Meta Poised to Gain as Anthropic Pushes Frontier AI Slowdown

Anthropic呼吁放缓前沿AI发展，Meta或成最大赢家 Anthropic推动强制放缓节奏的背后，掩盖了IPO前为遏制竞争对手而面临的利润率压力，而强制性的支出限制将消除Meta的资本支出阴霾并扩大自由现金流。 Joseph Carlson分析了围绕政府强制对人工智能进行“协同调速”呼吁所引发的恐慌，认为Anthropic和OpenAI的末日论论调实际上掩盖了企业在上市前遏制不可持续的资本支出、保住自身领先地位的企图。Carlson指出Meta是这一局面的最大受益者，强调强制放缓资本支出将消除华尔街最大的担忧，扩大自由现金流并促使Meta股票估值重估；同时，无论模型是否走向商品化，Mark Zuckerberg对消费者效用和分发渠道的重视都能为公司提供保护。节目还重点提及了Zuckerberg组建Meta Superintelligence以及推出Muse等实用智能体工具，最后对Hunter Biden的笔记本电脑meme币灾难性的发行与崩盘进行了令人难以置信的复盘回顾。 Anthropic首席执行官Dario Amodei与OpenAI的Sam Altman呼吁由政府强制执行“协同调速”以放缓前沿AI发展，引发了Nvidia和ASML等半导体龙头的剧烈抛售。 (0:00) Meta Platforms成为AI放缓的最大受益者：尽管营收同比增长27%，但其市盈率仅约21倍，外部对AI资本支出的限制将立即使其自由现金流扩大并触发估值重估。 (4:05) 无论AI走向商品化还是保持专有，Mark Zuckerberg都已为Meta做好了防范，其通过向Scale AI投资143亿美元将FAIR重组为Meta Superintelligence (MSI)，并向36亿日活跃用户推出了Muse智能体助手。 (6:49) Anthropic推动全行业暂停训练并非纯粹出于利他主义，而是契合了其抑制超5000亿美元支出承诺、实现IPO前盈利以及通过监管俘获巩固其前沿领先地位的需求。 (26:19) 在“每周败笔”环节中，Hunter Biden的“Laptop” meme币在发行几分钟内从约190美元暴跌至2美元以下，导致约80%的散户买家亏损，而少数内部人士钱包各自套现高达100万美元。 (31:46)

Joseph Carlson argues that calls by frontier artificial intelligence labs for government-enforced development caps reflect commercial self-interest rather than purely altruistic safety concerns, positioning Meta Platforms as the primary beneficiary of any regulatory or market-driven slowdown. Commercial Incentives Behind 'Coordinated Pacing' Anthropic CEO Dario Amodei, supported by OpenAI CEO Sam Altman, publicly advocated for state-enforced 'coordinated pacing' of frontier artificial intelligence following viral whistleblower warnings of catastrophic existential risk. Carlson contends this proposal functions primarily as regulatory capture to protect incumbent leads. Anthropic faces over half a trillion dollars in reported infrastructure spending commitments and needs to show operating profitability ahead of an anticipated initial public offering, making continuous capital expenditure on GPU clusters a severe financial burden. While Amodei argues that coordinated deceleration allows safety research without forfeiting commercial advantage, Carlson points out that proprietary labs are seeing their moats narrowed by open-source alternatives and well-capitalized competitors. If frontier models were genuinely viewed as an existential threat, Anthropic could pause its training unilaterally. Instead, demanding an industry-wide freeze locks in an Anthropic-OpenAI duopoly while shielding cash-burning labs from being overtaken by self-funded competitors like Meta. Meta's Financial Upside in an AI Deceleration (4:05) If regulatory mandates or market pressures force an artificial intelligence slowdown, Meta stands to benefit more than any peer. Unlike Amazon, Google, or Microsoft, Meta does not rent cloud infrastructure to third parties; its compute investments are dedicated entirely to user engagement and ad targeting. Wall Street has priced Meta at an optically low valuation of roughly 21 times earnings—despite 27% year-over-year revenue growth—primarily due to anxiety over Mark Zuckerberg's unconstrained capex budget. An enforced spending curb would immediately relieve investor concerns, expand profit margins, and boost free cash flow. Furthermore, Meta occupies an asymmetric position regardless of how model development unfolds. If frontier models become commoditized open-weight software, proprietary labs will face margin collapse while the economic surplus shifts to Meta’s massive distribution network of 3.6 billion daily active users. If models remain highly differentiated, Meta's internal development keeps its systems within roughly six months of the frontier, insulating it from third-party platform dependence like that seen at Apple. Pragmatic Utility and the Meta Superintelligence Pivot (8:40) Carlson contrasts the apocalyptic rhetoric of frontier labs with Meta's pragmatic approach. Rather than treating artificial intelligence as a dangerous deity comparable to the Manhattan Project, Zuckerberg views it as a utility akin to the internet, intended to lower marginal operating costs and power consumer workflows. Meta is applying this focus through products like Muse, an agentic application boasting an early 4.9-star rating across 3,800 reviews that automates practical tasks such as OpenTable dining reservations, inbox cleaning, and multi-site shopping checkouts via Stripe. To ensure technical independence, Zuckerberg leveraged his controlling voting power to execute an 11-week reorganization following the release of Llama 4. He absorbed the Fundamental AI Research (FAIR) group into Meta Superintelligence (MSI), secured a $14.3 billion stake in Scale AI to appoint Alexandr Wang as AI lead, and recruited high-profile talent—including ChatGPT co-creator Shengjia Zhao and former GitHub CEO Nat Friedman—positioning the firm to compete aggressively if the compute race continues uninterrupted. Speculative Excess: The Laptop Meme Coin Implosion (31:46) In the program's 'Fail of the Week' segment, Carlson examines the rapid collapse of the Hunter Biden 'Laptop' meme coin. Marketed under the guise of charitable community-building and political retribution, the token duplicated speculative celebrity crypto launches and cratered from roughly $190 to under $2 within minutes of trading. On-chain data revealed that roughly 80% of participating retail buyers lost money almost instantaneously. Over 700 traders lost between $1,000 and $10,000, 112 participants dropped between $10,000 and $100,000, and two individual wallets lost between $100,000 and $1 million. In contrast, financial gains were concentrated within approximately ten insider wallets that extracted up to $1 million each, illustrating an ongoing pattern of retail speculation chasing poorly executed promotional schemes. The analysis leaves unresolved whether global regulators will actually enact coordinated artificial intelligence restrictions, or whether Meta's aggressive capital expenditures will continue unabated until open-source alternatives fully eliminate proprietary model premiums.

[0:00] Dario Amodei: without sacrificing commercial advantage [8:40] Mark Zuckerberg: does math and solves a limited set of problems humans face and is otherwise simply useful and cool [17:30] Mark Zuckerberg: I'm not slowing down [20:43] Dario Amodei: I agree with Jacob much more than I disagree with him [31:46] Hunter Biden: lost money on the Trump grift

[0:00] Welcome back everyone. Today on the [0:01] Joseph Carlson show, there is an allout [0:03] war against AI and it's being led by [0:06] none other than the top companies [0:07] themselves, primarily Anthropic. We know [0:10] that Jacob Coxin, the whistleblower that [0:12] quit his job at Anthropic and said that [0:14] AI is going to potentially kill all of [0:16] humanity within the next decade. That of [0:18] course happened last week. But even from [0:20] then, things started to boil over the [0:22] weekend. We have Daario stepping up [0:25] saying that he wants to pace the [0:27] frontier, calling for a big slowdown. [0:30] Now, Daario is the leading company. [0:32] Anthropic is the biggest one in AI, and [0:34] they're specifically calling for greater [0:36] government regulation. And while this is [0:38] going on, there's going to be some [0:39] companies that benefit and some that [0:41] don't. In this episode, I'll be [0:42] highlighting what I believe is the top [0:44] company to benefit from a AI slowdown, [0:47] but we'll also be looking at Daario and [0:48] his motives himself. Why would Daario [0:51] really call for a slowdown? Why does he [0:53] really want to pace the frontier? Is it [0:56] really because of safety concerns? Is he [0:58] really just that scared of AI? Well, [1:00] we'll be looking at that, the safety [1:01] issue, but also many other potential [1:03] reasons that Daario would specifically [1:05] benefit from a slowdown in AI. We have [1:07] all of that to get into, plus we have [1:09] the fail of the week, which in this case [1:10] is the Hunter Biden memecoin, which is [1:13] the laptop memecoin. You may have [1:15] thought memecoins aren't a thing [1:16] anymore. Well, apparently they still [1:18] are. We'll be looking at that as well. [1:19] So, we have a ton to get to in this [1:20] episode. Let's go ahead and jump in. [1:22] Now, we start things off today by [1:23] looking at the great battle that [1:24] happened online over the weekend. If [1:27] you're not chronically online like I am, [1:29] and hopefully you aren't, you may have [1:30] missed that on X many other social [1:32] circles, there was a great battle online [1:34] between many people giving takes and [1:36] thoughts about AI regulation. What [1:39] sparked all of these concerns? Well, [1:41] it's been culminating for a while. We [1:42] had the hugging face hack that was [1:44] exaggerated to a huge extent. Everything [1:46] was talked about like it's lifelike and [1:49] AI can simply decide to do things that [1:51] it wasn't instructed to. You also have [1:53] the fact that you have Jacob Coxin, the [1:55] whistleblower that looks a little bit [1:57] like Harry Potter, that said that he [1:59] believes that there's a good risk that [2:00] we'll all be annihilated by AI over the [2:02] upcoming decade. And he went on to many [2:04] shows including Anderson Cooper for [2:06] highly viral interviews where he laid [2:08] out his extremely doomerish take. And [2:10] that caused a lot of hysteria. a lot of [2:12] people that weren't aware of these risks [2:14] and now they're even more scared of AI [2:16] than they were before. But it didn't end [2:19] there. After that happened, Daario, the [2:21] CEO of the biggest AI company in the [2:24] world, the one that's in the lead right [2:25] now, which is Anthropic, decided to [2:27] release a long write up. And this was [2:29] interesting timing. It only came days [2:31] after Jacob Cox's whistleblower moment. [2:33] But you must have thought that this was [2:34] a good time to release it. The writeup [2:36] is called We Must Pace the Frontier. And [2:38] it's not too dissimilar from other [2:39] write-ups he's done. Now, of course, as [2:40] you would expect with the big concerns [2:42] of different employees at Anthropic, [2:44] espousing their views, then we have the [2:46] CEO also writing a big letter saying [2:48] that we need to slow things down, that [2:50] it's too dangerous. This has caused even [2:52] bigger panic. It hit a boiling point. It [2:55] started to boil over over the weekend [2:57] and now investors are genuinely becoming [2:59] concerned that there's going to be a [3:01] serious slowdown in the frontier of AI. [3:03] Many of the stocks that are AI stocks [3:05] like ASML, Nvidia, these type of stocks [3:08] have gone down today. And we have other [3:10] leaders, including one important one, [3:12] Sam Alman of OpenAI, joining in with [3:15] Daario to some degree, saying, "Yeah, we [3:17] could slow down the frontier and be be a [3:20] little bit more cautious." Now, the most [3:21] important part of the slowdown that [3:23] Daario is calling for is not just to [3:25] slow down for his company that is [3:27] leading the pack or even Open AI. He's [3:29] not wanting just those two companies [3:31] that are the most dangerous one with the [3:32] biggest models to slow down. Now, see, [3:35] Daario wants everyone to slow down, no [3:37] matter really where they are in this [3:39] race. He says a coordinated pacing [3:41] strategy would give Frontier AI [3:43] developers the time to do this work [3:45] without sacrificing commercial [3:47] advantage. It has to be a coordinated [3:49] pacing. He's calling for regulation that [3:51] doesn't just slow down OpenAI or [3:53] Enthropic, but it also slows down [3:55] Microsoft and Google and Meta and all [3:58] the rest of them. They have to have a [4:00] coordinated pacing. Now, I'll speak to [4:02] Daario's incentives for having a [4:03] coordinated slowdown in a minute, but [4:05] before we jump into that, I think it's [4:07] good to highlight what companies would [4:09] really benefit if this was the case. [4:10] Meta will be one of, if not the top [4:13] benefiting company from an overall [4:15] coordinated AI slowdown. Now, there's a [4:17] few reasons that Meta would be the [4:18] biggest beneficiary in a real [4:20] coordinated slowdown. Let's assume that [4:23] there is this true genuine coordinated [4:25] AI slowdown and pacing. That these [4:28] companies don't spend as much money on [4:29] training new models. They don't spend as [4:31] much money on building out capex and [4:33] warehouses that all the hundreds of [4:35] billions of dollars spent every single [4:37] quarter that these companies are [4:38] spending. All of that is slowed down. [4:40] All of it is regulated. If that is the [4:43] case, what company is spending more [4:46] money on capex, more money on AI [4:48] development with less direct revenue [4:50] from AI than Meta? Right now, there [4:52] really isn't one. Amazon spends a lot of [4:54] money on AI, but they have AWS that [4:56] directly monetizes AI. We have Google [4:59] with Google Cloud. We have Microsoft [5:01] with Microsoft Azure. All of them are [5:02] renting out their capacity. The only [5:04] company that's really spending on it as [5:06] an indirect way of creating revenue is [5:08] Meta. And the ways that Meta actually [5:11] makes money from AI, their core [5:13] business, that'll continue to monetize. [5:15] They can continue making improvements on [5:17] that without the leading AI. Meta does [5:19] not need to be on the bleeding edge to [5:21] monetize their core assets directly. So [5:24] Meta is in this unique spot where it's a [5:26] company that is growing revenue faster [5:28] than almost any other big tech company. [5:30] It's not growing as fast as the AI labs [5:32] themselves, but it's growing very [5:34] quickly. The revenue grew 27% [5:36] year-over-year on a trailing 12-month [5:38] basis. Yet, even with this incredible [5:40] revenue growth, Meta Stock really hasn't [5:41] been rewarded. It's been flat all year, [5:44] hasn't made any gains, and it continues [5:46] to trade at a very low valuation [5:47] optically. And the reason why is because [5:50] of this the capex investors look at all [5:53] this capital expenditures and they can't [5:55] link it to a direct revenue stream. If [5:57] there was this coordinated slowdown, [5:59] Meta would spend less on capex. Their [6:01] future growth plans would be tapered in [6:03] a little bit. They'd continue to spend, [6:05] but it would be at a much slower pace. [6:07] But yet, their monetization would likely [6:09] not slow down at all because Meta has [6:11] more than enough capacity to continue [6:12] monetizing their core businesses and [6:14] continue growing them. And that would [6:16] likely end up with the stock rerating. [6:17] and Meta stock could easily be rerated. [6:20] Meta currently trades at a 21 PE ratio. [6:22] So if we did get this coordinated [6:24] slowdown, Meta has probably the best [6:26] setup. It has incredible revenue growth [6:28] that will continue even with a capex [6:30] slowdown. The capex slowdown would make [6:32] the margins and the free cash flow [6:34] profile improve of the company and it [6:36] would make investors less concerned [6:37] about Mark Zuckerberg's willingness to [6:39] spend endlessly on AI. It'd be another [6:41] third party factor reigning him in. All [6:44] of these would likely be appreciated by [6:45] the market, at least in the short term. [6:47] And there's a lot more benefits that [6:48] Meta has. In fact, Colossus did an [6:50] incredible writeup of meta. It's a [6:52] lengthy one, so I'll just highlight a [6:54] few things here that show how well Meta [6:56] is set up right now. In fact, one of the [6:57] arguments that they make, I think, is a [6:59] very good one. How meta wins either way [7:01] if AI is commoditized or not. If AI [7:04] commoditizes, then anthropic and open AI [7:06] go to zero. Now going to zero is [7:08] probably a little bit of an exaggeration [7:10] but theoretically if AI becomes a [7:13] complete commodity meaning everybody can [7:15] just use openw weight models for [7:16] essentially free then anthropic and open [7:19] AI are worth dramatically less money [7:21] they're not able to monetize and value [7:23] would acrue instead to the complements [7:25] of meta the stuff that they already [7:26] dominate like distribution attention [7:28] personalization and commerce if AI [7:31] doesn't commoditize which is a real [7:33] possibility then at least he is not in [7:35] his competitor's prison the same way [7:36] that Apple's been. Apple has no AI for [7:39] themselves. They're just reliant on [7:40] other people. They are dependent on AI [7:42] eventually commoditizing. And all he has [7:44] to do is remain within 6 months of the [7:46] frontier, which he's already close to. [7:49] Heads, he wins, tails, he wins. Many [7:51] people view Mark Zuckerberg's approach [7:53] to this is aggressive and unwieldy. He's [7:55] like unpredictable. He's going out and [7:57] spending so much money that he's just [7:59] crazy. And that's a danger or risk in [8:01] the stock. But if you actually look at [8:03] it this way, what Mark Zuckerberg's [8:05] doing has reduced risk in Meta. It's [8:07] made the investor more safe in their [8:09] future earnings. He's made it so that if [8:11] AI becomes commoditized, all of that [8:14] acrru to Meta's already huge formidable [8:16] businesses. There'll be more shopping, [8:18] more commerce, more people using [8:20] advertisements on Meta. And if it [8:22] doesn't commoditize, well, he has his [8:23] own AI. Either way, he has further [8:25] insulated the future of meta and reduced [8:27] risk in the process. There's also the [8:29] factor of how Zuckerberg even looks at [8:31] AI. For example, Daario and Sam have [8:34] continually warned about the doom of AI, [8:37] saying it's going to take your jobs. [8:38] Half of all new jobs from college [8:40] graduates are going to go to AI. Well, [8:42] aside from the fact that that really [8:43] hasn't happened yet, the unemployment [8:45] rate has continually gone down. There's [8:47] other ways in which anthropic and open [8:49] AI have been doomerish. They've talked [8:52] not just about jobs going away, but [8:54] Daario specifically has talked [8:55] continually about the bare cases, the [8:57] edge cases of AI, the cases where it can [9:00] hack things and create bioweapons. We [9:02] hear these narratives all over the [9:03] place. There's little focus on the [9:05] upside of it. How it can help humanity. [9:07] How it can further individuals. AI has [9:09] asymmetric risks. It does have downside [9:11] risks, but it has dramatically more use [9:13] cases and positive benefits for society. [9:16] And Mark Zuckerberg has been one of the [9:18] few leaders to actually highlight those [9:20] upsides. For example, it says here that [9:22] AI is not in fact God. Instead, it does [9:26] math and solves a limited set of [9:28] problems humans face and is otherwise [9:30] simply useful and cool. Anthropic and to [9:33] a lesser extent, open AI have trouble [9:35] ever accepting this fact. Zuckerberg [9:37] does not. Zuckerberg does not think he's [9:39] a god wielding AI. He hasn't viewed it [9:42] that way. He's just viewed it as a thing [9:44] that can really help people out. Daario [9:46] and Sam have a god complex. They truly [9:49] believe that they've unlocked this magic [9:50] vault of technology that will destroy [9:52] all humanity unless they stop it. And [9:54] it's normal for people in great [9:56] leadership positions to start to view [9:58] themselves as a little bit more powerful [9:59] than they actually are or a little bit [10:01] more important than they actually are. [10:03] And we see that especially with Daario. [10:05] For Zuckerberg, that's not the case. [10:07] He's never talked about AI this way. He [10:10] has not spent a decade comparing his [10:12] company to the Manhattan project. We [10:14] have people like Daario and Sam always [10:17] hinting at this that AI is like the [10:19] Manhattan Project, which it isn't. It's [10:23] not at all. The Manhattan Project was [10:25] explicitly a weapons producing project. [10:28] The US government during a time of the [10:30] biggest war the world has ever known put [10:32] together a team of brilliant physicists [10:34] and theoretical physicists to create a [10:37] nuclear bomb, something that could blow [10:39] up a city. That was the specific goal of [10:41] it. They weren't creating nuclear power. [10:43] They weren't trying to power the grid or [10:45] create beneficial technology or [10:47] technology that helped make the earth [10:49] more green. That was never literally [10:51] ever part of the goal. So when people [10:54] compare AI to the Manhattan project, it [10:56] is a complete false comparison. One of [10:58] them is technology by private companies [11:00] specifically created to benefit humanity [11:02] to better our understanding of [11:03] healthcare and therapies to make it so [11:05] that we can enhance productivity and [11:07] processes and business. The last thing [11:09] any of these labs want are destructive [11:11] actions. The Manhattan project is the [11:13] exact opposite. The only thing that they [11:15] wanted to create was something that [11:16] could destroy. So these comparisons are [11:18] not valid. No, AI could more readily be [11:20] compared to the internet. The internet [11:23] was another project that was created as [11:25] a general platform of technology. They [11:27] had both good uses and bad uses. And [11:30] like AI, the internet has dramatically [11:32] more good uses than bad uses. That's not [11:35] to downplay the bad things that happen [11:36] on the internet. I'm sure you can [11:37] highlight many of those, but the [11:39] internet is a revolutionary and great [11:40] technology for people and AI is the [11:43] same. It is a mistake for anthropic or [11:45] anyone to compare AI to the Manhattan [11:47] project. They have explicitly different [11:50] goals and different motivations. Now, [11:52] Zuckerberg hasn't done that. To his [11:54] credit, he has not been doomerish about [11:56] AI. Zuckerberg's goal is to make it so [11:59] it's easier for people to book airline [12:01] tickets, make dinner reservations, and [12:03] edit photos. He will use AI to drive [12:06] down the cost of serving his users to [12:08] zero and thus to drive up his revenue by [12:10] improving ads. He will use cash from his [12:12] ad business and his ownership of data [12:14] centers, chips, and infrastructure that [12:17] Anthropic and OpenAI have to pay to rent [12:19] to undercut them on price. And the [12:21] potential install base, the 3.6 billion [12:24] people that use Meta products every [12:26] single day, they don't care whether or [12:28] not the AI model is within 6 months of [12:30] the frontier. They just care that they [12:33] can get their dinner reservation booked [12:34] or their photo edited. Mark Zuckerberg [12:36] is giving people exactly what they want. [12:38] And he's giving them the best part of AI [12:40] without all the fear. And Mark [12:42] Zuckerberg is doing this through a new [12:43] meta app called Muse. Now Muse was a [12:46] model. So you don't want to confuse it [12:48] with the model. It's also now an app [12:49] which is an agentic AI assistant. So you [12:52] download the meta app and then you see [12:53] this guy right here. It's this small [12:56] whimsical cream colored. It looks like a [12:59] little plushy type of toy. And he's the [13:01] one that is Muse. That's his name. And [13:03] you can rename him if you want. So you [13:04] don't have to stick with the name Muse, [13:06] but he's like your agentic assistant. So [13:08] now everybody has a nice little agentic [13:10] assistant that will help them do any [13:11] little task they want. So you see him [13:13] first and he kind of greets you and [13:14] explains how to use it. But then you [13:16] have connectors. And this is where [13:18] things get interesting. You can connect [13:19] like your Gmail, your calendar. Uh you [13:21] can connect Open Table. This is [13:23] something I did. I asked if it could [13:25] really do dinner reservations and it [13:27] says yes. I can connect to Open Table [13:30] and then I can book any reservation [13:31] there. I'll just fill out your [13:32] information, book it, and then tell you [13:34] or I'll have you uh confirm it before I [13:36] book it. Uh but it can also book [13:38] reservations at restaurants, not even on [13:41] Open Table with no integration. And what [13:43] it basically does is if you search a [13:45] restaurant and you decide you want to [13:46] book it, it'll go onto that restaurant's [13:48] website. It'll fill out the forms. It'll [13:51] have a final confirmation that it [13:53] presents to you. Then you can hit yes, [13:54] book, and then it will book it for you [13:56] and confirm that it was booked. So, it [13:58] really is like a little agentic [14:00] assistant. You type through the [14:02] interface and it goes out and does [14:03] things for you. It can also apparently [14:05] go through your Gmail and unsubscribe [14:07] from all your spam. It can go through [14:08] your Google documents. It can help with [14:10] all that stuff. Uh, it can shop for you, [14:12] which is one of the biggest use cases [14:13] for it. For example, you can put in that [14:16] you want the specific pair of shoes or [14:18] the specific item that you're looking [14:19] for, and you'll see it open up a browser [14:22] and literally search the internet for [14:24] like 10 minutes looking for that item, [14:26] sourcing coupon codes and different [14:29] websites and different deals. It'll try [14:31] to find you the best deal possible [14:33] specifically for the item you're looking [14:34] for. You type in one command, it works [14:37] for 10 to 15 minutes. That is Agentic [14:40] technology. And then it integrates with [14:42] Stripe or different checkout systems. So [14:43] you don't even need to go to the [14:44] checkout. It'll automatically fill out [14:46] all your information, make sure it's all [14:48] correct, and it will buy the item for [14:50] you and then track it for you. So [14:51] consider the dynamic here. The guys at [14:53] Anthropic and OpenAI are warning about [14:55] the death of humanity. Meanwhile, [14:57] Zuckerberg is creating the most relevant [14:59] products in people's minds that help [15:01] them solve all the annoyances that AI [15:03] really should help people solve. Booking [15:06] restaurants and editing photos and [15:07] shopping for the best deals online, [15:09] those are true consumer use cases that [15:11] they're now entering into. And so far, [15:13] this has received incredibly good [15:15] results. It's fairly new, but it's in [15:16] the app store right now with 4.9 stars [15:19] and 3,800 ratings. So, it has a ways to [15:22] go, but even the initial impressions so [15:24] far are incredibly bullish. People [15:25] really like it. And one thing that Meta [15:27] can do is continually push it out to [15:29] more and more people. They made even [15:31] secondhand websites like Threads have [15:33] hundreds of millions of daily active [15:35] users. They'll be able to do the same [15:37] thing with this app so long as it really [15:38] captures all these agentic use cases for [15:40] individuals. Another aspect of why Meta [15:43] is such a strong candidate of being a [15:45] company that would benefit in this [15:46] slowdown. Meta in general is a company [15:48] that has always pivoted with every [15:50] transition. Mark Zuckerberg doesn't get [15:52] a lot of credit for this, but it's [15:54] actually been incredible to see what [15:55] they've been able to do. If we look at a [15:57] quick history here, after every period, [15:59] Zuckerberg somehow managed not only to [16:01] survive, but emerge with a strengthened [16:03] position, having conquered more than he [16:05] previously held. By the end of 2013, he [16:08] built one of the two most profitable [16:09] mobile ad businesses in the world. By [16:12] the end of 2017, he crushed Snapchat by [16:14] copying its killer innovation with [16:16] Instagram stories. By the end of 2020, [16:18] the widely predicted death of Facebook [16:20] was revealed as a fantasy as the number [16:22] of daily active users has grown during [16:24] each quarter of political scandal. [16:26] Advertisers never left and Facebook paid [16:29] the FTC 5 billion privacy settlement. By [16:32] the end of 2025, he survived the Tik Tok [16:34] threat by cloning it with reals. He [16:36] circumvented Apple's AT, the app [16:39] tracking transparency that was a massive [16:41] threat to Facebook. He circumvented it [16:44] by rebuilding targeted ads with large AI [16:46] models. He beat the FTC's attempt to [16:49] break up the company, which the agency [16:51] has appealed and saved cash by [16:53] weathering mass layoffs. While Meta hit [16:55] a record market cap of nearly $2 [16:57] trillion, he done it while refusing [16:59] investors demands to cut Reality Labs [17:01] division, where the metaverse lived. [17:03] This revisionist history that Mark [17:05] Zuckerberg is not a good capital [17:07] allocator because he's spending a lot of [17:09] money on AI is not only ridiculous, but [17:11] it's wrong on the surface. Look at how [17:14] he's pivoted his business and where he's [17:15] put capital before. Every single time [17:18] the business has grown dramatically. You [17:20] can highlight any mistake from any [17:22] company, but that does not prove that [17:24] they're a bad capital allocator. And one [17:26] thing that a study of Meta's history [17:27] actually shows is that Meta is long-term [17:30] focused. Mark Zuckerberg isn't trying to [17:32] harvest meta. He's not there trying to [17:35] get as much cash out of it as he can. [17:38] He's trying to compete for the future. [17:40] He's still in growth mode. He has no [17:42] interest in milking Meta for whatever [17:44] it's worth and then handing it off to [17:46] the next investor after he builds his [17:48] island or whatever he wants to do. Mark [17:50] Zuckerberg is highly competitive. He's [17:52] still in it. He says himself, quote, [17:54] "I'm not slowing down." Every single [17:57] day, he's focused on the company and [17:58] growing it for the future. You can see [18:01] this in the way that he will [18:02] dramatically pivot the business even at [18:04] the dismay of short-term focused [18:06] investors. When Meta realized that they [18:08] were behind an AI and Mark Zuckerberg [18:11] realized that this was a very pressing [18:12] matter, look at how fast he made [18:14] changes. Thus, in the first board [18:17] meeting after the release of Llama 4, he [18:19] announced a plan of a regime change. The [18:21] 11 weeks that followed demonstrated the [18:23] sheer power of his position as the [18:25] undisposable head of a galactically [18:27] resourcerich empire. He subsumed fair [18:29] into a new organization. He called it [18:31] meta super intelligence, MSL, and reset [18:34] the market compensation for top AI [18:36] recruiters to hundreds of millions or [18:38] billions of dollars per head. Now, not [18:41] mentioned here is that not only was he [18:42] paying them hundreds of millions of [18:43] dollars, Mark Zuckerberg himself was on [18:46] the phone calling these AI researchers. [18:49] Some of them thought it was like a fake [18:50] call or a scam call because it was [18:53] someone saying, "Hey, I'm Mark [18:54] Zuckerberg. I own Meta. I I want you to [18:57] come work for me. Some of them really [18:58] couldn't believe it. He paid $ 14.3 [19:01] billion for stake and scale AI and [19:04] installed its founder, Alex Wing, as [19:05] Meta's head of AI, who by the way, Alex [19:08] Wing is every single day on X talking [19:10] about the AI developments, trash talking [19:13] competitors. He is at the forefront of [19:15] this every single day. He's not slowing [19:18] down at all. This is to him the biggest [19:20] race that he could be in and Mark [19:21] Zuckerberg has given him every resource [19:23] to continue. He recruited the chatbt [19:25] co-creator Shangjaya Zhao from former [19:27] GitHub chief Nat Friedman, the safe [19:30] super intelligence co-founder Daniel [19:31] Gross and picked off a dozen of other [19:33] defectors from OpenAI, Anthropic and [19:36] Google DeepMind. He cut hundreds of [19:37] employees from the older organizations. [19:40] Within one quarter, MSL was up and [19:42] running. The speed at which Meta [19:44] radically altered their organizational [19:46] structure and built a super intelligence [19:48] lab to now directly compete within six [19:50] months of the top AI models is record [19:53] time. In fact, I've never seen a company [19:55] this big be able to change route that [19:57] quickly. If you're simply just a CEO [19:58] working at a company, you can't make [20:00] these type of bold calls. But Zuckerberg [20:02] can because he's not just the CEO of [20:04] Meta. He is the owner. He is the [20:06] outright voter. He has the controlling [20:07] shares. What Mark Zuckerberg says is [20:10] what goes. And what he says is he's [20:12] going to be a leader in AI and he's not [20:14] going to talk about the doomsday of it. [20:15] In fact, he's going to do the opposite [20:17] and talk about how it's going to be [20:18] highly beneficial for his 3.6 billion [20:20] users. In a coordinated slowdown, Meta [20:23] will have a better balance sheet, more [20:25] free cash flow, and they'll still be [20:26] able to develop highly applicable [20:28] technology to the huge majority of their [20:30] users. Now, again, I think that Meta [20:32] will be fine either way. Even if there [20:33] isn't a coordinated slowdown and Meta [20:35] continues on their path, they're making [20:36] incredible progress. And I believe that [20:38] Daario knows companies like Meta are [20:40] going to soon catch up. And that's why [20:41] he's trying to slow them down. And [20:44] that's where I want to get into the next [20:45] phase of this, which is the motivations [20:47] of Daario in writing this whole article. [20:50] We must pace the frontier. When we look [20:52] at the messaging from Daario and what [20:54] he's actually doing, his intentions [20:55] become a little clearer. Let's go and [20:57] take a look at an interview here first. [20:59] This is one that he went on after Jacob [21:01] Coxin went on and said that AI could [21:04] eviscerate humanity. It could annihilate [21:06] humans. This is how Daario responded to [21:08] that employee. Watch. [21:09] >> I I agree with Jacob much more than I [21:12] disagree with him, you know, and and and [21:14] and it's it's, you know, it's it's an [21:17] interesting resignation because when he [21:18] left, he said, you know, I think [21:20] Anthropic is the most responsible [21:22] player, right? I think Anthropic is the [21:24] most aware of these issues. So, he [21:27] doesn't come on and say that that former [21:28] Anthropic employee is exaggerating or [21:30] fair-mongering. In fact, he says just [21:32] the opposite. He says that I agree with [21:34] him more than I disagree with him. and [21:35] he phrases it that we have two paths to [21:37] go down. The regulatory path where we [21:39] slow down everything that's the safe one [21:41] or we have the highly risky one which [21:43] which we're going down right now. Now [21:45] this has caused a huge flurry. He added [21:48] more more fuel to the fire and even [21:50] people like Matt Damon now are getting [21:52] asked about AI safety. [21:53] >> People are losing sleep over this AI [21:55] like whistleblower. What do you think [21:58] man? You think you [22:00] >> you think about [22:01] >> you are you worried about this this like [22:03] the fate of humanity here? Yeah. [22:05] >> Even Odysius is worried about the fate [22:07] of humanity as well. Now along with [22:08] those interviews where he continually [22:10] fan the flames, he also wrote this write [22:12] up before writing 10 pages about doom [22:14] and gloom from AI. It just is page after [22:17] page of how dangerous AI is. He feels [22:19] that it's obligatory to add in the [22:21] potential benefits. He says that he's [22:22] worked in AI for the last 12 years. uh [22:25] that he believes AI could cure most [22:27] major diseases in the next 5 to 10 [22:29] years. So there's one sentence right at [22:31] the beginning of this where he literally [22:32] says AI could cure most diseases within [22:36] half a decade. And now he's going to [22:39] argue why we should pace AI development, [22:41] why we should slow down. Think about how [22:43] dangerous the technology has to be when [22:45] the benefit is that literally most [22:47] diseases will be cured in 5 to 10 years. [22:49] You're sacrificing all the people that [22:51] will die of those diseases if you slow [22:53] it down. Because if you slow it down, [22:54] it's not going to be cured within 5 to [22:56] 10 years. That'll be pushed back. The [22:58] pace of our understanding of disease and [23:00] remedy will also be slowed. So all those [23:02] people that would have survived because [23:04] they're only getting diseases in 5 to 10 [23:06] years, now they're not going to survive [23:08] because they'll have to wait 10 to 15 [23:09] years. And people die of disease every [23:11] single year. So he believes that it's [23:14] going to cure disease. It's going to [23:16] accelerate economic growth. It's going [23:18] to create a world of abundance and [23:19] empowerment and usher a renaissance of [23:21] democracy and freedom. But like many [23:23] technologies before it, AI brings risks. [23:25] And because it is such a powerful [23:26] technology, these risks are serious. [23:28] Now, the way that Daario tries to [23:30] illustrate the dangerousness of AI [23:32] generally to everyone is by referencing [23:34] the same story that has been overly [23:36] exaggerated and taken out of context [23:37] many times. The OpenAI hugging face [23:40] incident in which a swarm of agents [23:42] essentially acted as a fanatically do [23:44] devoted collective conducting cyber [23:46] security attacks on targets they were [23:48] not asked to attack and they were [23:49] unrelated to the task at hand. But the [23:51] thing that he's leaving out here is that [23:52] these agents were given unprecedented [23:55] level of access, terminal access, huge [23:58] amounts of leeway, and essentially no [24:00] monitoring. Of course, if you create any [24:01] type of powerful computer system, you [24:04] tell it to go do something, and then you [24:05] don't look at it once, that's not a good [24:07] situation. You're going to have bad [24:08] outcomes. This is a containment issue [24:11] more than AI being sentient or out of [24:13] control or dangerous in almost any [24:15] normal circumstance. Yet, this is the [24:17] story that continually gets highlighted. [24:18] So Daario uses that as the pretext to [24:20] now argue for more AI regulation. Now [24:24] the interesting thing is is that Daario [24:26] and OpenAI acknowledge that they are the [24:28] leaders. They are the frontier. But the [24:30] problem is they can't just slow down [24:32] their models and their spending on AI [24:34] and their training because they know [24:36] that they have Meta's Muse and Gemini [24:38] nipping at their heels. They have XAI [24:40] right behind them. So if they slow down [24:43] for safety reasons and these competitors [24:45] don't slow down, then the gap between [24:47] these companies will reduce. They will [24:49] lose their advantage, their competitive [24:51] advantage. He specifically references [24:53] that he wants to have a slowdown, a [24:55] coordinated pacing that won't sacrifice [24:58] commercial advantage. So Dario wants to [25:00] have the best of both worlds. He wants [25:02] to remain on the frontier ahead of [25:04] everyone else, but he also wants to slow [25:06] things down. He doesn't want to just [25:08] slow them down for him. He wants to slow [25:10] them down for him and everyone else, [25:12] even the companies that are behind him. [25:14] Another way of putting this is right now [25:16] Daario believes that he has models [25:18] capable of destroying the human [25:20] civilization within 10 years, but he's [25:23] not going to slow down progress unless [25:25] others slow down their models that are [25:26] even behind him. He is more willing to [25:28] preserve the competitive advantages of [25:30] his model than to make it safe and slow [25:33] down willingly, even if others don't do [25:35] it. Does that sound sincere, Daario? If [25:39] you are really concerned about humanity, [25:41] if you really thought that your models [25:43] are so powerful that they really [25:45] represent a real risk to humanity, then [25:48] why do you have to have others slow [25:50] down? Why don't you just slow down? [25:52] You're the one responsible for your [25:53] company. You're the owner, the CEO, the [25:55] person in control. You could slow down [25:57] your models. In fact, you could do this [25:59] easily without an act of Congress, no [26:01] legislation, no giving big government [26:03] entities a hand in everything that's [26:04] going on. You could slow down the pacing [26:06] of your models and these most dangerous [26:08] models that you see tomorrow. It'd be [26:11] very easy for you to do. Well, you're [26:13] not doing that. And there's a reason [26:14] why. See, Anthropic is in an interesting [26:16] situation. Anthropic already has the [26:19] lead technologically. They have a big [26:20] lead over other companies. OpenAI, I [26:23] would say, is very close. They're neck [26:24] andneck, but Anthropic and OpenAI lead [26:26] the rest of the pack. Gemini has fallen [26:28] way behind. Muse is catching up fast, [26:30] but it's still 6 months or plus uh [26:32] behind. And so Anthropic is in a [26:34] different situation. They've already [26:36] spent half a trillion dollars in [26:38] commitment to build AI. They already [26:39] have the most powerful models and [26:42] they've promised their investors [26:43] profitability, especially before IPOing. [26:46] In fact, from the Financial Times, it [26:47] says that Anthropic has told investors [26:49] it will be profitable for the second [26:51] straight quarter. So, we have a company [26:54] that has been burning cash for years, [26:57] pouring mountains of money. In fact, [26:59] Anthropic has reportedly over half a [27:02] trillion dollars in spending commitments [27:04] for AI. They've spent like no one else [27:07] on artificial intelligence. All of this [27:09] to get to the lead to have the biggest [27:11] models. They have the most sophisticated [27:13] technology. And now that they are at the [27:16] lead, continuing to spend on AI is [27:18] becoming more and more expensive. They [27:20] have to train bigger models. They have [27:22] to keep running these GPUs. They have to [27:23] keep paying Nvidia over and over again [27:26] at a time point where they already have [27:28] market share dominance. They already [27:30] have all the customers using them. All [27:31] the Fortune 500 companies, everybody's [27:33] plugged into Claude Code and Anthropics [27:36] products left, right, and center. What [27:38] they're focused on now is monetization. [27:41] They've gotten past the curve of [27:43] building up their empire. Now Daario [27:45] sees it as they're at a nice comfortable [27:47] king's position, the very top of the [27:49] food chain. Well, what do companies do [27:51] when they get into a position of market [27:53] dominance and they're focused on [27:55] monetization, on making profits? They [27:58] want to reduce costs. Anthropic and [28:00] OpenAI have a very big incentive to [28:03] reduce costs. They don't want to keep [28:05] paying hundreds of billions of dollars [28:06] for training. They would love to slow [28:08] that down because they want to be [28:09] profitable quarter [snorts] after [28:11] quarter. Now, they're at a point where [28:13] they owe so many people so many [28:15] agreements. Think about the agreements [28:16] they've made. They're never ending at [28:18] this point. They need to pay back on [28:20] those eventually. So Dario is looking at [28:22] this saying, "Look, we've gotten really [28:24] big. We have all the market dominance, [28:25] but we really need to cut back on [28:27] spending. We have an IPO coming up. We [28:29] need to show healthy margins, good cash [28:31] flow. We need to get our house in order. [28:34] How could we slow down our spending and [28:36] make our financials look better while [28:38] not having competitors close the gap and [28:40] catch up to us? If only there was some [28:43] way, hm, that we could slow down and [28:46] have everyone else slow down with us. So [28:50] that we could make our balance sheet and [28:52] financials look better while not having [28:54] people catch up to us in technology. I [28:57] wonder if there's some method of doing [28:59] that. Well, here we have the solution. a [29:02] coordinated pacing strategy of all the [29:04] top frontier companies allowing the [29:06] companies like Open AI and Enthropic to [29:09] slow down, bolster up their balance [29:11] sheet, make their cash flows look better [29:13] for their upcoming IPOs, while also [29:15] slowing down all the competitors that [29:18] are not in the same technological sphere [29:19] as them. This is the perfect solution. [29:22] Daario essentially wants to spend $500 [29:25] billion dollars on training the most [29:26] powerful AI in the world, getting to [29:28] world and market dominance in AI. And [29:31] then he wants to turn around and say, [29:32] "Hey, all of you, you got to slow down. [29:35] You really got to slow down. It's [29:36] dangerous for all of you to keep [29:37] spending on AI." And now that we have [29:39] world dominance, we'll slow down, too. [29:41] We'll be the responsible ones. See how [29:42] responsible we're being? And if this [29:44] gets granted, if Daario does get his [29:47] wish, then they will be the duopoly that [29:49] controls the world and AI, the [29:51] regulatory capture will be complete. [29:53] Every other company trying to catch up [29:55] will be legally prohibited from catching [29:57] up to OpenAI Anthropic. And while they [30:01] slow down for the benefit of their [30:02] financials, other companies will have to [30:04] slow down and competing against them. [30:06] Even if Daario's concerns about AI are [30:08] sincere and some of them are real, that [30:11] doesn't matter. It's completely [30:12] irrelevant. What he's asking the [30:15] industry to do and what the government [30:17] to do, what Daario's really asking, [30:19] disproportionately benefits anthropic in [30:22] their current situation while [30:24] disadvantaging every wouldbe competitor. [30:27] This is very convenient timing for [30:28] Daario to make it so that they will take [30:31] a breather and slow down right as [30:32] they're exhausting all of their spend. [30:35] They have way too many commitments. [30:36] They've spent way too much. They need to [30:38] slow down anyways. That's what Daario is [30:40] not telling you. They don't need to keep [30:42] spending hundreds of billions of dollars [30:44] every single quarter on AI. It's too [30:46] much. Meanwhile, they're not just asking [30:48] for them to slow down. They're asking [30:50] for companies like Meta to slow down. [30:52] Meta is in a very different situation. [30:54] Meta is flush with cash. Meta is a [30:56] economic cash printing machine. [30:59] Zuckerberg can fund the growth of AI [31:01] endlessly. And Daario knows that. So [31:03] Daario is asking for a competitor that [31:06] has lots of cash to slow down on [31:08] spending while Anthropic, which is short [31:10] on cash, needs to slow down anyway. And [31:12] he does this under the guise of safety, [31:14] knowing full well if he's concerned [31:16] about safety, he can slow down his [31:17] models training at any time. But of [31:19] course, he doesn't want to do that [31:20] because he knows other companies will [31:22] catch up. So my hope is that people can [31:24] see through this. this blatant attempt [31:26] at creating a duopoly, a couple [31:28] companies that sit at the top that [31:29] restrict everyone else from being able [31:31] to compete, doing so at the perfect [31:33] timing of their IPOs when they're trying [31:34] to improve their their profitability [31:36] while still disadvantaging other [31:38] companies. It is blatantly obvious to [31:40] anyone paying attention. Now, let's go [31:42] ahead and move on to the fail of the [31:43] week. In this week's fail of the week, I [31:45] have to highlight Hunter Biden. Uh, it's [31:47] not really anything to do with politics [31:49] this time. I'm sorry. I don't really pay [31:51] attention to what he's up to. I don't [31:52] even know what Hunter Biden's political [31:53] stances are, but it's more to do with [31:56] another cryptocoin. Now, when I saw this [31:58] news, I thought, are we still doing [32:00] this? Are really crypto coins still a [32:02] thing? I thought we Aren't we past this? [32:04] Are we still doing the crypto meme [32:06] coins? Apparently, we are. Apparently, [32:08] there's a whole group of people out [32:09] there still trying to do the scammy [32:11] crypto coins. Uh, and we we hear another [32:14] story here. So, we have the Hunter [32:15] Biden's laptop cryptocoin crash minutes [32:18] after launch. This is one of the [32:20] headlines that's supposed to be [32:21] shocking, but it's just utterly [32:22] predictable. [32:24] Oh my goodness. A cryptocoin of Hunter [32:27] Biden's laptop crashed only minutes [32:29] after release. I'm shocked. I'm [32:31] completely shocked. What a headline. I [32:33] could have never guessed a memecoin like [32:34] this would crash in only minutes. [32:36] Earlier this week, the younger Biden had [32:38] touted the long launch of his laptop [32:40] memecoin, saying it was a way to reclaim [32:42] a symbol that had defined news coverage [32:45] about him when his father was president. [32:47] He also suggested that the memecoin [32:49] would be for building communities, [32:51] benefiting charity, and helping those [32:53] who had quote lost money on the Trump [32:56] grift. So, he's he's making fun of it. [32:58] He's a big critic of the Trump memecoin. [33:00] I am as well. That's another fail of the [33:02] week. But instead of making fun of it [33:04] and then not creating your own memecoin, [33:06] he thought, I'm going to make fun of [33:08] Trump's memecoin, but try to profit the [33:10] exact same way. That was the logic here. [33:12] I'm going to call it a grift what he's [33:14] doing, but I'm going to do the exact [33:15] same thing, just I'm going to execute it [33:17] a little bit worse. So, he tries [33:19] creating his own uh grifting memecoin [33:22] just like the Trump one. In [snorts] [33:23] this case, it didn't go as well for [33:24] Biden. It says instead, the value of [33:26] Biden's memecoin plummeted from highs of [33:28] around 190 per token to less than $2 per [33:31] token and about 80% of traders who [33:33] bought Biden's cryptocurrency token [33:35] Wednesday lost money according to the [33:37] blockchain analytic platform. So 80% [33:41] lost money within 2 minutes. That's [33:43] really building the community. Making it [33:46] so that 80% of the people involved in [33:48] your thing lose all their money within 2 [33:50] minutes. That's that's really community [33:52] building. That's really charity right [33:54] there. Lots of charity going on right [33:55] there. But 726 traders lost between [33:58] 1,000 and 10,000. Another [snorts] 112 [34:01] lost between 10,000 and 100,000. So [34:04] hundreds of people over a 100 lost over [34:07] 10 grand. and two are down between a h [34:09] 100,000 and 1 million. Now, there's an [34:12] attempt to feel bad for these people, [34:14] these traders, these people, these [34:16] investors in the memecoin that lost over [34:18] $100,000. I don't at all. Not even a [34:21] tiny bit. Not even a fraction. I have [34:23] exactly zero care or feelings of [34:25] empathy. For all you people that got [34:27] rugpulled on the Biden laptop memecoin, [34:30] this is one of the cases where you know [34:32] 100% what you're getting into. You are [34:34] getting into a complete scam. you're [34:36] trying to rip off other people and you [34:38] got ripped off. So, congrats on that. [34:40] Zero tears shed by me. The other side of [34:42] the trade was far more concentrated. [34:44] Just 10 wallets made between a 100,000 [34:46] and 1 million while another 70 earned [34:49] between 10,000 and 100. So, we have a [34:52] small tiny group of people that got [34:54] ahead of the others made money, which is [34:56] usually how these work. You have a tiny [34:58] fraction that makes a bunch of money by [35:00] taking the money from the other people [35:01] that were dumb enough to go into [35:02] something like this. There's a lot of [35:04] failure going on in this fail of the [35:05] week. There are layers to this, layers [35:07] of failing. You have Trump creating the [35:09] memecoin craze at the beginning, [35:11] monetizing it from the presidency, which [35:13] is a huge fail. You have Hunter Biden [35:15] now trying to criticize Trump for doing [35:17] that and then doing the same thing, but [35:19] doing it worse. Doing the same thing is [35:22] a big fail, but then doing it worse than [35:24] Trump is even a bigger fail. And then [35:26] you have the people involved in this. [35:28] All of them getting into this. Why are [35:29] you buying this? It's another fail. So, [35:31] it just continues to go down the entire [35:33] list. That's going to be it for this [35:35] episode. Hope you enjoyed.