1 00:00:00,020 --> 00:00:04,820 Ejaaz: Three weeks ago, rumors broke that a major AI lab had built a model more powerful, 2 00:00:05,060 --> 00:00:09,520 Ejaaz: more dangerous, and more expensive than any AI model that we had seen before. 3 00:00:09,780 --> 00:00:13,480 Ejaaz: We didn't know which model lab it would be. We didn't know what the model was called. 4 00:00:13,720 --> 00:00:18,840 Ejaaz: And then just a few days ago, Anthropic leaked a model called Claude Mythos, 5 00:00:18,960 --> 00:00:22,400 Ejaaz: which is supposedly more powerful than any model that they've ever built before, 6 00:00:22,540 --> 00:00:26,280 Ejaaz: a tier above Opus 4.6, which is what we see today. 7 00:00:26,540 --> 00:00:30,840 Ejaaz: This model is actually so good that it is considered a cyber security threat 8 00:00:30,840 --> 00:00:33,100 Ejaaz: and can't be rolled out to the public just yet. 9 00:00:33,240 --> 00:00:36,900 Ejaaz: But it's not just Anthropic that's building a model that is close to AGI like this. 10 00:00:37,240 --> 00:00:41,740 Ejaaz: OpenAI has a model codenamed Spud, Google has a model codenamed Agent Smith, 11 00:00:41,900 --> 00:00:43,800 Ejaaz: and there's many more to come this year. 12 00:00:43,940 --> 00:00:48,100 Josh: But the Anthropic leak wasn't intentional. This was discovered by accident last 13 00:00:48,100 --> 00:00:52,620 Josh: Thursday, March 26th, by a Fortune reporter who discovered that Anthropic's 14 00:00:52,620 --> 00:00:54,920 Josh: content management system had a configuration error. 15 00:00:54,920 --> 00:00:57,240 Josh: And for those who aren't familiar, the content management system, 16 00:00:57,360 --> 00:00:59,660 Josh: it's how the web server serves files. 17 00:00:59,860 --> 00:01:04,320 Josh: And within that, there is a config error that leaked nearly 3,000 unpublished 18 00:01:04,320 --> 00:01:07,600 Josh: assets sitting in this publicly searchable database. 19 00:01:07,920 --> 00:01:10,640 Josh: Anyone could find them. So two independent security researchers, 20 00:01:10,640 --> 00:01:12,600 Josh: they went through, they confirmed. 21 00:01:12,860 --> 00:01:17,940 Josh: And among these files were two blog posts of two models named Claude Mythos 22 00:01:17,940 --> 00:01:20,880 Josh: and a new tier named Capybara. 23 00:01:21,290 --> 00:01:24,050 Josh: Anthropic immediately removed access to all of 24 00:01:24,050 --> 00:01:27,210 Josh: this as soon as it came out but then later on an anthropic spokesperson 25 00:01:27,210 --> 00:01:30,030 Josh: confirmed that it represents a step change in 26 00:01:30,030 --> 00:01:33,750 Josh: ai performance and is the most capable model we've ever built so they confirmed 27 00:01:33,750 --> 00:01:38,450 Josh: what we're seeing here is real now the problem is like this image suggests on 28 00:01:38,450 --> 00:01:42,530 Josh: screen we're missing a lot of information this is a leak that something like 29 00:01:42,530 --> 00:01:46,310 Josh: this exists but we don't we're not sure exactly what what we do know is that 30 00:01:46,310 --> 00:01:49,070 Josh: there is the new model tier, Ejaz, like you mentioned, named Capybara. 31 00:01:49,250 --> 00:01:53,310 Josh: It is the new tier that sits above Opus. So now the lineup will kind of look 32 00:01:53,310 --> 00:01:56,990 Josh: like Haiku, Sonnet, Opus, and then Capybara at the top. 33 00:01:57,170 --> 00:01:59,810 Josh: It doesn't really sound quite right. Maybe that's an experimental name. 34 00:01:59,910 --> 00:02:00,830 Josh: They might find something better. 35 00:02:01,070 --> 00:02:04,270 Josh: And then Mythos is the specific model name within that tier. 36 00:02:04,410 --> 00:02:07,630 Josh: So you can think of Capybara as the weight class and Mythos as like the fighter. 37 00:02:07,830 --> 00:02:08,670 Josh: It's the specific model. 38 00:02:08,950 --> 00:02:12,710 Josh: Now, according to the leaked documents, this dramatically outperforms Claude 39 00:02:12,710 --> 00:02:17,070 Josh: Opus 4.6 on basically everything, but particularly coding, academic reasoning, 40 00:02:17,270 --> 00:02:19,050 Josh: and the cybersecurity benchmarks. 41 00:02:19,250 --> 00:02:22,630 Josh: And I think the cybersecurity one is one of the more interesting points here, 42 00:02:22,750 --> 00:02:26,170 Josh: because it's so powerful as cybersecurity that one of the main reasons why they 43 00:02:26,170 --> 00:02:29,650 Josh: can't release it is to actually prevent people from using it maliciously. Is that right? 44 00:02:29,990 --> 00:02:33,430 Ejaaz: Yeah. So actually, if we rewind to about a month and a half ago, 45 00:02:34,190 --> 00:02:38,330 Ejaaz: Anthropik's head of AI security, who's actually a legend in the industry, 46 00:02:38,790 --> 00:02:42,850 Ejaaz: gave a talk about Claude Opus 4.6, when it had just released. 47 00:02:43,090 --> 00:02:49,850 Ejaaz: And his talk described how the model was pointed at five to 10 very popular 48 00:02:49,850 --> 00:02:53,110 Ejaaz: open source code bases with no instructions given. 49 00:02:53,350 --> 00:02:56,410 Ejaaz: And what the model did was very, very interesting. 50 00:02:56,710 --> 00:03:02,270 Ejaaz: It scanned all those code bases and discovered 500 major security flaws. 51 00:03:03,260 --> 00:03:08,780 Ejaaz: Expert human AI security researchers couldn't discover in decades that they'd 52 00:03:08,780 --> 00:03:11,360 Ejaaz: been staring and using these exact code bases. 53 00:03:11,620 --> 00:03:15,600 Ejaaz: So Claude did in a couple of hours what many security researchers couldn't do. 54 00:03:15,680 --> 00:03:18,980 Ejaaz: We're talking about like millions of compute hours and time spent staring at 55 00:03:18,980 --> 00:03:20,200 Ejaaz: these code bases, testing it. 56 00:03:20,600 --> 00:03:25,760 Ejaaz: Claude, Opus 4.6 managed to figure this out. Now, this created a lot of excitement, 57 00:03:25,920 --> 00:03:27,140 Ejaaz: but also a lot of concern. 58 00:03:27,400 --> 00:03:31,380 Ejaaz: Now, because these AI security researchers had a good heart, 59 00:03:31,600 --> 00:03:33,100 Ejaaz: they weren't using this maliciously. 60 00:03:33,200 --> 00:03:35,900 Ejaaz: But if you could imagine that if this model had been placed to, 61 00:03:36,300 --> 00:03:40,440 Ejaaz: say, a malicious actor, they could have exploited these for many different reasons. 62 00:03:40,640 --> 00:03:44,840 Ejaaz: And so these exploits were surfaced and they were fixed. But the question now 63 00:03:44,840 --> 00:03:49,560 Ejaaz: becomes, what if a more powerful model was made more readily available to anyone 64 00:03:49,560 --> 00:03:55,160 Ejaaz: or an attacker, for example, a foreign adversary that could discover and exploit any future bugs? 65 00:03:55,500 --> 00:03:58,320 Ejaaz: That's the concern that's around that i have personally around 66 00:03:58,320 --> 00:04:01,740 Ejaaz: clode mythos or capybara this model is supposedly 67 00:04:01,740 --> 00:04:05,000 Ejaaz: meant to be a tier above anything that we've ever seen before apparently it 68 00:04:05,000 --> 00:04:08,580 Ejaaz: is amazing at discovering and exploiting exploits 69 00:04:08,580 --> 00:04:12,800 Ejaaz: so if it is let's say two orders of magnitude let's be conservative two orders 70 00:04:12,800 --> 00:04:16,760 Ejaaz: of magnitude better than opus 4.6 we could have a real problem on our hands 71 00:04:16,760 --> 00:04:21,700 Ejaaz: and so what anthropic has done now is they've started to slow release this secret 72 00:04:21,700 --> 00:04:26,860 Ejaaz: model mythos and capybara to cybersecurity experts first. 73 00:04:27,080 --> 00:04:30,680 Ejaaz: Why? Because they want them to figure out how they can harden their own defense 74 00:04:30,680 --> 00:04:32,840 Ejaaz: systems before they publicly release this model. 75 00:04:32,980 --> 00:04:37,500 Ejaaz: And someone, maybe a nefarious attacker might use it for unachievable gain. 76 00:04:37,640 --> 00:04:40,720 Josh: I think it's ironic that the company building what it describes as an AI with 77 00:04:40,720 --> 00:04:44,940 Josh: unprecedented cybersecurity capabilities leaked it because someone misconfigured their blog. 78 00:04:45,820 --> 00:04:49,000 Josh: Like the irony there is too strong. And you have to wonder, you have to really 79 00:04:49,000 --> 00:04:52,320 Josh: ask yourself the question, well, what if this model so smart that it's leaking 80 00:04:52,320 --> 00:04:56,820 Josh: itself if it's like poking holes to like let people secretly find it i don't know 81 00:04:57,430 --> 00:05:01,750 Josh: The one thing for sure is that, one, this model is going to be incredibly expensive 82 00:05:01,750 --> 00:05:02,950 Josh: to run currently, at least. 83 00:05:03,210 --> 00:05:05,990 Josh: That's part of the reason why we're not seeing it now. But the second is it's 84 00:05:05,990 --> 00:05:07,330 Josh: going to be unbelievably powerful. 85 00:05:07,530 --> 00:05:11,650 Josh: And the progress that we've had in the last year is going to probably look like 86 00:05:11,650 --> 00:05:14,090 Josh: nothing compared to what we're going to get for the next three quarters. 87 00:05:14,290 --> 00:05:17,190 Josh: The market also very much felt the effects of this because, oh my God, 88 00:05:17,290 --> 00:05:19,810 Josh: these stock charts look absolutely horrendous. 89 00:05:19,810 --> 00:05:23,290 Ejaaz: Yeah, CrowdStrike, which is like the major cybersecurity firm, 90 00:05:23,470 --> 00:05:25,990 Ejaaz: was down a couple billion on the news. 91 00:05:26,230 --> 00:05:28,990 Ejaaz: And Palo Alto Networks, which is another similar company that competes in this 92 00:05:28,990 --> 00:05:30,730 Ejaaz: firm, also suffered from this. 93 00:05:30,850 --> 00:05:34,910 Ejaaz: Now, these two charts that I'm looking at right now for these specific companies, 94 00:05:35,010 --> 00:05:38,210 Ejaaz: Josh, gives me a little PTSD or deja vu. 95 00:05:38,350 --> 00:05:43,190 Ejaaz: Because we were talking about this, I think, four weeks ago when Anthropic released 96 00:05:43,190 --> 00:05:45,350 Ejaaz: their security review clawed feature. 97 00:05:45,350 --> 00:05:49,650 Ejaaz: Which, you know, wasn't anything to do about Mythos, but basically helped review 98 00:05:49,650 --> 00:05:54,370 Ejaaz: the Vibe code that you produced using Claude. And so cybersecurity stocks dumped again. 99 00:05:54,730 --> 00:05:58,190 Ejaaz: This is happening seemingly on a monthly basis at this point. 100 00:05:58,350 --> 00:06:02,330 Josh: Even though these charts are down quite a bit, I'm not sure how concerned the 101 00:06:02,330 --> 00:06:06,650 Josh: market needs to be immediately because it appears as if this new model that's coming, 102 00:06:06,770 --> 00:06:10,250 Josh: this new cybersecurity specialist is really compute intensive, 103 00:06:10,250 --> 00:06:15,910 Josh: so much so that it's almost going to be impossible for them to run across all 104 00:06:15,910 --> 00:06:20,070 Josh: the accounts currently without some serious compression and iteration and figuring 105 00:06:20,070 --> 00:06:21,050 Josh: out how to run this more optimally. 106 00:06:21,430 --> 00:06:24,630 Josh: And it seems like we're starting to see those growing pains, right? It's like, 107 00:06:25,060 --> 00:06:27,820 Josh: As they're training models like this, as they're running them on their own servers, 108 00:06:27,840 --> 00:06:30,100 Josh: it's starting to affect the average user. 109 00:06:30,280 --> 00:06:33,080 Josh: I know sometimes I'll wake up and I'll feel like my opus is running a little 110 00:06:33,080 --> 00:06:35,860 Josh: bit dumber than it was the day before. And we actually have data that backs this up. 111 00:06:36,200 --> 00:06:42,340 Ejaaz: Yeah, so basically over the weekend, Clawed servers basically went down or were majorly impaired. 112 00:06:42,960 --> 00:06:46,260 Ejaaz: There were a bunch of different outages. People were reporting very, 113 00:06:46,420 --> 00:06:49,000 Ejaaz: very reduced quality in their interactions with Clawed. 114 00:06:49,120 --> 00:06:53,100 Ejaaz: And this has been kind of like a repeating trend over the last couple of weeks. 115 00:06:53,100 --> 00:06:55,000 Ejaaz: And now we might have the answer why. 116 00:06:55,340 --> 00:06:59,540 Ejaaz: Typically, major AI labs, the last public bit of information that we had was 117 00:06:59,540 --> 00:07:03,300 Ejaaz: from OpenAI's 2025 run of a major model. 118 00:07:03,520 --> 00:07:07,360 Ejaaz: They dedicated 30% of their available compute to a training run. 119 00:07:07,620 --> 00:07:11,540 Ejaaz: Now, the rumors state that for Claude Mythos, they've dedicated even more and 120 00:07:11,540 --> 00:07:14,780 Ejaaz: that's like the major architectural breakthrough that they've made. 121 00:07:15,240 --> 00:07:19,440 Ejaaz: If they've done that, that might be the reason why we aren't being able to use 122 00:07:19,440 --> 00:07:23,180 Ejaaz: the best version of Claude as consumers because they're too busy using the compute 123 00:07:23,180 --> 00:07:26,200 Ejaaz: to train the next step or tier in model. 124 00:07:27,060 --> 00:07:30,940 Ejaaz: I don't know if this is a good or bad thing, but one thing it definitely like 125 00:07:30,940 --> 00:07:34,360 Ejaaz: screams at me is like, we need a ton more compute. 126 00:07:34,500 --> 00:07:38,720 Josh: Big time. And it's amazing to think about how far we've come just in the last 127 00:07:38,720 --> 00:07:40,760 Josh: three months leading up to this moment here. 128 00:07:40,840 --> 00:07:43,640 Josh: I mean, when you think about over the winter break is when people really start 129 00:07:43,640 --> 00:07:44,700 Josh: to take vibe coding seriously. 130 00:07:44,940 --> 00:07:50,240 Josh: And since then, companies have gone from a very small percentage of code to almost 100% of code. 131 00:07:50,360 --> 00:07:54,980 Josh: I mean, this is saying 80% plus of all code deployed is written by CloudCo just for Anthropic. 132 00:07:54,980 --> 00:07:59,640 Josh: It's unbelievable we started with opus 4.5 which 133 00:07:59,640 --> 00:08:02,620 Josh: was released in november and then opus 4.6 came 134 00:08:02,620 --> 00:08:05,620 Josh: in february which took us from a 200 000 token contacts 135 00:08:05,620 --> 00:08:10,200 Josh: went into a million and now whatever this new thing is is going to really drive 136 00:08:10,200 --> 00:08:15,080 Josh: up the coding capabilities in a really big way and i think it's probably worth 137 00:08:15,080 --> 00:08:18,680 Josh: checking in on which model is going to be the strongest model which company 138 00:08:18,680 --> 00:08:21,380 Josh: has the best model through the end of june and thanks to polymarket we have 139 00:08:21,380 --> 00:08:22,800 Josh: some interesting stats on this. 140 00:08:22,920 --> 00:08:27,480 Josh: So the people are betting that Anthropic has a 66% chance of having the best 141 00:08:27,480 --> 00:08:29,860 Josh: AI model in June, which is huge. 142 00:08:30,020 --> 00:08:33,940 Josh: And that number has increased very significantly recently. If you look just 143 00:08:33,940 --> 00:08:38,460 Josh: back in February, it was Google who was the heavy favorite with a almost 80% 144 00:08:38,460 --> 00:08:41,480 Josh: chance or 70% chance of having the best model. 145 00:08:41,600 --> 00:08:45,440 Josh: That has changed recently in a big way, perhaps because of this leak. 146 00:08:45,560 --> 00:08:48,500 Josh: But I'm not sure if this is fully up to date and 147 00:08:48,500 --> 00:08:53,160 Josh: it may be missing some information because we have some news on open ai and 148 00:08:53,160 --> 00:08:56,480 Josh: google who are planning to release something really important too and thank 149 00:08:56,480 --> 00:08:58,980 Josh: you for probably for sponsoring that part of the show but let's talk about open 150 00:08:58,980 --> 00:09:03,380 Josh: ai there's a new code name spud model that's coming and this is probably going 151 00:09:03,380 --> 00:09:05,940 Josh: to be the mythos competitor so what is this looking like yeah 152 00:09:05,940 --> 00:09:09,520 Ejaaz: Um that's the issue we don't really know all of these models we don't have the 153 00:09:09,520 --> 00:09:11,400 Ejaaz: the specs we need the specs to talk about them. 154 00:09:12,090 --> 00:09:15,910 Ejaaz: There's a few trends or patterns that are happening amongst the hottest, 155 00:09:16,190 --> 00:09:18,810 Ejaaz: or should I say, top two or three AI labs. 156 00:09:19,370 --> 00:09:23,630 Ejaaz: We've got Anthropic Releasing Mythos, which is their AGI or pre-AGI model, 157 00:09:23,810 --> 00:09:25,070 Ejaaz: a massive, massive leap ahead. 158 00:09:25,450 --> 00:09:30,110 Ejaaz: OpenAI is working on the same thing. They've been secretively working on a larger model. 159 00:09:30,270 --> 00:09:32,670 Ejaaz: This has gone through a few different names. If you remember, 160 00:09:32,770 --> 00:09:35,850 Ejaaz: Josh, by the end of the year, I think it was referred to as codename Sprout. 161 00:09:36,070 --> 00:09:39,650 Ejaaz: And now it's referred to as Spud. So I don't know if that implies that it's 162 00:09:39,650 --> 00:09:41,150 Ejaaz: grown massively since then. 163 00:09:41,330 --> 00:09:44,970 Ejaaz: It's growing. But these models are supposedly meant to be anywhere between 10 164 00:09:44,970 --> 00:09:47,530 Ejaaz: to 20 trillion parameter models. 165 00:09:47,690 --> 00:09:51,670 Ejaaz: Now, for context, the largest models that we currently look at right now is 166 00:09:51,670 --> 00:09:52,830 Ejaaz: between one to two trillion. 167 00:09:53,010 --> 00:09:57,070 Ejaaz: So this is a major order of magnitude larger model. 168 00:09:57,530 --> 00:10:00,570 Ejaaz: They're going to be compute intensive. They're going to be very expensive to serve. 169 00:10:00,690 --> 00:10:04,190 Ejaaz: So we need to figure out how to scale AI infrastructure and a bunch of other things. 170 00:10:04,310 --> 00:10:08,110 Ejaaz: But OpenAI's model is codenamed Spud, and it's meant to be the competitor to 171 00:10:08,110 --> 00:10:15,170 Ejaaz: Mythos. People are anticipating that it might be something like GPT 5.5 or rather GPT 6. 172 00:10:15,370 --> 00:10:19,690 Ejaaz: So again, a tier above what we see today. It's going to be advanced in coding, 173 00:10:19,910 --> 00:10:22,250 Ejaaz: reasoning, and a lot of the things Anthropics is as well. 174 00:10:22,710 --> 00:10:26,690 Ejaaz: When I look at this, Josh, personally to me, this seems to be, 175 00:10:26,790 --> 00:10:30,970 Ejaaz: one, a massive bid to try and leapfrog each other. 176 00:10:31,070 --> 00:10:34,810 Ejaaz: And number two, maybe try and juice their numbers ahead of a potential IPA. 177 00:10:34,890 --> 00:10:37,350 Ejaaz: I don't know whether your reaction to this is the same, but that's like my gut 178 00:10:37,350 --> 00:10:39,130 Ejaaz: reaction when I read news like this. 179 00:10:39,410 --> 00:10:42,610 Josh: Yeah, it's probably both. They want to juice up things before the IPO, 180 00:10:42,710 --> 00:10:43,790 Josh: but they also just want to win. 181 00:10:43,930 --> 00:10:47,950 Josh: And I have some pretty strong speculations just based on vibes of what this is going to look like. 182 00:10:48,170 --> 00:10:51,890 Josh: I think we've been seeing this recent convergence around OpenAI, 183 00:10:52,150 --> 00:10:56,470 Josh: particularly on focus and on really dialing in what they're focused on. 184 00:10:56,590 --> 00:11:01,210 Josh: And we saw a big move last week when they removed Sora. They totally destroyed Sora. 185 00:11:01,490 --> 00:11:06,330 Josh: They moved a lot of the teams together. They made their chief of product, 186 00:11:06,330 --> 00:11:11,690 Josh: um the chief of like agi release and it appears as if they're building a mega 187 00:11:11,690 --> 00:11:12,970 Josh: app based on the rumors so 188 00:11:13,510 --> 00:11:17,990 Josh: Part of the reason why I have a difficult time using OpenAI's products is there's 189 00:11:17,990 --> 00:11:19,270 Josh: kind of spread out everywhere. 190 00:11:19,810 --> 00:11:23,530 Josh: There's like the Sora app was one, there's Codex, then there's their browser, 191 00:11:23,850 --> 00:11:26,130 Josh: then there's ChatGPT, and there's a lot of different software. 192 00:11:26,350 --> 00:11:29,510 Josh: And the same is true with their models, or it was at least, where there was 193 00:11:29,510 --> 00:11:33,290 Josh: GPT 5.3 Codex, and there was 5.3 High, Mid, Low. 194 00:11:33,630 --> 00:11:37,290 Josh: There's all these different models that really complicate things and confuse things. 195 00:11:37,470 --> 00:11:42,870 Josh: With 5.4, they made a singular model. Now 5.4 does your coding and it does the reasoning all in one. 196 00:11:42,870 --> 00:11:47,390 Josh: And what I suspect with this new model, Codename Spud, is going to be the kind 197 00:11:47,390 --> 00:11:50,870 Josh: of pinnacle of this focus, where I'm hoping they release this with their new 198 00:11:50,870 --> 00:11:53,050 Josh: application, with a singular model. 199 00:11:53,230 --> 00:11:56,150 Josh: So there's one model that is all-knowing. There's one application, 200 00:11:56,170 --> 00:11:59,390 Josh: similar to what Anthropic does with the Cloud Desktop app, that has all of the 201 00:11:59,390 --> 00:12:00,470 Josh: functionality under one roof. 202 00:12:00,650 --> 00:12:03,590 Josh: And I think they're going to probably use this as a point to really... 203 00:12:04,320 --> 00:12:08,460 Josh: Lean into that focus instead of distributing this across a lot of different areas. 204 00:12:08,580 --> 00:12:12,260 Josh: And I'm hopeful that that will meaningfully change OpenAI more so than it'll 205 00:12:12,260 --> 00:12:16,120 Josh: change Anthropik because it actually changes the way that users interface with 206 00:12:16,120 --> 00:12:17,780 Josh: the product and it becomes a much better product. 207 00:12:17,980 --> 00:12:22,320 Ejaaz: Yeah, I think for the majority of last year, I was pretty upset with the way 208 00:12:22,320 --> 00:12:25,580 Ejaaz: that SAM and OpenAI were focusing on so many different things. 209 00:12:25,680 --> 00:12:27,740 Ejaaz: I was just like, just focus on creating a really good model. 210 00:12:28,140 --> 00:12:31,680 Ejaaz: You're being left behind in coding, Anthropik's eating your lunch, like figure this out. 211 00:12:32,080 --> 00:12:34,920 Ejaaz: And then since their code read of like, what was it, November last year, 212 00:12:35,040 --> 00:12:39,300 Ejaaz: they've been like reallocating compute, money, data, and all their resources 213 00:12:39,300 --> 00:12:42,900 Ejaaz: to focus on building the best general model and the best coding model. 214 00:12:43,000 --> 00:12:44,920 Ejaaz: So we're starting to see the fruits of that labor. 215 00:12:45,100 --> 00:12:48,440 Ejaaz: I have a lot of faith now in OpenAI that they're going to produce a really good 216 00:12:48,440 --> 00:12:51,740 Ejaaz: product that will compete with the likes of Anthropic, which have been eating their lunch. 217 00:12:51,860 --> 00:12:55,840 Ejaaz: When I look at like the last week, it seems like it's pretty negative for OpenAI. 218 00:12:55,960 --> 00:12:57,120 Ejaaz: You mentioned that they killed Sora. 219 00:12:57,300 --> 00:13:00,360 Ejaaz: They also killed the $1 billion deal that they had signed with Disney. 220 00:13:00,620 --> 00:13:04,560 Ejaaz: And they also shut down ChatGPT adult mode and a bunch of like consumer shopping 221 00:13:04,560 --> 00:13:06,720 Ejaaz: apps and their like app marketplace as well. 222 00:13:06,820 --> 00:13:09,480 Ejaaz: They're just focused on these few things right now. 223 00:13:09,620 --> 00:13:14,820 Ejaaz: But then the other thing is Sam is also kind of defaulting on a few of the major 224 00:13:14,820 --> 00:13:16,480 Ejaaz: GPU and data center deals, right? 225 00:13:16,640 --> 00:13:20,660 Ejaaz: So we had the OpenAI and Oracle Abilene deal fall through where they couldn't 226 00:13:20,660 --> 00:13:23,660 Ejaaz: finance it for a variety of different reasons. 227 00:13:24,100 --> 00:13:28,600 Ejaaz: Then the other thing is they're defaulting on purchasing up to 40% of the world's 228 00:13:28,600 --> 00:13:32,000 Ejaaz: memory supply because they haven't figured out their finances right now. 229 00:13:32,140 --> 00:13:36,400 Ejaaz: So I think that OpenAI is going through kind of like a puberty period where 230 00:13:36,400 --> 00:13:39,060 Ejaaz: they're figuring their stuff out and where to reallocate resources. 231 00:13:39,280 --> 00:13:40,480 Ejaaz: But I think they're going to pull through. 232 00:13:40,760 --> 00:13:43,440 Josh: And it also seems like this is indeed a serious breakthrough. 233 00:13:43,620 --> 00:13:46,200 Josh: I mean, Sam, in an internal memo that got leaked out to employees, 234 00:13:46,520 --> 00:13:49,120 Josh: he said things are moving faster than many of us expected. 235 00:13:49,580 --> 00:13:53,200 Josh: And he called it a very strong model that can really accelerate the economy. 236 00:13:53,700 --> 00:13:59,260 Josh: That seems like pretty large claims to make internally with employees who are 237 00:13:59,260 --> 00:14:01,620 Josh: also kind of in the know and aware of what's going on. 238 00:14:01,860 --> 00:14:06,220 Josh: And I just think that a lot of us who are sitting outside these labs are not 239 00:14:06,220 --> 00:14:10,940 Josh: entirely wrapping our head around how much progress is actually about to hit 240 00:14:10,940 --> 00:14:13,280 Josh: us over the next couple of months with these new model releases. 241 00:14:13,460 --> 00:14:15,620 Josh: It seems like they're step function improvements. 242 00:14:15,980 --> 00:14:19,980 Josh: And one of the employees from OpenAI actually hinted that Spud contains a capability 243 00:14:19,980 --> 00:14:23,240 Josh: that is very different from what we've seen before. So while there aren't specifics, 244 00:14:23,460 --> 00:14:27,020 Josh: there are clearly a lot of these huge novel breakthroughs incoming, 245 00:14:27,280 --> 00:14:29,040 Josh: which is worth looking out for. 246 00:14:29,260 --> 00:14:33,220 Josh: There's one final model release, model leak that we have from Google, 247 00:14:33,260 --> 00:14:36,440 Josh: who has been doing well, kind of chugging along slowly in the background. 248 00:14:36,540 --> 00:14:38,260 Josh: And this is called Agent Smith. 249 00:14:38,580 --> 00:14:40,920 Josh: It's a secret AI tool. Do you have any information on this one, EJS? 250 00:14:41,020 --> 00:14:44,980 Ejaaz: Yeah, so there was like a leaked report from an insider at Google. 251 00:14:45,520 --> 00:14:49,020 Ejaaz: Apparently, Google employees are using a new internal tool called AgentSmith 252 00:14:49,020 --> 00:14:53,340 Ejaaz: that can automate tasks such as coding, according to three people that were familiar with it. 253 00:14:53,700 --> 00:14:57,820 Ejaaz: The way that this product is supposed to work is within their Vibe coding platform 254 00:14:57,820 --> 00:15:03,020 Ejaaz: called Antigravity, which exists today but hasn't really had a major upgrade 255 00:15:03,020 --> 00:15:07,440 Ejaaz: for, let's say, a couple months now, which is like an eternity in the AI world. 256 00:15:07,440 --> 00:15:11,680 Ejaaz: So they're releasing a new AI model called Agent Smith that is supposed to take 257 00:15:11,680 --> 00:15:16,300 Ejaaz: a multi-agent approach and use an upgraded version of Gemini 3.1. 258 00:15:16,460 --> 00:15:20,160 Ejaaz: So it's probably not going to be 3.1. It might be 3.5 or maybe even 4. 259 00:15:20,280 --> 00:15:23,620 Ejaaz: Again, another order of magnitude leap up. So what we're seeing here is Google 260 00:15:23,620 --> 00:15:27,840 Ejaaz: working on an AI coding model competitor to try and catch up to Anthropic and 261 00:15:27,840 --> 00:15:29,500 Ejaaz: the likes of OpenAI's codecs. 262 00:15:29,740 --> 00:15:33,500 Ejaaz: You've got OpenAI trying to reallocate resources and focus on building the best 263 00:15:33,500 --> 00:15:36,440 Ejaaz: general model and catch up with Anthropic, which they have at coding. 264 00:15:36,440 --> 00:15:41,320 Ejaaz: Then you have Anthropic trying to keep these two at bay and make the next order 265 00:15:41,320 --> 00:15:44,800 Ejaaz: of magnitude up spending all their compute but coming at the expense of serving 266 00:15:44,800 --> 00:15:48,780 Ejaaz: their existing users which they're adding like a million a day reporting you know 267 00:15:49,090 --> 00:15:53,250 Ejaaz: Claude servers being down and reduced quality of usage. So this is a very, 268 00:15:53,430 --> 00:15:56,730 Ejaaz: I can like feel the tension in the air between these three companies right now. 269 00:15:56,850 --> 00:15:58,050 Ejaaz: I don't know what Mets is doing. 270 00:15:58,230 --> 00:16:00,910 Ejaaz: I don't know where Grok is. I'm rooting for them. I hope they catch up. 271 00:16:01,010 --> 00:16:05,550 Ejaaz: But it seems to be these three major competitors right now that are in the running for winning this race. 272 00:16:05,710 --> 00:16:08,770 Josh: They're firing. I mean, in the last 90 days, since we started this year to now, 273 00:16:08,870 --> 00:16:11,390 Josh: we went from 200,000 context windows to a million. 274 00:16:11,670 --> 00:16:14,750 Josh: We went from these coding assistants to compiler writing 275 00:16:14,750 --> 00:16:17,830 Josh: agents who are completely capable of writing a very small amount 276 00:16:17,830 --> 00:16:21,070 Josh: now over a quarter of google's production software and 80 plus 277 00:16:21,070 --> 00:16:23,950 Josh: of anthropic software everything we learned this week the frontier is 278 00:16:23,950 --> 00:16:26,910 Josh: going to keep moving faster and faster so we're in 279 00:16:26,910 --> 00:16:29,590 Josh: for a crazy q2 q3 q4 just a 280 00:16:29,590 --> 00:16:32,690 Josh: crazy 2026 and as all these things happen as these 281 00:16:32,690 --> 00:16:35,610 Josh: ipos start to happen and they get even more fundraising to deploy 282 00:16:35,610 --> 00:16:38,410 Josh: these ai data centers at scale things are really going 283 00:16:38,410 --> 00:16:42,610 Josh: to get weird in a hurry but we will be here to cover it as always um if you 284 00:16:42,610 --> 00:16:45,730 Josh: enjoyed this episode please don't forget share it with your friends uh like 285 00:16:45,730 --> 00:16:50,190 Josh: it on youtube don't forget to subscribe if you listen on a podcast player like 286 00:16:50,190 --> 00:16:53,870 Josh: spotify or rss you could rate us five stars there it's always really appreciated 287 00:16:53,870 --> 00:16:56,650 Josh: you just any final notes before we sign off for the day we've 288 00:16:56,650 --> 00:17:00,690 Ejaaz: Been absolutely killing it on our side uh loads of new subscribers loads of 289 00:17:00,690 --> 00:17:04,870 Ejaaz: new listeners thank you guys so much for for joining us um and yeah i have a 290 00:17:04,870 --> 00:17:07,650 Ejaaz: request because we always like to give out homework at the end of the episode 291 00:17:07,650 --> 00:17:13,050 Ejaaz: um if you're listening to this and you are a insider at anthropic open air or 292 00:17:13,050 --> 00:17:16,070 Ejaaz: google and you are willing to give an anonymous tip to our accounts, 293 00:17:16,250 --> 00:17:20,070 Ejaaz: please spin up an Anon account on x slash Twitter and DM us. 294 00:17:20,170 --> 00:17:20,930 Ejaaz: I would love to hear from you. 295 00:17:21,350 --> 00:17:23,810 Josh: That'd be great. Well, yeah, thank you guys for watching. We'll see you in the next one.