something that none of them have, which is the personal data 48 00:03:00,410 --> 00:03:05,430 Ejaaz: and very incriminating data of 3.4 billion daily active users. 49 00:03:05,550 --> 00:03:11,230 Ejaaz: So not only do they have a crap ton of data, they also have the data being refreshed every single day. 50 00:03:11,350 --> 00:03:14,890 Ejaaz: Now, I'm personally not a user of Facebook, but I use WhatsApp. 51 00:03:14,890 --> 00:03:17,490 Ejaaz: I use Instagram every single day. 52 00:03:17,670 --> 00:03:21,950 Ejaaz: So they have already collected a bunch of data around what type of conversations 53 00:03:21,950 --> 00:03:26,050 Ejaaz: I'm having, what type of media that I'm liking, what kind of memes that I enjoy. 54 00:03:26,250 --> 00:03:30,790 Ejaaz: And they can use this to engineer the perfectly crafted AI assistant that no 55 00:03:30,790 --> 00:03:32,870 Ejaaz: other lab can do. And this is what they're doing. 56 00:03:33,030 --> 00:03:36,270 Ejaaz: So the benchmarks that do matter, if you look at this table over here, 57 00:03:36,570 --> 00:03:40,930 Ejaaz: they excel or this model excels in visual reasoning and multimodality. 58 00:03:40,930 --> 00:03:47,330 Ejaaz: So it can look at a picture and understand it and perceive it better than Opus 4.6 and ChatGPT 5.4. 59 00:03:47,450 --> 00:03:50,590 Ejaaz: And so if they're going to end up building a model that ultimately can feed 60 00:03:50,590 --> 00:03:53,950 Ejaaz: you a better algorithm or feed you the best content, this is going to be the 61 00:03:53,950 --> 00:03:55,450 Ejaaz: model that enables that. And that's what they're doing. 62 00:03:55,840 --> 00:03:58,740 Josh: Yeah, the personal AGI, it's a good way to put it. I don't think this needs 63 00:03:58,740 --> 00:04:02,600 Josh: to be everything for everyone. This just needs to be better than nothing. 64 00:04:02,800 --> 00:04:05,880 Josh: Because the reality is that a lot of people still aren't using these AI tools 65 00:04:05,880 --> 00:04:06,640 Josh: in their day-to-day life. 66 00:04:06,880 --> 00:04:08,980 Josh: What I found interesting is the examples that they showcase. 67 00:04:09,120 --> 00:04:12,220 Josh: Now, we don't actually know what it looks like applied to these social media 68 00:04:12,220 --> 00:04:14,240 Josh: networks. We're not sure what that integration is going to look like. 69 00:04:14,300 --> 00:04:17,220 Josh: But they showed a few examples in the blog post that they shared, 70 00:04:17,340 --> 00:04:19,140 Josh: which is it has the visual intelligence. 71 00:04:19,140 --> 00:04:21,340 Josh: When you point your camera at something like a refrigerator, 72 00:04:21,540 --> 00:04:25,520 Josh: say, it knows what's in there. it can calculate the amount of protein in that 73 00:04:25,520 --> 00:04:29,140 Josh: block of cheese or the amount of calories in that hamburger that's sitting there. 74 00:04:29,320 --> 00:04:30,820 Josh: And I think that's really interesting. 75 00:04:31,120 --> 00:04:34,840 Josh: They had another example of someone who was doing yoga and you could actually 76 00:04:34,840 --> 00:04:39,020 Josh: analyze the positions of the arms and the legs and the body and you could see 77 00:04:39,020 --> 00:04:40,600 Josh: if they were doing it properly. 78 00:04:40,860 --> 00:04:44,280 Josh: So the visual element of it, I think the native multimodality is something that's 79 00:04:44,280 --> 00:04:45,220 Josh: pretty noteworthy and interesting. 80 00:04:45,620 --> 00:04:48,160 Josh: Yeah, here's the example of the food. Here's the yoga pose. 81 00:04:48,660 --> 00:04:51,220 Josh: I'm not sure what that looks like fully integrated into... 82 00:04:52,160 --> 00:04:57,080 Josh: Our social media experiences today, outside of what could be just like a much stronger algorithm. 83 00:04:57,260 --> 00:05:00,680 Josh: But they did release meta.ai, which is a website that you can go to to actually 84 00:05:00,680 --> 00:05:02,240 Josh: try this out yourself and run some tests. 85 00:05:02,380 --> 00:05:05,600 Josh: So I would encourage anyone who's curious, who wants early access to the model 86 00:05:05,600 --> 00:05:09,760 Josh: before it hits your Instagram feeds or WhatsApp to go actually on the website 87 00:05:09,760 --> 00:05:11,920 Josh: and check it out for yourself. It's available for anyone with an account. 88 00:05:12,500 --> 00:05:15,860 Ejaaz: Yeah. And it's been a labor of love for Zuck. I think he famously said on an 89 00:05:15,860 --> 00:05:22,160 Ejaaz: interview about a month ago that he is willing to spend the entire wallet on 90 00:05:22,160 --> 00:05:27,540 Ejaaz: AI CapEx and training the best AI model versus risking losing out on this race entirely. 91 00:05:27,740 --> 00:05:31,560 Ejaaz: He would rather make an incredibly expensive mistake than produce a subtale model. 92 00:05:31,760 --> 00:05:34,820 Ejaaz: They spent the last nine months pre-training this model. Fun fact, 93 00:05:34,980 --> 00:05:36,300 Ejaaz: typically pre-training takes 94 00:05:36,300 --> 00:05:38,920 Ejaaz: around a couple of months and then you kind of go into post-training. 95 00:05:39,180 --> 00:05:41,820 Ejaaz: So they've invested a lot of time, money and heart into this. 96 00:05:42,120 --> 00:05:45,740 Ejaaz: So that's why I'm kind of like, I'm kind of split as to why this model isn't 97 00:05:45,740 --> 00:05:47,080 Ejaaz: as good enough. They delayed it. 98 00:05:47,380 --> 00:05:53,000 Ejaaz: They delayed its own launch about a month as well. But the CapEx spend keeps on going. 99 00:05:53,280 --> 00:05:58,220 Ejaaz: They today announced a partnership with CoreWeave to the tune of $21 billion. 100 00:05:58,560 --> 00:06:01,240 Ejaaz: So they're just pushing ahead with all the data center type stuff. 101 00:06:01,400 --> 00:06:06,660 Ejaaz: But you mentioned a model earlier on, Josh, which kind of went viral a couple of weeks ago. 102 00:06:06,820 --> 00:06:10,800 Ejaaz: And this is the scary part of this entire story. I would love to think that 103 00:06:10,800 --> 00:06:16,020 Ejaaz: Meta is in my best interests and wants me to become a better productive human. 104 00:06:16,220 --> 00:06:20,800 Ejaaz: But stuff like this, this new model, Tribe V2, which basically reads your brain 105 00:06:20,800 --> 00:06:25,260 Ejaaz: signals and predicts what type of content is going to stimulate different parts 106 00:06:25,260 --> 00:06:29,640 Ejaaz: of your brain and make you more engaged in certain types of content, doesn't give me hope. 107 00:06:29,860 --> 00:06:33,880 Ejaaz: Because presumably, they're going to use this model to understand what types 108 00:06:33,880 --> 00:06:37,780 Ejaaz: of content stimulate different people's appetites for videos, 109 00:06:38,040 --> 00:06:39,840 Ejaaz: reels, memes, or whatever that might be. 110 00:06:40,020 --> 00:06:44,000 Ejaaz: And then AI generate this exact amount of content that they can feed you in 111 00:06:44,000 --> 00:06:45,720 Ejaaz: the timeline, because they already have the distribution, right? 112 00:06:45,740 --> 00:06:48,620 Ejaaz: They already have the social media platforms, they can create and perfect the 113 00:06:48,620 --> 00:06:51,020 Ejaaz: content and own that without relying on creators. 114 00:06:51,160 --> 00:06:54,520 Ejaaz: They now have a recursive dopamine loop that they can just trap you in. 115 00:06:54,860 --> 00:06:56,500 Ejaaz: Doesn't give me hope. Maybe that's a do my take. 116 00:06:56,940 --> 00:07:00,980 Josh: Well, I'm curious what the stated intention of this model is, right? 117 00:07:01,060 --> 00:07:05,040 Josh: It's like, why are you trying to understand how the brain reacts to certain 118 00:07:05,040 --> 00:07:07,020 Josh: type of content? And sure, it's interesting. 119 00:07:07,240 --> 00:07:10,080 Josh: I know they're working on their metaglasses and they have that like wrist strap 120 00:07:10,080 --> 00:07:13,280 Josh: that's kind of like a neural interface that detects your hand movements and 121 00:07:13,280 --> 00:07:16,620 Josh: understand this information is probably more helpful to their future hardware products. 122 00:07:16,760 --> 00:07:19,700 Josh: But also, I mean, the clear and obvious use case 123 00:07:19,700 --> 00:07:23,060 Josh: for this is understanding what types of impulses trigger 124 00:07:23,060 --> 00:07:26,440 Josh: when you watch a specific type of content and then funneling that 125 00:07:26,440 --> 00:07:29,720 Josh: into your feed on a daily basis and luke who 126 00:07:29,720 --> 00:07:32,340 Josh: is behind the scenes he was mentioning earlier that in a way 127 00:07:32,340 --> 00:07:35,020 Josh: they become like the this high-end drug dealer that can feed 128 00:07:35,020 --> 00:07:37,920 Josh: you like some dopamine here or some cortisol here and they know exactly 129 00:07:37,920 --> 00:07:41,000 Josh: how a video is going to hit your brain and in fact they can optimize those 130 00:07:41,000 --> 00:07:44,720 Josh: videos through a very tight feedback loop to improve them to a point in which 131 00:07:44,720 --> 00:07:48,980 Josh: it can guarantee that the part of your brain that they want to fire will fire 132 00:07:48,980 --> 00:07:52,560 Josh: and that level of deep understanding not only from the data and preferences 133 00:07:52,560 --> 00:07:55,860 Josh: they've collected over the last decade but also now understanding the biological 134 00:07:55,860 --> 00:07:58,220 Josh: brain and truly at a deep level how it works. 135 00:07:58,400 --> 00:08:02,600 Josh: I can see the good use cases for it, but man, it gets scary quick because they 136 00:08:02,600 --> 00:08:04,940 Josh: have their Vibes app. I don't know if anyone remembers it. 137 00:08:05,470 --> 00:08:09,230 Josh: I'm not sure anyone's even used it, but it's the short form scroller that is 138 00:08:09,230 --> 00:08:13,070 Josh: AI generated content only. And this is like injecting that with steroids. 139 00:08:13,350 --> 00:08:16,370 Josh: And I could see a world in which this content gets very good very quick. 140 00:08:16,450 --> 00:08:17,670 Josh: And is it a good or bad thing? 141 00:08:17,890 --> 00:08:22,330 Josh: That is TBD. You can argue that I would prefer to get good ads that are personalized 142 00:08:22,330 --> 00:08:25,570 Josh: for me that actually sell me things that I'm interested in versus nonsense. 143 00:08:25,830 --> 00:08:30,750 Josh: But at the same time, it's not my decision because they will have more understanding 144 00:08:30,750 --> 00:08:33,490 Josh: of how my brain works than even I will. And that's a little unnerving. 145 00:08:33,950 --> 00:08:37,050 Ejaaz: I'm going to argue on the side of Skynet. It's a bad thing. 146 00:08:37,490 --> 00:08:41,530 Ejaaz: Prove me wrong, Suck, if you're watching this or anyone from Meta. I would love to know. 147 00:08:41,710 --> 00:08:45,830 Ejaaz: Tell me the counter argument. But the good news is, even if they are going after 148 00:08:45,830 --> 00:08:49,730 Ejaaz: your brain and hijacking your dopamine circuits, you might be able to get paid for it. 149 00:08:50,010 --> 00:08:55,130 Ejaaz: This was a story that broke a few weeks ago where this kid, or I think she's 150 00:08:55,130 --> 00:09:01,110 Ejaaz: like a young adult now, basically won a court case against Meta and YouTube 151 00:09:01,110 --> 00:09:03,190 Ejaaz: where she got paid $3 million. 152 00:09:03,730 --> 00:09:07,970 Ejaaz: For the effects of social media, specifically Meta's platforms and YouTube, 153 00:09:08,750 --> 00:09:13,710 Ejaaz: had on her depression and personality, her core years that formed her adolescence. 154 00:09:14,010 --> 00:09:15,330 Ejaaz: She got a three million paycheck. 155 00:09:15,790 --> 00:09:19,130 Ejaaz: I'm wondering if I could do this because I definitely suffered from a lot of this growing up. 156 00:09:19,610 --> 00:09:22,390 Josh: Yeah, what a come up. I mean, it sets a somewhat dangerous precedent, 157 00:09:22,510 --> 00:09:25,110 Josh: right? It's like I spend a lot of time on my computer and my mental health, 158 00:09:25,230 --> 00:09:26,050 Josh: sure, it could be improved. 159 00:09:26,270 --> 00:09:30,110 Josh: Am I eligible for a $3 million raise? And if not me, 160 00:09:30,250 --> 00:09:33,210 Josh: I certainly know some other people nearby that definitely qualify 161 00:09:33,210 --> 00:09:36,170 Josh: for this if that's the parameter that counts so three 162 00:09:36,170 --> 00:09:39,150 Josh: million dollar paycheck i don't know not a bad deal dangerous precedent 163 00:09:39,150 --> 00:09:41,870 Josh: if people are going to be able to start claiming that um as they go because 164 00:09:41,870 --> 00:09:45,490 Josh: i mean again this company has it's done a lot of great things but social media 165 00:09:45,490 --> 00:09:49,750 Josh: as a whole has generated a lot of damage specifically around um the meta-owned 166 00:09:49,750 --> 00:09:56,550 Josh: companies and we'll see what happens meta has been disappointing they continue to be disappointing 167 00:09:57,290 --> 00:10:02,190 Josh: But I am hopeful that a founder-run company led by someone like Zuck can figure 168 00:10:02,190 --> 00:10:07,030 Josh: out a way to really turn this into something special and meaningful and positively impactful. 169 00:10:07,270 --> 00:10:10,550 Josh: So we'll see. I think that is the meta story. 170 00:10:10,830 --> 00:10:15,330 Josh: Now we have another story that has left me a little unnerved, 171 00:10:15,530 --> 00:10:19,770 Josh: which is the annual run rate revenue story between Anthropic and OpenAI and 172 00:10:19,770 --> 00:10:21,610 Josh: how they're actually counting their revenue. 173 00:10:21,730 --> 00:10:25,270 Josh: Because the headline is that what? Anthropic just went from $19 billion to $30 174 00:10:25,270 --> 00:10:28,530 Josh: billion in annual revenue in like a month. 175 00:10:28,710 --> 00:10:33,030 Ejaaz: Yeah. So the breaking news here was that Anthropic had signed a multi-billion 176 00:10:33,030 --> 00:10:36,850 Ejaaz: dollar deal with Google to basically use their GPUs to train Claude. 177 00:10:36,930 --> 00:10:41,370 Ejaaz: But there was a hidden nugget in this article, which revealed that Anthropic's 178 00:10:41,370 --> 00:10:45,050 Ejaaz: revenue run rate, their ARR, has officially hit $30 billion. 179 00:10:45,530 --> 00:10:49,570 Ejaaz: Now, for context here, at the end of last year, it was $9 billion. 180 00:10:50,030 --> 00:10:53,830 Ejaaz: At the start of the year, it had just about hit around $12 billion. 181 00:10:54,350 --> 00:11:00,850 Ejaaz: Last month, it hit $19 billion. In a single month, it has gained $11 billion. 182 00:11:01,270 --> 00:11:05,350 Ejaaz: Now, that's the result of Claude Opus being amazing and the rumored Mythos model, 183 00:11:05,450 --> 00:11:10,030 Ejaaz: which is now confirmed, we did a whole episode about this, being as good as people claimed. 184 00:11:10,210 --> 00:11:15,450 Ejaaz: So people are obviously buying Claude subscriptions and running their revenue rate up to $30 billion. 185 00:11:15,830 --> 00:11:18,690 Ejaaz: I thought this was amazing until we had a conversation this morning, 186 00:11:18,790 --> 00:11:20,690 Ejaaz: Josh, where you said that this is fake news? 187 00:11:20,910 --> 00:11:25,050 Josh: This is nonsense. The accounting is technically GAAP compliant, 188 00:11:25,050 --> 00:11:28,950 Josh: but the way that they get there is so vastly different that you really can't 189 00:11:28,950 --> 00:11:30,070 Josh: compare the two companies together. 190 00:11:30,330 --> 00:11:34,590 Josh: I mean, this example is showing Anthropic is at $30 billion of annual revenue 191 00:11:34,590 --> 00:11:36,510 Josh: and OpenAI is at $24 billion. 192 00:11:37,030 --> 00:11:40,390 Josh: But then explain to me why OpenAI is valued at twice the current valuation of 193 00:11:40,390 --> 00:11:42,750 Josh: Anthropic. It's an accounting problem. 194 00:11:43,070 --> 00:11:46,670 Josh: Now, open ai has a deal with microsoft where open ai shares 20 of its revenue 195 00:11:46,670 --> 00:11:51,050 Josh: with microsoft in the financial statements it counts those sales before that 196 00:11:51,050 --> 00:11:55,670 Josh: deduction but for like azure cloud customers buying open ai models open ai only 197 00:11:55,670 --> 00:11:58,210 Josh: books 20 of that cut as the revenue 198 00:11:58,690 --> 00:12:01,990 Josh: And Anthropic, they have a deal with AWS, Google, and Microsoft, 199 00:12:01,990 --> 00:12:05,390 Josh: and all three of those cloud providers resell cloud to their customers. 200 00:12:05,870 --> 00:12:10,210 Josh: Anthropic books the entire thing as revenue, and then marks off the 80% cut 201 00:12:10,210 --> 00:12:12,830 Josh: as a line item in marketing expenses. 202 00:12:13,150 --> 00:12:17,250 Josh: And that is such a huge difference. I mean, both of those are technically GAAP 203 00:12:17,250 --> 00:12:20,090 Josh: compliant, but they produce very different top line numbers. 204 00:12:20,210 --> 00:12:22,270 Josh: And I think this is really important. And a lot of people are missing this, 205 00:12:22,270 --> 00:12:28,230 Josh: That when they see Anthropic projecting a $30 billion annual recurring revenue, 206 00:12:28,470 --> 00:12:32,570 Josh: that's only projecting out what they've currently done for the last four weeks. 207 00:12:32,710 --> 00:12:37,030 Josh: And it's counting all revenue, including the amount that they're going to have 208 00:12:37,030 --> 00:12:40,450 Josh: to give back to AWS, Google Cloud, and Microsoft Azure. 209 00:12:40,650 --> 00:12:44,050 Josh: And that's like a huge accounting difference. That is a big problem that I don't 210 00:12:44,050 --> 00:12:45,470 Josh: think a lot of people are taking note of. 211 00:12:46,150 --> 00:12:50,270 Ejaaz: I have so many thoughts on this. Number one, how is this compliant and legal? 212 00:12:50,470 --> 00:12:52,030 Ejaaz: Why are they allowed to do that? 213 00:12:52,470 --> 00:12:57,590 Ejaaz: Number two, this is financial accounting crime. They shouldn't be able to do that. 214 00:12:57,810 --> 00:13:01,470 Ejaaz: But something isn't adding up for me, which is one thing that I see a lot of 215 00:13:01,470 --> 00:13:05,030 Ejaaz: these articles comparing is how much revenue they're making, 216 00:13:05,130 --> 00:13:06,350 Ejaaz: but also how much they're burning. 217 00:13:06,630 --> 00:13:11,490 Ejaaz: So for example, with OpenAI, they're making, let's say, $25 billion this year, 218 00:13:11,630 --> 00:13:14,530 Ejaaz: but they're also burning $25 billion dollars. 219 00:13:14,770 --> 00:13:19,330 Ejaaz: And so if Anthropic is moving the line item for their cost center to just say 220 00:13:19,330 --> 00:13:24,010 Ejaaz: marketing, shouldn't their burn rate still be higher than what is being reported on? 221 00:13:24,130 --> 00:13:27,810 Ejaaz: Like something doesn't make sense. Like, is the Financial Times just wrong and 222 00:13:27,810 --> 00:13:32,170 Ejaaz: they're missing this line item completely? Or is Anthropic genuinely just not 223 00:13:32,170 --> 00:13:36,730 Ejaaz: burning as much still and they're still on track to making a profit by 2028? 224 00:13:36,810 --> 00:13:38,590 Ejaaz: Because that's what all the projections have it at, right? 225 00:13:38,730 --> 00:13:42,330 Ejaaz: They're going to be making a profit or turning a profit much sooner than OpenAI 226 00:13:42,330 --> 00:13:45,650 Ejaaz: is who are spending way too much on capex so there's something i'm missing though. 227 00:13:45,650 --> 00:13:50,610 Josh: Yeah i wonder if they're not counting it as burn because it's it's technically 228 00:13:50,610 --> 00:13:56,530 Josh: not like they they are earning money on it just 20 of what they are actually 229 00:13:56,530 --> 00:14:00,810 Josh: putting on the accounting book so like they're not actually burning cash they're 230 00:14:00,810 --> 00:14:03,070 Josh: just accounting for more cash than they're going to right but the 231 00:14:03,070 --> 00:14:07,410 Ejaaz: 80 is being put on the marketing expenses book right well i'm not misunderstanding that. 232 00:14:07,820 --> 00:14:11,180 Josh: Uh, that one we're going to have to talk to an accountant about. I'm not sure. 233 00:14:11,340 --> 00:14:14,360 Josh: I do know that we are comparing apples to oranges though. And we look at open 234 00:14:14,360 --> 00:14:16,740 Josh: AI and Anthropics run rate. And I think this is something that, 235 00:14:16,740 --> 00:14:18,540 Josh: um, an IPO is going to fix. 236 00:14:18,680 --> 00:14:22,780 Josh: Once we have all of these documents publicly stated again, a lot of this is insider reporting. 237 00:14:22,940 --> 00:14:26,580 Josh: You need a subscription just to access this document that we're showing on the screen. 238 00:14:26,700 --> 00:14:32,420 Josh: It is this like really messy, uh, article also in the article or really messy accounting. 239 00:14:32,560 --> 00:14:35,480 Josh: And also in the article, they mentioned how they count their revenue, 240 00:14:35,680 --> 00:14:40,620 Josh: which is just projecting their prior four weeks out for a multiple of 13 to 241 00:14:40,620 --> 00:14:41,800 Josh: account for all 52 weeks. 242 00:14:41,920 --> 00:14:46,640 Josh: So there's a lot of handwaving going on in order to make these numbers go up into the right. 243 00:14:46,760 --> 00:14:49,500 Josh: I just think it's important when you look at these headlines to really take 244 00:14:49,500 --> 00:14:53,280 Josh: it with a grain of salt and understand that Anthropic actually didn't gain $11 245 00:14:53,280 --> 00:14:56,440 Josh: billion in revenue in like a week. That's not happening. 246 00:14:56,600 --> 00:15:01,880 Josh: They're just doing these funny accounting methods to make things look like they are going great. 247 00:15:01,980 --> 00:15:04,780 Josh: And they are going great, just perhaps not as great as people perceive. 248 00:15:04,980 --> 00:15:08,000 Ejaaz: And it might be getting a lot greater for OpenAI. 249 00:15:08,160 --> 00:15:11,520 Ejaaz: I mean, talking of like the IPO rumor mill and like boosting valuations, 250 00:15:11,700 --> 00:15:13,360 Ejaaz: it's all marketing, right? It's all storytelling. 251 00:15:13,660 --> 00:15:18,520 Ejaaz: And there was a story that Axios broke this morning before we started recording 252 00:15:18,520 --> 00:15:23,460 Ejaaz: that OpenAI plans to release their next model. Now they're dubbing it SPUD. 253 00:15:23,640 --> 00:15:27,060 Ejaaz: Some people are calling it GPT 5.5. Some people are calling it GPT 6. 254 00:15:27,120 --> 00:15:31,700 Ejaaz: But apparently it's going to be so good that they have to do a limited release 255 00:15:31,700 --> 00:15:35,560 Ejaaz: that they can't release it publicly because it's too good or too dangerous. 256 00:15:35,760 --> 00:15:40,500 Ejaaz: Now, if that sounds similar, it's because that's exactly what Anthropic did 257 00:15:40,500 --> 00:15:45,040 Ejaaz: this week, announcing Claude Mythos, their next AGI-like model, 258 00:15:45,200 --> 00:15:47,680 Ejaaz: which they're not releasing publicly because it's a cybersecurity risk. 259 00:15:47,800 --> 00:15:49,880 Ejaaz: It discovered 1,000-plus security vulnerabilities. 260 00:15:50,340 --> 00:15:55,080 Ejaaz: And what was interesting about that entire news cycle earlier this week is someone 261 00:15:55,080 --> 00:15:58,700 Ejaaz: commented and said, well, if it's that good and if it's that expensive, 262 00:15:59,260 --> 00:16:03,360 Ejaaz: we're probably not gonna get to use this model for our next another couple of months, basically. 263 00:16:03,920 --> 00:16:07,500 Ejaaz: And OpenAI's head of model training, Thibault, basically said, 264 00:16:07,890 --> 00:16:12,390 Ejaaz: Uh, I wouldn't count on that. He said, um, dot, dot, dot, which implies that 265 00:16:12,390 --> 00:16:15,690 Ejaaz: OpenAI is going to release a very similarly capable model very soon. 266 00:16:15,830 --> 00:16:19,150 Ejaaz: So that's what, uh, that's the news that basically Axios broke. 267 00:16:19,270 --> 00:16:24,730 Ejaaz: But I just want to point out, I want to check myself here that it might also be fake news. 268 00:16:24,990 --> 00:16:28,910 Ejaaz: Um, this post from Dan Shipper apparently spoke to someone within OpenAI and 269 00:16:28,910 --> 00:16:32,450 Ejaaz: he basically said his, his contact basically said that we're talking about a 270 00:16:32,450 --> 00:16:36,050 Ejaaz: different model here that is, uh, hyper-focused on cybersecurity specifically. 271 00:16:36,050 --> 00:16:38,330 Ejaaz: Typically, and it'll be a separate release to GPT-6. 272 00:16:38,530 --> 00:16:41,930 Ejaaz: So at this point, I have no idea what's going on, but I know that OpenAI's valuation 273 00:16:41,930 --> 00:16:45,590 Ejaaz: for their IPO has probably gone up in the space of time that this has happened. 274 00:16:45,930 --> 00:16:52,950 Josh: Yeah, I can't imagine they don't already have a model that is close to Mythos, if not already there. 275 00:16:53,070 --> 00:16:57,170 Josh: And if they don't, then they're, I'm sure, just weeks behind actually having something like that. 276 00:16:57,410 --> 00:17:01,110 Josh: What I found interesting about this story is it says that they're working explicitly 277 00:17:01,110 --> 00:17:05,450 Josh: on a cyber product, which implies that it's for cybersecurity. 278 00:17:06,050 --> 00:17:09,670 Josh: And with Claude Mythos, they weren't actually training it on cyber at all. 279 00:17:09,770 --> 00:17:11,810 Josh: It was just a downstream effect of really powerful code. 280 00:17:12,010 --> 00:17:16,810 Josh: And when I think about the pivot that OpenAI has had recently from retail to 281 00:17:16,810 --> 00:17:19,350 Josh: enterprise, a lot of that focus has been around code. 282 00:17:19,450 --> 00:17:21,890 Josh: So I wonder if they're just building a really strong coding model and this is 283 00:17:21,890 --> 00:17:25,130 Josh: a downstream effect, or if they're genuinely training something explicitly on 284 00:17:25,130 --> 00:17:28,370 Josh: cybersecurity. And I have to imagine, if it's trained explicitly on cybersecurity... 285 00:17:29,240 --> 00:17:31,840 Josh: It probably will become better than mythos and then 286 00:17:31,840 --> 00:17:34,580 Josh: you have to ask the question what happens three months from now 287 00:17:34,580 --> 00:17:37,520 Josh: when these tools actually do become available and also who's going 288 00:17:37,520 --> 00:17:40,580 Josh: to decide when to release them if open ai is taking the clod route or 289 00:17:40,580 --> 00:17:43,800 Josh: the anthropic route and they're keeping it private well now suddenly 290 00:17:43,800 --> 00:17:47,280 Josh: we have everyone's worst nightmare where the labs are kind of all working together 291 00:17:47,280 --> 00:17:53,540 Josh: they all have the most powerful stuff and there is no counterbalance to whatever 292 00:17:53,540 --> 00:17:57,380 Josh: they decide to do with it and they in a way become those king makers and when 293 00:17:57,380 --> 00:18:01,440 Josh: you think about like the Department of War contract that we had this whole episode about is a big mess. 294 00:18:01,680 --> 00:18:05,160 Josh: Anthropic kind of has all the leverage. And if OpenAI moves with them, 295 00:18:05,320 --> 00:18:09,880 Josh: they have the ability to crack nation state software at will. 296 00:18:10,020 --> 00:18:13,120 Josh: And that seems really powerful. And they are now the gatekeepers. 297 00:18:13,300 --> 00:18:16,100 Josh: And I don't know, Chris, a lot of interesting questions as we move into this 298 00:18:16,100 --> 00:18:20,220 Josh: next paradigm of Blackwell models that are unbelievably powerful. 299 00:18:20,440 --> 00:18:24,640 Ejaaz: But as we mentioned earlier, Project Glasswing, Anthropic mentioned their breaking 300 00:18:24,640 --> 00:18:28,140 Ejaaz: tier model, Claude Mythos, which is coming out. We did an entire episode about 301 00:18:28,140 --> 00:18:30,560 Ejaaz: this earlier this week. Definitely go and check this out. 302 00:18:31,080 --> 00:18:34,940 Ejaaz: Now, I want to end this segment on a bit of chart crime, Josh, 303 00:18:35,020 --> 00:18:36,980 Ejaaz: because now that you've told me about the fake revenue numbers, 304 00:18:37,160 --> 00:18:40,580 Ejaaz: I can't look at this Wall Street Journal analysis and think, 305 00:18:40,980 --> 00:18:45,840 Ejaaz: is Anthropik's numbers fake? On the left here, we see yearly AI model training costs. 306 00:18:45,980 --> 00:18:48,400 Ejaaz: And it basically shows that OpenAI is spending way too much money. 307 00:18:48,540 --> 00:18:50,340 Ejaaz: And Anthropik is spending a fraction of that. 308 00:18:50,440 --> 00:18:53,780 Ejaaz: But now I'm realizing that that's probably in projection with their revenue. 309 00:18:53,780 --> 00:18:56,820 Ejaaz: And if you look at the bottom left over here, it shows that they're kind of the same. 310 00:18:57,000 --> 00:19:00,880 Ejaaz: So I think there is some chart crime or county crime happening with Anthropic, 311 00:19:01,040 --> 00:19:03,440 Ejaaz: and I want that to be talked about more. 312 00:19:03,660 --> 00:19:06,560 Ejaaz: But it's not all rosy with OpenAI. 313 00:19:06,860 --> 00:19:09,560 Ejaaz: They did have the UK Stargate project go down, right? 314 00:19:09,920 --> 00:19:12,700 Josh: The whole Project Stargate thing is really just a huge disappointment. 315 00:19:13,020 --> 00:19:16,400 Josh: It was supposed to be this giant grand buildout domestically, internationally. 316 00:19:16,860 --> 00:19:20,980 Josh: I think they also did this in Saudi Arabia, somewhere in the Middle East. 317 00:19:20,980 --> 00:19:23,540 Josh: They were planning to build one. They didn't do it in the UK. 318 00:19:23,780 --> 00:19:27,760 Josh: They're not doing it in the U S they're not doing it. Project Stargate was really 319 00:19:27,760 --> 00:19:30,580 Josh: just as big, like raw, raw project that was initiated with the government. 320 00:19:31,540 --> 00:19:34,820 Josh: Elon famously said the post the day that it was announced, you don't have the 321 00:19:34,820 --> 00:19:38,800 Josh: money to do this. Turns out he was right. No one actually wants to foot the bill for this. 322 00:19:39,140 --> 00:19:42,320 Josh: Logistically, it's very technical and challenging. And it turns out it's just 323 00:19:42,320 --> 00:19:43,820 Josh: really hard to build things at scale, 324 00:19:44,560 --> 00:19:47,680 Josh: particularly internationally in places like Europe and the UK, 325 00:19:47,680 --> 00:19:52,480 Josh: where there's a lot of regulatory issues, both environmental and energy related, 326 00:19:52,680 --> 00:19:54,700 Josh: that just make doing these things incredibly difficult. 327 00:19:54,860 --> 00:19:57,720 Josh: And we're seeing it here domestically. We have a story a little bit later about 328 00:19:57,720 --> 00:20:00,340 Josh: AI data centers and how they're just having a really tough time getting them 329 00:20:00,340 --> 00:20:03,760 Josh: online. And this is probably one of the future trends that we're going to look 330 00:20:03,760 --> 00:20:07,700 Josh: out for is just the idea that building these data centers is hard. 331 00:20:08,250 --> 00:20:11,890 Josh: And not always just because of the technicalities, but also because of the regulation 332 00:20:11,890 --> 00:20:15,510 Josh: and legislation associated with this and the public sentiment. It's not looking good. 333 00:20:16,150 --> 00:20:20,870 Josh: There is certainly a rift in the world right now between people who want to 334 00:20:20,870 --> 00:20:22,110 Josh: build these things and people who do not. 335 00:20:22,510 --> 00:20:27,570 Ejaaz: Shifting gears slightly, I wish Anthropix Claude Mythos was the main headline, 336 00:20:27,750 --> 00:20:29,970 Ejaaz: but they're still shipping other products somehow. 337 00:20:30,190 --> 00:20:33,330 Ejaaz: Maybe they're using an AGI-like model to do so, right? They announced a new 338 00:20:33,330 --> 00:20:37,770 Ejaaz: product called Claude Manage Agents, which is basically, think of AWS, 339 00:20:38,310 --> 00:20:39,730 Ejaaz: but for spinning up AI agents. 340 00:20:39,850 --> 00:20:44,190 Ejaaz: It's a platform that allows you to design and architect an agent through a single 341 00:20:44,190 --> 00:20:45,110 Ejaaz: prompt or a couple of prompts. 342 00:20:45,210 --> 00:20:48,630 Ejaaz: You can amend the memory, a bit of its own custom design, and then launch it, 343 00:20:48,670 --> 00:20:50,270 Ejaaz: which may not sound novel, right? 344 00:20:50,310 --> 00:20:53,890 Ejaaz: You could create AI agents before, but there's a distinct difference here. 345 00:20:54,070 --> 00:20:59,150 Ejaaz: Typically, when you create an AI agent, you can't scale it to your millions of users. 346 00:20:59,250 --> 00:21:02,330 Ejaaz: Let's say if you're a Fortune 500 company, because you need to set up a bunch 347 00:21:02,330 --> 00:21:06,370 Ejaaz: of other production and dev tooling environments in order to support that. 348 00:21:06,470 --> 00:21:09,370 Ejaaz: That typically takes anywhere between three to six months. 349 00:21:09,570 --> 00:21:13,750 Ejaaz: Now, I ran a bit of the math on this. Typically, an app development at scale 350 00:21:13,750 --> 00:21:18,230 Ejaaz: for, say, a million users costs around $50,000, depending on the specific type 351 00:21:18,230 --> 00:21:19,550 Ejaaz: of feature or product you want to launch. 352 00:21:19,730 --> 00:21:26,150 Ejaaz: This reduces it down to $100. That's like a 500x reduction, and you can do it in under an hour. 353 00:21:26,330 --> 00:21:30,130 Ejaaz: But there is a new cool thing that this product unlocks. Josh, can you guess it? 354 00:21:30,790 --> 00:21:33,550 Josh: I'm going to guess the end of white call it work or am I being dramatic? 355 00:21:34,450 --> 00:21:39,630 Ejaaz: Well, that's also that, but they figured out a new revenue run rate for them. 356 00:21:39,910 --> 00:21:41,830 Josh: Of course they did. Of course they did. 357 00:21:42,010 --> 00:21:44,390 Ejaaz: Right now. Now listen to this genius plan, right? 358 00:21:44,870 --> 00:21:51,490 Ejaaz: Typically, every single AI model provider charges you based on the amount of tokens you use. 359 00:21:51,710 --> 00:21:55,650 Ejaaz: Tokens in, tokens out, you pay a subscription, or you pay an API cost. 360 00:21:56,130 --> 00:22:00,690 Ejaaz: Anthropic, for this product, is charging you for the amount of time that your 361 00:22:00,690 --> 00:22:03,730 Ejaaz: agent takes to think of a solution. 362 00:22:03,930 --> 00:22:09,550 Ejaaz: So the tokens it's using to think is now being charged to the tune of $0.08 363 00:22:09,550 --> 00:22:13,790 Ejaaz: per call. Now, if you assume that, and I actually don't need to assume, 364 00:22:13,890 --> 00:22:17,530 Ejaaz: they used a live example of Sentry processing 1 million bug reports. 365 00:22:17,710 --> 00:22:22,670 Ejaaz: If each agent session is 10 minutes, that's 166,000 session hours, 366 00:22:22,830 --> 00:22:27,570 Ejaaz: which turns out to be around $13,000 per run in fees. 367 00:22:27,710 --> 00:22:31,130 Ejaaz: That all adds up massively. If you're a massive corporate enterprise where you 368 00:22:31,130 --> 00:22:35,710 Ejaaz: distribute this platform to whatever, 50 plus teams, you end up making millions of dollars. 369 00:22:36,330 --> 00:22:40,490 Ejaaz: Genius unlock anthropic well done i'm clapping for you also you're going after 370 00:22:40,490 --> 00:22:43,950 Ejaaz: all the other startups out there you probably killed out a bunch of different 371 00:22:43,950 --> 00:22:47,610 Ejaaz: agent harness startups that will value their billions of dollars well done. 372 00:22:47,610 --> 00:22:50,290 Josh: If you use a computer you just have to assume that 373 00:22:50,290 --> 00:22:53,210 Josh: you're months away from no longer needing to use a computer for anything you 374 00:22:53,210 --> 00:22:56,150 Josh: don't want to do yeah um it will just be automated you 375 00:22:56,150 --> 00:22:59,350 Josh: can have the computer watch your screen emulate your decision making processes 376 00:22:59,350 --> 00:23:02,150 Josh: do all the clicking and thinking that you would need to do in order 377 00:23:02,150 --> 00:23:04,950 Josh: to accomplish whatever you're trying to do and that's been 378 00:23:04,950 --> 00:23:07,950 Josh: it's been a recent realization particularly with mythos about how close we are 379 00:23:07,950 --> 00:23:11,810 Josh: to this reality and i have to imagine that all these features are being built 380 00:23:11,810 --> 00:23:15,430 Josh: with that model and are therefore resulting in this incredibly fast iteration 381 00:23:15,430 --> 00:23:18,870 Josh: loop that anthropic's having where every day we get some like groundbreaking 382 00:23:18,870 --> 00:23:23,570 Josh: technological impact if you just push this out three months six months even 12 months 383 00:23:24,190 --> 00:23:27,650 Josh: there's no way that you're going to need to use your computer for anything you 384 00:23:27,650 --> 00:23:30,230 Josh: don't want to it It will just automate the entire process for you. 385 00:23:30,670 --> 00:23:35,370 Josh: So Anthropic with another big win. Quad is just on fire. The rate of acceleration 386 00:23:35,370 --> 00:23:37,190 Josh: is truly through the roof. 387 00:23:37,610 --> 00:23:39,310 Ejaaz: Josh, are you a Black Mirror fan? 388 00:23:40,040 --> 00:23:42,040 Josh: Oh, huge fan. Watched every episode. 389 00:23:42,260 --> 00:23:47,640 Ejaaz: Huge fan. So what if I told you there is now a robotic lamp that you can buy 390 00:23:47,640 --> 00:23:52,640 Ejaaz: that turns into pincers that can fold your clothes, make your bed, 391 00:23:52,720 --> 00:23:54,160 Ejaaz: and maybe make you a cup of coffee? 392 00:23:54,340 --> 00:23:56,920 Josh: You buy that? I say 100% chance it stares me through the heart when I'm sleeping. 393 00:23:57,100 --> 00:24:00,100 Josh: But I'm going to hope that's not the case because we actually did have this 394 00:24:00,100 --> 00:24:04,400 Josh: founder on the show a while back to talk about this product, which is now out? 395 00:24:04,900 --> 00:24:07,180 Josh: Yes. Kind of. On a second, Cali, explain what's going on here. 396 00:24:07,420 --> 00:24:11,720 Ejaaz: Okay. So what you're seeing on your screen is a lamp which basically extends 397 00:24:11,720 --> 00:24:15,500 Ejaaz: into a robotic arm and it can do a bunch of chores for you. 398 00:24:15,620 --> 00:24:20,020 Ejaaz: So what you're seeing on the screen is this lady is apparently putting her clean 399 00:24:20,020 --> 00:24:23,960 Ejaaz: laundry on the front of her bed and her lamps are now activating. 400 00:24:24,220 --> 00:24:28,980 Ejaaz: And now you can see their claws coming out and they can start folding your clothes 401 00:24:28,980 --> 00:24:32,020 Ejaaz: and making your bed or starting your record player. 402 00:24:32,200 --> 00:24:37,840 Ejaaz: And the point is robots are going to be pretty pervasive in human society. 403 00:24:38,260 --> 00:24:41,980 Ejaaz: They may not necessarily look like humanoids. And that's the point that Sincere 404 00:24:41,980 --> 00:24:44,200 Ejaaz: and Aaron Tan, the founder, is aiming at. 405 00:24:44,320 --> 00:24:47,740 Ejaaz: Now, the last time that we spoke about this founder, he only had a mock-up. 406 00:24:48,020 --> 00:24:51,580 Ejaaz: And to be honest, the lamp looked pretty scary. The pincers were much larger. 407 00:24:51,800 --> 00:24:55,580 Ejaaz: It's nice to see that they've now got like a kind of curved metal piece over 408 00:24:55,580 --> 00:24:59,720 Ejaaz: the pincers. So maybe they can't necessarily stab you. It's more colorful, it's more amenable. 409 00:25:00,320 --> 00:25:04,160 Ejaaz: It looks kind of slow, but the good news is you can start ordering this right now. 410 00:25:04,240 --> 00:25:06,500 Ejaaz: I signed up on the wait list and I got an email the other day saying, 411 00:25:06,580 --> 00:25:08,920 Ejaaz: hey, you can now order this thing. I don't know when they're gonna start delivering 412 00:25:08,920 --> 00:25:11,760 Ejaaz: it, but it might be something that I try out. Are you trying it out? 413 00:25:12,840 --> 00:25:15,080 Josh: No, I will not be trying this out. Okay. 414 00:25:16,270 --> 00:25:22,070 Josh: I love founders who are trying cool things. And I think building a narrow use 415 00:25:22,070 --> 00:25:23,550 Josh: robot like this is so cool. 416 00:25:23,710 --> 00:25:27,910 Josh: The design is awesome. It's a lamp that does robot things. And that's very cool. 417 00:25:28,050 --> 00:25:31,910 Josh: When I look at the trailer here, and this isn't really showing many use cases, 418 00:25:31,910 --> 00:25:34,010 Josh: one of which is the folding of laundry. 419 00:25:34,190 --> 00:25:37,250 Josh: You kind of see the actuators there. They're a little like fidgety. 420 00:25:37,410 --> 00:25:38,330 Josh: They're not quite smooth. 421 00:25:38,570 --> 00:25:42,330 Josh: They're like, it doesn't seem that it's moving very quick. I can't imagine this many use cases. 422 00:25:42,470 --> 00:25:46,070 Josh: Like we're looking at a robotic lamp that is being manually adjusted and turned. 423 00:25:46,070 --> 00:25:49,270 Josh: Like why is that happening there's a lot of questions i have about what the 424 00:25:49,270 --> 00:25:52,170 Josh: actual use cases of something like this are and how effective it is at those 425 00:25:52,170 --> 00:25:56,650 Josh: use cases but again i love the idea that people are trying new things and trying 426 00:25:56,650 --> 00:26:00,070 Josh: to build something unique with beautiful design that's actually effective inside 427 00:26:00,070 --> 00:26:03,430 Josh: of a home and as soon as you get yours delivered i'm coming over and i'm trying 428 00:26:03,430 --> 00:26:05,010 Josh: it out because i want to see i 429 00:26:05,010 --> 00:26:08,870 Ejaaz: Might be dead so i don't know maybe you're recovering my corpse at that point 430 00:26:08,870 --> 00:26:14,790 Ejaaz: josh but in other news space x ai is back and they've signed a massive partnership 431 00:26:14,790 --> 00:26:19,950 Ejaaz: with Intel, American made fabricator of AI chips. 432 00:26:20,170 --> 00:26:23,610 Ejaaz: And the question becomes, why on earth are they doing this? Well, 433 00:26:23,790 --> 00:26:24,890 Ejaaz: there's a few different reasons. 434 00:26:25,070 --> 00:26:28,490 Ejaaz: Number one, Elon Musk announced something called the TerraFab, 435 00:26:28,610 --> 00:26:33,710 Ejaaz: which is pretty much the most ambitious AI chip manufacturing project that is 436 00:26:33,710 --> 00:26:38,650 Ejaaz: ever going to be achieved if it does get achieved over the next, say, five to 10 years. 437 00:26:38,930 --> 00:26:43,570 Ejaaz: The idea is to achieve one terawatts worth of compute, 80% of those AI chips 438 00:26:43,570 --> 00:26:46,790 Ejaaz: are actually getting sent out to space to harvest the sun's energy to train 439 00:26:46,790 --> 00:26:48,590 Ejaaz: AI models, presumably Grok. 440 00:26:48,690 --> 00:26:51,730 Ejaaz: Now, if all of that sounds insane, we have a bunch of episodes that you can 441 00:26:51,730 --> 00:26:53,690 Ejaaz: go and check out and we explain everything. 442 00:26:53,850 --> 00:26:56,050 Ejaaz: But the point is, why Intel specifically? 443 00:26:56,250 --> 00:27:00,850 Ejaaz: I think there's two main reasons. Number one, you need these AI chips to be 444 00:27:00,850 --> 00:27:03,250 Ejaaz: American-made and American-manufactured. 445 00:27:03,450 --> 00:27:09,130 Ejaaz: AI has become a huge geopolitical weapon and taiwan the threat of taiwan being 446 00:27:09,130 --> 00:27:11,150 Ejaaz: taken over by china and tsmc being. 447 00:27:11,730 --> 00:27:17,110 Ejaaz: Within Taiwan is a massive threat to the US AI production of AI models, 448 00:27:17,250 --> 00:27:19,930 Ejaaz: GPUs, etc. NVIDIA realized heavily on TSMC. 449 00:27:20,370 --> 00:27:24,330 Ejaaz: Intel is the closest American-made lab or manufacturing plant that we can get 450 00:27:24,330 --> 00:27:26,910 Ejaaz: to building A-grade AI chips. 451 00:27:27,150 --> 00:27:29,550 Ejaaz: But why is Elon signing up with Intel specifically? 452 00:27:29,810 --> 00:27:33,530 Ejaaz: There was this little nugget that I saw Robert Scoble post about, 453 00:27:33,650 --> 00:27:37,990 Ejaaz: which is there's this compound called gallium nitride, which basically makes 454 00:27:37,990 --> 00:27:41,730 Ejaaz: these AI chips radiation-hardened, which, there you go, 455 00:27:41,890 --> 00:27:45,150 Ejaaz: is going to make it perfectly suitable to launch these AI chips into space. 456 00:27:45,290 --> 00:27:47,430 Ejaaz: So Elon's already thinking way in advance. 457 00:27:47,610 --> 00:27:50,950 Ejaaz: Robert got into a bit of trouble because he published that Elon had liked this 458 00:27:50,950 --> 00:27:54,590 Ejaaz: tweet, aka confirming that this was partially the reason why they did that. 459 00:27:54,910 --> 00:27:58,470 Ejaaz: But yeah, this might be the next unlock for achieving the TerraFap. It's pretty cool. 460 00:27:59,010 --> 00:28:02,490 Josh: It's one of the most ambitious projects I think any company is undertaking on earth. 461 00:28:02,750 --> 00:28:07,730 Josh: There's no one who's really trying to offset this monopoly that exists on chip 462 00:28:07,730 --> 00:28:08,910 Josh: fabrication and production. 463 00:28:09,110 --> 00:28:11,850 Josh: And I think one of the most important things that's underrated about the TerraFab 464 00:28:11,850 --> 00:28:15,110 Josh: is the fact that they have a separate staging facility separate from the TerraFab 465 00:28:15,110 --> 00:28:18,170 Josh: that has all of the required pieces needed to make these chips. 466 00:28:18,290 --> 00:28:20,470 Josh: It has the lithography, it has the masking, it has the packaging. 467 00:28:20,750 --> 00:28:25,010 Josh: And what that allows you to do is iterate very quickly on the actual design 468 00:28:25,010 --> 00:28:29,010 Josh: of these chips. A lot of times a chip gets submitted and then a year goes by 469 00:28:29,010 --> 00:28:31,650 Josh: or even longer until you actually have the full thing completed. 470 00:28:31,930 --> 00:28:34,910 Josh: This compresses that iteration cycle because it's all into one roof and allows 471 00:28:34,910 --> 00:28:39,510 Josh: them to make chips that are far better very quick because they can do all the testing in one place. 472 00:28:39,590 --> 00:28:43,150 Josh: And they can even do that prior to the TerraFab going fully online because it's 473 00:28:43,150 --> 00:28:44,310 Josh: a small sample set of that. 474 00:28:44,510 --> 00:28:47,610 Josh: The TerraFab is going to be hard. I'm sure there's going to be a lot of negative 475 00:28:47,610 --> 00:28:49,830 Josh: press as they make mistakes and as things get delayed. 476 00:28:50,070 --> 00:28:53,330 Josh: But the outcome of the terafab is so 477 00:28:53,330 --> 00:28:56,210 Josh: profound that it's hard to imagine a world in which tesla 478 00:28:56,210 --> 00:28:59,350 Josh: does accomplish this at scale or spacex ai does accomplish 479 00:28:59,350 --> 00:29:03,070 Josh: this at scale and they are collectively not the most valuable company in the 480 00:29:03,070 --> 00:29:05,830 Josh: world because that implies not only do they have the chips but they have the 481 00:29:05,830 --> 00:29:10,230 Josh: robots they have the satellites they have the spaceships they have all the infrastructure 482 00:29:10,230 --> 00:29:15,550 Josh: required for this next generation of embodied ai of space trained intelligence 483 00:29:15,550 --> 00:29:18,970 Josh: and super super intelligence and I'm really bullish on it. 484 00:29:19,430 --> 00:29:22,430 Josh: Intel's badass. I'm glad they're helping them, and I'm just so stoked for the 485 00:29:22,430 --> 00:29:24,890 Josh: tariff app in general. This is going to be a fun one to follow over the next few years. 486 00:29:25,250 --> 00:29:29,710 Ejaaz: Yeah. I have a huge Intel bag, so please, please, please keep signing all these partnerships. Please. 487 00:29:30,470 --> 00:29:35,270 Ejaaz: Now, the reason why Josh and I started this show, the reason why we do Limitless 488 00:29:35,270 --> 00:29:36,990 Ejaaz: is we're very optimistic about the tech. 489 00:29:37,110 --> 00:29:41,410 Ejaaz: Now, we know that there's a lot of doomerous takes, but the fact is we believe 490 00:29:41,410 --> 00:29:44,070 Ejaaz: AI is going to change the world for good, and there are many different ways 491 00:29:44,070 --> 00:29:47,010 Ejaaz: that we think that's going to happen, and we're going to track every single news story. 492 00:29:47,760 --> 00:29:52,200 Ejaaz: Supports that. But there are obviously people in the world that don't believe that's the case. 493 00:29:52,460 --> 00:29:56,060 Ejaaz: And unfortunately, this week, we had a pretty serious story where someone fired 494 00:29:56,060 --> 00:30:01,300 Ejaaz: 13 shots into the home of an Indianapolis counselor with a note reading, 495 00:30:01,780 --> 00:30:04,060 Ejaaz: no data centers left at the scene. 496 00:30:04,640 --> 00:30:08,520 Ejaaz: Now, we don't know the exact motivations of this person, because I don't believe 497 00:30:08,520 --> 00:30:09,580 Ejaaz: they've been caught just yet. 498 00:30:09,640 --> 00:30:13,240 Ejaaz: But you can hypothesize what the takes are, which is stuff that we've covered 499 00:30:13,240 --> 00:30:16,480 Ejaaz: on the show before, which is people are worried that data centers are going 500 00:30:16,480 --> 00:30:19,780 Ejaaz: to empower AI models that are going to eventually replace them or take their job. 501 00:30:19,940 --> 00:30:22,900 Ejaaz: They're worried about the energy costs. They're worried about the water consumption. 502 00:30:23,320 --> 00:30:28,220 Ejaaz: Now, the issue with this is, number one, AI data centers take up less water 503 00:30:28,220 --> 00:30:30,420 Ejaaz: than your average golf course. 504 00:30:30,680 --> 00:30:34,160 Ejaaz: That's like in your neighborhood itself. We did a whole episode covering this. 505 00:30:34,480 --> 00:30:38,760 Ejaaz: Number two, for the electricity charges that kind of increase for people's bills, 506 00:30:38,940 --> 00:30:42,580 Ejaaz: a lot of governments and states have mandated that the AI labs responsible for 507 00:30:42,580 --> 00:30:46,080 Ejaaz: this pay for that extra surplus so that it doesn't actually hit you. 508 00:30:46,360 --> 00:30:49,400 Ejaaz: Also, we're working on different ways to deal with this electricity consumption, 509 00:30:49,600 --> 00:30:54,260 Ejaaz: like launching GPUs into space. So it's just a sad and very concerning story to see. 510 00:30:54,460 --> 00:30:57,940 Ejaaz: There's a growing contingency of people that are against data centers, 511 00:30:57,980 --> 00:31:01,060 Ejaaz: and I understand the concerns, but this shouldn't be the way to deal with it. 512 00:31:01,360 --> 00:31:04,660 Josh: This is dark. You know what's way worse than not having data centers? 513 00:31:04,980 --> 00:31:08,800 Josh: Trying to getting mythos first and then using it to attack all of our infrastructure 514 00:31:08,800 --> 00:31:12,640 Josh: and then using it to iterate and build even more powerful models that are even 515 00:31:12,640 --> 00:31:15,600 Josh: more dangerous, more harmful, and then applying that back to us. 516 00:31:15,720 --> 00:31:20,820 Josh: And I think the impact of that is far greater than the impact of some patch 517 00:31:20,820 --> 00:31:24,220 Josh: of grass that is so detached from most towns getting turned into a data center. 518 00:31:24,420 --> 00:31:30,200 Josh: And I think the moral dilemma here is that people are saying they want one thing 519 00:31:30,200 --> 00:31:31,660 Josh: and then doing something else. 520 00:31:31,900 --> 00:31:39,760 Josh: And it's disturbing to see the sheer size of the population that doesn't want 521 00:31:39,760 --> 00:31:43,420 Josh: to move forward as it relates to AI and progress, 522 00:31:43,700 --> 00:31:49,480 Josh: totally unaware of the fact that this moves on whether we're a participant or not. And... 523 00:31:49,890 --> 00:31:54,650 Josh: As these things become more powerful, there's a lot more profound downstream 524 00:31:54,650 --> 00:32:02,470 Josh: effects of not having these models in our core, having them like having the power on our side. 525 00:32:02,550 --> 00:32:07,510 Josh: And I hope that this becomes more of a realization for a lot more people. 526 00:32:07,690 --> 00:32:10,050 Josh: There's this clear divide happening now where it's like people who are using 527 00:32:10,050 --> 00:32:13,510 Josh: these AI tools to further empower themselves and to do better work in their 528 00:32:13,510 --> 00:32:16,450 Josh: lives or to handle more things in their personal lives. and then those who don't. 529 00:32:16,610 --> 00:32:20,090 Josh: And that K-curve that's going to come out of this in the economy, 530 00:32:20,230 --> 00:32:25,850 Josh: in society, just throughout our general day-to-day lives is going to become pretty wide. 531 00:32:26,150 --> 00:32:29,570 Josh: And I hope a lot of people really reflect on 532 00:32:30,610 --> 00:32:36,750 Josh: What it looks like to like actively be on the wrong side, be on the slowdown 533 00:32:36,750 --> 00:32:38,790 Josh: side of history, be on the, what is it? 534 00:32:38,890 --> 00:32:41,930 Josh: The deceleration side of history and like what the actual implications of that 535 00:32:41,930 --> 00:32:43,590 Josh: are as we continue progress forward. 536 00:32:43,890 --> 00:32:49,850 Josh: I don't know. It makes me sad, but hopefully it's something that will change over time. 537 00:32:50,270 --> 00:32:53,570 Josh: And perhaps it's just a messaging thing. It's funny. A lot of people that hate 538 00:32:53,570 --> 00:32:55,150 Josh: SpaceX were very excited about Artemis. 539 00:32:55,370 --> 00:33:00,510 Josh: So perhaps like the goals are the same. We just need to package them differently. 540 00:33:00,610 --> 00:33:03,890 Josh: In a way that's more digestible that these mobs can get behind. 541 00:33:04,190 --> 00:33:06,790 Josh: And I don't know, maybe it's a messaging thing. Maybe it's a moral thing. 542 00:33:07,030 --> 00:33:12,150 Josh: But that's where we'll leave you at the end of this week on the conclusion of the AI Roundup. 543 00:33:12,270 --> 00:33:15,090 Josh: That's four episodes in a single week about all of the hottest topics. 544 00:33:15,190 --> 00:33:16,650 Josh: If you missed anything, you can go back and watch them. 545 00:33:17,590 --> 00:33:20,670 Josh: I think it's been a big week. I mean, we had a few huge models released. 546 00:33:20,810 --> 00:33:23,710 Josh: We had Mythos. We had OpenAI versus Anthropic. 547 00:33:24,070 --> 00:33:28,150 Josh: We had Gemma 4, which was very cool and a very powerful model. 548 00:33:28,310 --> 00:33:31,150 Josh: But that wraps up everything for this week. Gijos, any final thoughts before 549 00:33:31,150 --> 00:33:33,230 Josh: we let these lovely listeners go? 550 00:33:33,790 --> 00:33:38,410 Ejaaz: Nope. Thank you guys for watching and listening. I'm curious if you guys have 551 00:33:38,410 --> 00:33:41,170 Ejaaz: any thoughts on any of the topics that we've discussed today, 552 00:33:41,210 --> 00:33:45,130 Ejaaz: or if there are any topics that you think we have missed or that you want to hear more of. 553 00:33:45,290 --> 00:33:48,970 Ejaaz: We are trying to cover any and every breaking topic, as well as some novel analysis 554 00:33:48,970 --> 00:33:51,250 Ejaaz: into the actual tools. It's one thing announcing the tools. 555 00:33:51,470 --> 00:33:54,070 Ejaaz: It's another thing using them. We're going to be doing more demos in the future. 556 00:33:54,470 --> 00:33:58,910 Ejaaz: But yeah, that's the end of the agenda for this week. And we will see you next 557 00:33:58,910 --> 00:34:00,570 Ejaaz: week. Thank you guys so much for listening. 558 00:34:00,990 --> 00:34:01,810 Josh: See you guys in the next one.