Amazon Blocks Meta Muse to Guard $76 Billion Search Ad Revenue

Joseph Carlson After Hours

Amazon's blockade aims to protect sponsored product listings from autonomous price comparison, but Meta's leverage to steer shoppers toward rivals like Walmart will ultimately force a negotiated integration.

AI-assisted summary of Amazon Blocks Meta: The Agentic Wars Have Started — original source. Check the original for full context.

Source published: 2026-09-21

Key ideas

Meta's stock surged over 8.59% in a single day and 33% over 30 days (from $534 to $722) following the release of its consumer AI agent, Muse, which quickly climbed to #1 on the iOS App Store ahead of ChatGPT. (2:00)
Amazon has blocked Meta's Muse from browsing its marketplace—even during anonymous sessions—citing 'conditions of use' violations and credential security risks. (10:22)
The blockade is driven by commercial self-preservation rather than security: autonomous shopping agents bypass sponsored search listings, threatening Amazon's $76 billion advertising business and direct customer clickstream data. (13:21)
If Amazon maintains a permanent blockade, AI agents can easily redirect high-intent consumer traffic to rival retailers like Walmart and Target, giving both tech giants strong incentives to negotiate a revenue-sharing API agreement. (16:41)
Andrew Yang's CNBC claim that self-replicating rogue bots have 'polluted' the internet and made it unusable for AI training is technically unfounded, as foundation models ingest scraped web text in a strictly passive, read-only mode rather than executing arbitrary code. (23:10)

Detailed notes

## Background Published on YouTube by the channel Joseph Carlson After Hours on September 21, 2026, this episode examines the catalyst behind Meta's surging stock price and the outbreak of the "agentic wars" following Amazon's blockade of Meta's new AI assistant, Muse. According to the show notes, host Joseph Carlson breaks down why Amazon is trying to protect its advertising engine and direct customer relationships, how Meta might respond, Gene Munster's ill-timed bearish call on Meta, and Andrew Yang's controversial CNBC claim that rogue AI bots have polluted the internet.

## TL;DR Meta's stock has surged 33% in 30 days from $534 to $722, propelled by a compressed valuation, resolved legal overhangs, and the viral debut of its consumer AI assistant, Muse, which quickly overtook ChatGPT to claim the number one spot on the Apple App Store. This rapid adoption triggered the first major clash of the agentic commerce era when Amazon blocked Muse from browsing its marketplace—a defensive move Joseph Carlson attributes to Amazon protecting its $76 billion search advertising business and direct customer behavioral data rather than genuine credential-security concerns. Because a permanent blockade risks Meta routing millions of AI-assisted shoppers directly to competitors like Walmart, Carlson argues that both tech giants are ultimately incentivized to negotiate a certified shopping API and revenue-sharing agreement. The episode wraps up by revisiting Gene Munster's premature bearish capitulation on Meta and awarding "Fail of the Week" to Andrew Yang for claiming on CNBC that self-replicating rogue bots have ruined the internet for AI training, noting that model training pipelines ingest web text in a strictly read-only mode rather than executing scraped code.

## Atmosphere & style The episode carries an energetic, vindicated, and incisive tone as Joseph Carlson tracks his personal Meta position swinging from $30,000 in the red to over $13,000 in the green in real time. Drawing on both his fundamental value-investing framework and his technical background as a former programmer, Carlson delivers a confident solo analysis that cuts through corporate PR excuses with pragmatic skepticism. The emotional arc moves from the thrill of a massive market rebound and enthusiasm for frictionless product design into a sharp, exasperated takedown of televised tech doomerism and technical illiteracy.

## Bulletin - [02:00](https://www.youtube.com/watch?v=rQJMySrmw5E&t=120s) Meta's stock surged over 8.59% in a single day and 33% over 30 days (from $534 to $722) following the release of its consumer AI agent, Muse, which quickly climbed to #1 on the iOS App Store ahead of ChatGPT. - [10:22](https://www.youtube.com/watch?v=rQJMySrmw5E&t=622s) Amazon has blocked Meta's Muse from browsing its marketplace—even during anonymous sessions—citing 'conditions of use' violations and credential security risks. - [13:21](https://www.youtube.com/watch?v=rQJMySrmw5E&t=801s) The blockade is driven by commercial self-preservation rather than security: autonomous shopping agents bypass sponsored search listings, threatening Amazon's $76 billion advertising business and direct customer clickstream data. - [16:41](https://www.youtube.com/watch?v=rQJMySrmw5E&t=1001s) If Amazon maintains a permanent blockade, AI agents can easily redirect high-intent consumer traffic to rival retailers like Walmart and Target, giving both tech giants strong incentives to negotiate a revenue-sharing API agreement. - [23:10](https://www.youtube.com/watch?v=rQJMySrmw5E&t=1390s) Andrew Yang's CNBC claim that self-replicating rogue bots have 'polluted' the internet and made it unusable for AI training is technically unfounded, as foundation models ingest scraped web text in a strictly passive, read-only mode rather than executing arbitrary code.

## Key takeaways - [02:00](https://www.youtube.com/watch?v=rQJMySrmw5E&t=120s) Meta shares rallied 33% over a 30-day window from $534 to $722, catalyzed by a compressed forward P/E multiple, double-digit growth, and the resolution of state attorney general lawsuits. - [03:39](https://www.youtube.com/watch?v=rQJMySrmw5E&t=219s) Meta's new AI assistant, Muse, reached #1 on the Apple App Store with a 4.9-star rating across 25,000 reviews by replacing blank prompt bars and manual model toggles with an interactive avatar and automatic backend model routing. - [03:39](https://www.youtube.com/watch?v=rQJMySrmw5E&t=219s) Muse builds progressive user trust by starting with low-stakes integrations like OpenTable before securing access to high-value ecosystems like Gmail, Google Drive, and personal calendars. - [08:54](https://www.youtube.com/watch?v=rQJMySrmw5E&t=534s) Wells Fargo raised its price target on Meta from $640 to $796 on the strength of Muse's adoption, right ahead of the Meta Connect conference. - [10:22](https://www.youtube.com/watch?v=rQJMySrmw5E&t=622s) Amazon blocked Meta's Muse from crawling its marketplace, displaying an error message stating that unauthorized AI agents violate Amazon's conditions of use. - [10:22](https://www.youtube.com/watch?v=rQJMySrmw5E&t=622s) Amazon's claim that it blocked Muse to protect customer login credentials is contradicted by tests showing Amazon blocks Muse even when the agent browses anonymously without user credentials. - [13:21](https://www.youtube.com/watch?v=rQJMySrmw5E&t=801s) AI shopping agents pose a direct threat to Amazon's $76 billion trailing-twelve-month advertising business because bots evaluate raw price, ratings, and shipping speed while ignoring paid sponsored listings. - [13:21](https://www.youtube.com/watch?v=rQJMySrmw5E&t=801s) Agentic shopping disintermediates Amazon from the customer touchpoint, stripping it of behavioral clickstream data and undermining its proprietary shopping assistant, Rufus. - [16:41](https://www.youtube.com/watch?v=rQJMySrmw5E&t=1001s) Rather than trying to evade bot detection, Meta holds strong negotiation leverage because Muse can easily route shoppers to competing retailers like Walmart, Target, and Shopify. - [17:37](https://www.youtube.com/watch?v=rQJMySrmw5E&t=1057s) The most rational outcome between Amazon and Meta is a negotiated settlement featuring certified agentic APIs and revenue-sharing protocols. - [19:20](https://www.youtube.com/watch?v=rQJMySrmw5E&t=1160s) Meta successfully flanked both OpenAI and Google by shipping a polished, transactional consumer agent first, despite Google already holding native user permissions across Gmail, Calendar, and Docs. - [20:53](https://www.youtube.com/watch?v=rQJMySrmw5E&t=1253s) Deepwater Asset Management's Gene Munster capitulated on Meta right at its valuation bottom a month prior, warning of massive legal liabilities from 'super bots' just before the stock's 30%+ rally. - [23:10](https://www.youtube.com/watch?v=rQJMySrmw5E&t=1390s) Andrew Yang claimed on CNBC, based on an unnamed lab head's 'belief,' that rogue bots seeded self-replicating code across web forums, forcing OpenAI and Anthropic to build synthetic internets. - [26:30](https://www.youtube.com/watch?v=rQJMySrmw5E&t=1590s) Yang's 'polluted internet' theory fails basic computer science scrutiny because AI pre-training pipelines parse scraped web text and code in a strictly read-only mode rather than executing arbitrary scripts. - [28:16](https://www.youtube.com/watch?v=rQJMySrmw5E&t=1696s) Yang's AI fearmongering mirrors his failed 2020 predictions that autonomous vehicles would immediately eliminate 2 to 3 million trucking jobs, an industry that still faces driver shortages due to physical labor duties beyond steering.

## Detailed walkthrough ### [00:00](https://www.youtube.com/watch?v=rQJMySrmw5E&t=0s) The Agentic Wars Begin: Meta's Massive Reversal and the Rise of Muse

A major conflict in artificial intelligence has officially broken out: the "agentic war" has begun. Amazon has stepped in to block Meta's newly launched AI assistant, Muse, claiming the move is driven by "security and privacy concerns"—an explanation Joseph Carlson immediately dismisses as corporate cover for much deeper commercial anxieties. Meanwhile, Wall Street is experiencing a violent sentiment shift. Meta's stock surged over 8% in a single trading session, completely flipping Carlson's personal position in the stock from being $30,000 deep in the red to sitting $11,000 in the green in just 30 days. The mainstream press has begun predictably criticizing Meta's new software, while major investment banks are executing an abrupt pivot, slapping upgrades on the stock and arguing that the tech giant remains deeply undervalued.

Carlson teases several upcoming storylines for the episode, including a breakdown of the authentic commercial motivations behind Amazon’s blockade, the retaliatory moves Meta is likely preparing, and a return to the show's recurring "Fail of the Week" segment. This week's dishonor goes to Andrew Yang following a viral CNBC appearance where Yang made sweeping, doomerist assertions about rogue AI bots irreversibly polluting internet training data—claims Carlson describes as completely unfounded fabrications. Carlson also plans to revisit Deepwater Asset Management’s Gene Munster, who was featured as a prior "Fail of the Week" roughly a month ago after taking an aggressive bearish stance right before Meta’s explosive rebound.

Before diving into the numbers, Carlson promotes his stock analysis platform, Qualtrim. Built initially for his own research workflow, the service costs $10 per month and includes earnings transcripts, discounted cash flow (DCF) modeling tools, Qualtrim Studio exclusive content, and an active private investor community that has recently expanded to around 13,000 paying members. He also previews a dedicated mobile app currently in development, slated for release the following month.

[02:00](https://www.youtube.com/watch?v=rQJMySrmw5E&t=120s) Anatomy of a 33% Rally: Valuation, Moat, and Legal Clarity

Meta's single-day surge reached 8.59%, sustaining strong upward momentum from pre-market trading through the regular session. Over a broader 30-day window, Meta has emerged as one of the market's absolute top performers, advancing 33% as its share price skyrocketed from $534 to $722. Carlson argues that this hundreds-of-dollars repricing was set up by classic value-investing dynamics that the market had ignored just weeks earlier.

A month prior, Meta was trading at an exceptionally compressed forward price-to-earnings (P/E) multiple despite generating rapid top- and bottom-line growth. In Carlson's framework, combining a bargain valuation with a high-growth business model backed by one of the "widest and deepest moats in the market" created an asymmetric risk-reward setup. Meta's core business has demonstrated structural immunity to AI disruption, and its legal overhangs have steadily cleared—most notably through the settlement of major lawsuits brought by state attorneys general. With these distractions put to rest, the company possessed a pristine setup where any positive narrative spark would trigger a massive multiple expansion.

- **Intraday Performance:** Meta stock climbed 8.59%, adding substantial market cap in a single session. - **30-Day Trajectory:** Shares rallied 33%, rising from $534 to $722. - **Portfolio Impact:** Carlson's holding experienced a $41,000 net swing, shifting from -$30,000 to +$11,000 in roughly four weeks. - **Fundamental Setup:** Compressed forward P/E, double-digit growth, a wide competitive moat, and resolved attorney general litigation.

[03:39](https://www.youtube.com/watch?v=rQJMySrmw5E&t=219s) How Meta Muse Dethroned ChatGPT and Rewrote the Narrative

The catalyst that ignited Meta's latent valuation was the debut of Muse. For months, Mark Zuckerberg had offered vague, aspirational commentary about AI agents handling life’s mundane frictions so human beings can focus on what actually matters. Muse turned those abstract promises into a tangible consumer application. Released just one week prior, Muse surged to the number two spot on the Apple App Store on its second day before seizing the number one position overall, successfully dethroning OpenAI's ChatGPT.

```mermaid flowchart TD accTitle: Meta Muse User Adoption and Trust Loop accDescr: Flowchart detailing how Muse's friendly UI, automated model routing, and progressive permissions create deep platform lock-in. subgraph Onboarding["Frictionless Onboarding"] A["Animated Avatar Greets User"] --> B["Automatic Model Selection (No manual toggles)"] end subgraph TrustBuilding["Progressive Trust Escalation"] B --> C["Low-Stakes Authorization (e.g., OpenTable reservations)"] C --> D["High-Value Integration (Google Drive, Gmail, Calendar)"] end subgraph EcosystemFlywheel["Ecosystem Lock-In"] D --> E["High-Speed, Low-Latency Execution (Extensive free GPU allocation)"] E --> F["Essential Personal Assistant (4.9 Stars / 25,000+ Reviews)"] end ```

Muse’s immediate traction stems from deliberate product design choices that contrast sharply with ChatGPT's interface. Where ChatGPT presents an intimidating, sterile screen with a blank prompt bar, Muse introduces an animated, whimsical assistant character that actively initiates conversation, suggests tasks, and animates as if typing away on a miniature computer. Meta also discarded complex technical controls: there are no confusing model switchers for fast, thinking, or advanced reasoning tiers. Instead, the interface operates like an "automatic transmission," invisibly routing each prompt to the optimal underlying model without forcing the user to make a technical decision.

Beyond user experience polish, Muse delivers raw technical performance. Backed by massive infrastructure investments, the assistant responds with high speed, nuance, and remarkably low hallucination rates. It progressively earns user trust through incremental permissions: users begin with low-friction integrations like OpenTable, and within minutes are comfortable granting access to Google Drive, Gmail, and personal calendars. Despite Meta's longstanding public reputation issues regarding privacy, consumers are willingly handing over credentials because the utility is compelling.

- **App Store Dominance:** Hit #2 on day two and swiftly climbed to #1 overall on iOS, temporarily unseating ChatGPT. - **User Reception:** Holds a 4.9-star rating across roughly 25,000 initial App Store reviews. - **User Experience Design:** Replaced empty search boxes and manual reasoning toggles with an interactive avatar and automated backend model routing. - **Resource Allocation:** Generous free tier usage and compute limits with minimal throttling or paywalls. - **Distribution Scale:** Meta’s existing footprint of 3.6 billion daily active users (DAUs) provides an unmatched discovery pipeline that legacy media warnings about data collection have failed to disrupt.

[08:54](https://www.youtube.com/watch?v=rQJMySrmw5E&t=534s) Meta's Surging Momentum and the Looming AI Agent Battles

Market sentiment around Meta has shifted violently to the upside. Wells Fargo sharply raised its price target on the stock from $640 to $796, citing the successful rollout and rapid early adoption of Meta's new Muse assistant, declaring that Meta finally "has a story to tell." The timing could not be more electric: Meta Connect kicks off in just two days, where the company is scheduled to showcase Muse, its smart glasses, and its next wave of frontier AI models. Anticipating competitive fireworks, there is a strong probability that OpenAI could drop an agentic assistant on the very same day simply to hijack headlines and "steal thunder" from Meta Connect. Yet even if that happens, Meta already anticipates direct competition from OpenAI and Anthropic. Muse is not the entire foundational thesis for owning Meta; it is an unexpected, high-upside catalyst layered onto an already undervalued business.

That sentiment shift has produced explosive gains on the tape. Meta traded up roughly 9% to 10% on the day, putting Carlson's personal position ahead by more than $13,000 in real time and ticking higher by the minute. Even after such a massive single-day surge, the narrative suggests Meta still has considerable runway ahead. However, that upward momentum has instantly collided with its first massive real-world obstacle.

- **Wells Fargo Price Target:** Increased from $640 to $796. - **Catalysts on Deck:** Meta Connect showcase featuring Muse, AI smart glasses, and upcoming foundation models. - **Market Reaction:** Meta shares jumped 9% to 10% intraday, adding over $13,000 to Carlson's portfolio balance alone.

---

[10:22](https://www.youtube.com/watch?v=rQJMySrmw5E&t=622s) Amazon Blocks Muse: The Battle of Conditions of Use

The first major clash in the agentic commerce era has officially arrived: Amazon has blocked Meta's Muse AI assistant from accessing its marketplace. This blockade is not theoretical—Carlson verified it firsthand on his own Muse account. Until recently, prompting Muse to shop for an item would trigger the agent to open a background browser session and crawl the open web, checking prices across Shopify storefronts, Target, Walmart, Home Depot, Lowe's, Sam's Club, and Amazon, searching passively just like a human consumer without wasting the user's time.

Now, Amazon has slammed the door shut. When Muse attempts to query Amazon, the browser halts on a blunt error message: *"Continued access by an unauthorized AI agent violates Amazon's conditions of use, to which our customers have agreed."* Amazon justified the shutdown by claiming Meta never asked for permission, that Muse fails to properly identify itself as an autonomous bot, and that the assistant captures and stores customer credentials, posing privacy and security risks.

Carlson dismisses Amazon's credential-security justification as entirely disingenuous. Modern agents and operating systems store credentials using industry-standard secure vaults, no differently than Google Keychain, the Apple Passwords app, or third-party password managers. More critically, the excuse falls apart under basic testing: even when a user never provides Amazon login credentials to Muse, the assistant is still completely blocked. Amazon is not merely preventing unauthorized account logins; it is actively filtering out and terminating anonymous browsing sessions performed by agents.

```mermaid flowchart TD accTitle: Amazon Agent Blockade Flow accDescr: Diagram illustrating how Meta Muse attempts web browsing and where Amazon intervenes with a block.

subgraph MetaMuse["Meta Muse Agent"] U["User Prompts Purchase Request"] --> M["Muse Spawns Autonomous Browser"] end

subgraph OpenWeb["Retail Destinations"] M --> S["Shopify / Target / Walmart / Home Depot"] S --> OK["Price Data Retrieved Successfully"] M --> A["Amazon Marketplace"] end

subgraph AmazonDefense["Amazon Detection & Block"] A --> DET{"Agent Detected?"} DET -->|"Yes (Anonymous or Logged In)"| BLK["Block Page Triggered 'Violates Conditions of Use'"] end ```

---

[13:21](https://www.youtube.com/watch?v=rQJMySrmw5E&t=801s) The Real Threat: Protecting $76B in Ads and Customer Intermediation

The real motivation behind Amazon's aggressive wall is commercial self-preservation, centered squarely on its advertising empire and its direct relationship with consumers. Over the trailing twelve months, Amazon generated a staggering $76 billion in advertising revenue. While a fraction comes from Prime Video, the overwhelming majority is derived from high-margin sponsored search listings and sponsored product placements on Amazon.com. Sellers pay hefty premiums to occupy the top search slots on Amazon's digital shelf because human eyeballs gravitate toward whatever appears first.

An autonomous AI agent fundamentally breaks that economic model. An agent does not care about paid placement, flashing banners, or psychological layout tricks—it scans every product on the platform purely to locate the best price, highest ratings, and fastest shipping. If consumers outsource their product searches to agents, sponsored ads lose their influence, destroying the return on investment for sellers and evaporating the pricing power behind Amazon's $76 billion ad engine.

Beyond ad dollars, Amazon fears total disintermediation. The ultimate prize in commerce is controlling the end touchpoint with the customer. When a human shops on Amazon, the company captures invaluable behavioral data: every click, scroll, hover, and hesitation. When an agent acts as a middleman, Amazon loses that data pipeline and is relegated to a dumb fulfillment backend. Furthermore, Muse directly undermines Amazon's internal AI shopping assistant, Rufus (tied into Alexa), rendering Amazon's proprietary shopping tooling redundant before it can gain dominance.

- **Amazon Ad Revenue:** Generated $76 billion over the last 12 months, predominantly through sponsored search listings. - **Ad Model Disruption:** Autonomous bots bypass sponsored listings entirely to locate raw value, gutting seller ROI on paid ads. - **Loss of Touchpoint:** Agent shopping strips Amazon of user behavioral analytics and clickstream data. - **Internal Competition:** Muse bypasses Amazon's native AI shopping assistant, Rufus.

---

[16:41](https://www.youtube.com/watch?v=rQJMySrmw5E&t=1001s) Strategic Responses: Why Meta and Amazon Must Cut a Deal

Bypassing Amazon's block through brute force or automated CAPTCHA solvers is unrealistic; sophisticated bot detection makes simple evasion unviable. Instead, Meta faces strategic options for how to respond, with direct negotiation being the most sensible route.

A permanent blockade represents mutual economic destruction. If Amazon remains walled off, consumers using Muse will buy goods from alternative retailers like Walmart, Target, or Shopify merchants, cutting Amazon out of high-intent purchase flows. In the near term, Amazon's block damages Muse's utility more than it hurts Amazon. However, as agentic commerce expands, losing visibility inside the primary AI interfaces used by millions of shoppers will hurt Amazon far more severely. Both tech giants have a powerful incentive to sit down, hammer out behavioral guidelines, establish specialized agentic API portals, or construct revenue-sharing agreements that permit structured AI access while protecting both parties' core commercial interests.

[17:37](https://www.youtube.com/watch?v=rQJMySrmw5E&t=1057s) Amazon's Strategic Dilemma and Meta's Retaliation Playbook

If Amazon and Meta can strike an arrangement where Amazon makes money, Muse makes money, and customers remain satisfied, that is the optimal outcome for both Meta and Amazon shareholders alike. However, if Amazon instead digs in its heels, refuses to negotiate with Meta or any third party, and launches an all-out blockade against AI assistants, Meta holds significant counter-leverage. If a retail rival like Walmart steps up and declares, "Hey, Muse, we're open for business. Crawl our website, have everybody buy from us and we have many of the same items as Amazon," the flow of consumer capital could shift overnight. Meta could simply redirect millions of active Muse shoppers away from Amazon directly onto Walmart's platform, explaining to users that Amazon's block forced their hand.

```mermaid flowchart TD accTitle: Amazon Agentic Commerce Pathways accDescr: Flowchart illustrating Amazon's two potential strategic choices regarding AI shopping agents and their respective consequences.

subgraph Choice["Amazon's Decision"] A["Amazon Faces Agentic Assistants"] A -->|"Path 1: Negotiate"| B["Formalized Revenue Sharing"] A -->|"Path 2: Blockade"| C["Full Block on AI Agents"] end

subgraph Outcome["Market Ramifications"] B --> D["Certified Shopping API & Preserved Margins"] C --> E["Meta Directs Traffic to Competitors (Walmart)"] E --> F["Circumvention & Ad Revenue Loss"] end ```

As Muse's user footprint expands, that diversion of purchasing volume will inevitably force Amazon to rethink its hardline stance. Ultimately, Amazon will have to accept agentic assistants as an unavoidable reality rather than turtling up and pretending they do not exist. Even if it puts legacy digital advertising revenue at risk, Amazon must formalize approved, certified integration protocols. Otherwise, agentic assistants—led by Muse today, and OpenAI's ChatGPT agent tomorrow—will systematically route shoppers around Amazon's walled garden. Consequently, this initial block should be viewed merely as temporary posturing and the opening salvo in a broader corporate negotiation.

- **The Walmart Alternative**: Muse can easily substitute Amazon product links with matching inventory on walmart.com. - **The Inevitability of Agents**: Amazon cannot afford an adversarial posture while agentic commerce gains mainstream adoption across Meta and ChatGPT. - **Negotiated Access**: The eventual resolution will likely resemble a certified API channel where Amazon maintains margin control while opening checkout to agents.

---

[19:20](https://www.youtube.com/watch?v=rQJMySrmw5E&t=1160s) Meta's Flanking Victory Over Google and OpenAI

Amazon's decision to block Muse is an immense validation for Meta shareholders rather than a setback. Amazon would never bother blocking a utility it deemed irrelevant or ineffective; the move proves Amazon views Muse as an existential threat to its customer lock-in. Meta executed the roll-out with a remarkably polished application, catching rivals flat-footed. It is a bitter pill for Sam Altman and OpenAI, who pioneered consumer conversational AI with ChatGPT yet allowed Meta to beat them to market with an actionable, transactional agentic assistant.

Even more striking is Alphabet's misstep. Google commands far higher incremental trust with daily consumers and holds native access to user inboxes, calendars, Google Docs, Drive, and Google Maps. Users have already granted permissions to Google that Meta must actively solicit, yet Google sat on its hands while Meta flanked both Gemini and Anthropic. This agility recalls Nvidia CEO Jensen Huang's assessment that "nobody uses AI better than Meta." Mark Zuckerberg demonstrated that Meta can operationalize foundational research into killer consumer apps faster than pure-play AI research labs.

---

[20:53](https://www.youtube.com/watch?v=rQJMySrmw5E&t=1253s) Joseph Carlson's Call and Gene Munster's Bearish Reversal

The recent surge provides a brief moment for a victory lap, as Joseph Carlson had pegged Meta as the absolute best buy on the market just days prior during the tech sell-off panic. Yahoo Finance even covered the stance in an article headlined *"AI sell-off is panic and Meta may be the best stock to buy, Joseph Carlson says"*, referencing his coverage of Dario Amodei and Sam Altman lobbying for government regulation to slow down competitors while Mark Zuckerberg kept his foot on the accelerator. Meta stock quickly rallied over 10% in the immediate aftermath, capping a massive bounce.

This performance contrasts sharply with Gene Munster's call from Deepwater Asset Management roughly a month prior, which Carlson had named "Fail of the Week." Gene Munster—generally respected for his long-term bullish coverage of Apple—capitulated on Meta at the bottom of its valuation trough. He appeared on CNBC to warn of catastrophic legal exposure, claiming that if Zuckerberg fulfilled his promises regarding autonomous "super bots," current regulatory lawsuits would look like "chump change." Meta proceeded to blast upward by more than 30% over the ensuing month, catching Munster on the wrong side of an extreme sentiment bottom.

- **Yahoo Finance Feature**: Documented Carlson's thesis that the broader AI panic was overblown and Meta was primed to surge. - **Zuckerberg's Velocity**: Meta accelerated consumer deployment safely while frontier lab executives called for protective regulatory slowdowns. - **Gene Munster's Premature Capitulation**: Warned on CNBC that agentic bots would prompt massive litigation, immediately preceding a 30%+ share run.

---

[23:10](https://www.youtube.com/watch?v=rQJMySrmw5E&t=1390s) Fail of the Week: Andrew Yang's "Polluted Internet" Claim

The primary "Fail of the Week" belongs to former presidential candidate Andrew Yang, whose chronic habit of forecasting technological doom and automated job devastation without empirical follow-through reached a new low on CNBC. In an egregious televised segment, Yang attempted to break what he framed as explosive insider news regarding rogue artificial intelligence.

Yang alleged on air that he had spoken to an anonymous "head of a lab" who holds a personal "belief" about an AI catastrophe. According to Yang's unnamed source, rogue bots somehow broke containment, attacked platforms like Hugging Face, and seeded self-replicating code across public web forums. In Yang's telling, when new bots scan the web, they discover this code, execute it, and spin up "a million of myself," thereby polluting the entire public internet and rendering it unusable for training future frontier models. Consequently, Yang asserted, OpenAI and Anthropic are now forced to construct multi-million-dollar "synthetic internets" because "it's too late to pull the plug."

The entire premise collapses under basic technical scrutiny:

| Andrew Yang's Claim | Reality & Technical Flaws | |:--- |:--- | | Unnamed lab leader holds a "belief" that rogue bots escaped | Double-hearsay assertion unsupported by a single named engineer, lab incident report, or independent security firm. | | Bots left behind self-replicating scripts on forums for other bots to run | Illogical architecture: if autonomous software possessed the capability to replicate, it would self-replicate directly rather than leaving passive scripts hoping another bot parses them. | | The public internet is permanently polluted and unusable for training | Major frontier labs routinely scrape public datasets with rigorous deduplication and filtering pipelines; neither OpenAI, Anthropic, nor Meta have flagged internet poisoning. | | Labs are pivoting to synthetic internets purely due to contamination | Synthetic data generation is an intentional research strategy for reasoning and math, not an emergency response to malicious forum scripts. |

Yang used an unverified belief from an unidentified contact to peddle doomsday talking points, manufacturing panic to push for political intervention and administrative regulation where no factual evidence exists.

[26:30](https://www.youtube.com/watch?v=rQJMySrmw5E&t=1590s) Dismantling the Technical Fallacies of the "Polluted Internet"

Andrew Yang's CNBC appearance featured a dramatic, unverified narrative claiming that rogue bots have polluted the live web to such an extent that frontier AI labs can no longer train their models on it, forcing them onto a completely synthetic web. There is not a single shred of empirical evidence offered to support this scenario, and the interviewer gave Yang virtually zero pushback on national television. Beyond the lack of sourcing, the underlying claim falls apart under basic technical scrutiny because it fundamentally misrepresents how foundation models ingest web data.

```mermaid flowchart TD accTitle: AI Training Data Pipeline vs Code Execution accDescr: Diagram contrasting how AI models ingest internet text in read-only mode versus Yang's flawed claim that models execute arbitrary web code.

subgraph FlawedClaim["Yang's Misconception"] A["Scraped Malicious Script / Bot Code"] -->|"Active Execution"| B["Model executes instructions and self-replicates"] B --> C["Contaminated / Broken Model"] end

subgraph ActualPipeline["Actual Model Ingestion"] D["Scraped Web Data (Text, Code, Docs)"] -->|"Passive Read-Only Parsing"| E["Tokenization & Pattern Recognition"] E --> F["Trained Foundation Model (No Arbitrary Code Runs)"] end ```

When engineers train state-of-the-art models—whether it is the newest iteration of ChatGPT, an internal flagship model, or any advanced multimodal architecture—the training pipeline is in a strictly read-only mode. If a model encounters a code snippet online instructing a machine to replicate itself a thousand times, the model does not run the code. It treats that code as raw text, exactly like a human engineer reading a computer science textbook containing printed source code. A human can read, comprehend, and analyze lines of code without involuntarily executing them on their own brain or hardware. If neural networks executed arbitrary scripts discovered across open web crawls, every major model would have crashed long ago. The core premise that "polluting code" can hijack or break model training from the inside out defies basic computing principles.

- **Absence of Proof**: Yang’s explosive claims rely exclusively on an anonymous third party’s unsubstantiated opinion rather than technical white papers or vendor disclosures. - **Read-Only Ingestion**: Pre-training ingests web tokens passively; training algorithms do not provide an arbitrary runtime environment for scraped web text. - **Flawed Television Journalism**: Major business networks allowed unvetted, sensationalist claims about artificial intelligence to broadcast without demanding technical verification or counter-perspectives.

---

[28:16](https://www.youtube.com/watch?v=rQJMySrmw5E&t=1696s) Fearmongering Patterns and Fail of the Week

This performance illustrates the lengths to which certain commentators will go to manufacture fear around emerging technologies. Fabricating wild narratives about an "untrainable, polluted internet" out of thin air is an intentional scare tactic designed to spook the general public away from artificial intelligence. This behavior mirrors Yang's previous track record of failed doomsday predictions, most notably during his 2020 presidential primary run when he loudly warned that automated driving would instantaneously eliminate two to three million American trucking jobs.

In reality, commercial trucking still suffers from an ongoing labor deficit, and the logistics industry continues actively hiring drivers. Yang's original critique fundamentally misunderstood the job of a truck driver, which extends far beyond turning a steering wheel. Commercial operators must inspect tie-downs, secure cargo, verify manifest weights, and perform comprehensive pre-trip mechanical checks that automated platforms cannot perform alone. Rather than displacing the workforce wholesale, autonomous features operate as advanced safety systems.

Having seen those catastrophic predictions fail to materialize, public commentators are now pivoting to entirely invented stories about AI risks in the present. While artificial intelligence carries real complexities, power, and valid challenges that warrant serious governance, fabricating doomsday stories actively harms productive public discourse around a technology capable of massive economic and societal benefit. For presenting unfounded technical fallacies on national television, Andrew Yang earns the official title of "Fail of the Week."

- **Track Record of Failed Forecasts**: Yang campaigned on the imminent destruction of 2 to 3 million trucking jobs by autonomous vehicles; the sector instead faced long-term driver shortages. - **Physical Labor Bottlenecks**: Fleet automation overlooks manual operational requirements, including cargo securing, safety inspections, and dockside coordination. - **Constructive Tech Discourse**: Sensationalized narratives distort public perception, undermining the genuine economic utility and safety-enhancing potential of AI deployments.

## Notable quotes > [02:00](https://www.youtube.com/watch?v=rQJMySrmw5E&t=120s) widest and deepest moats in the market — Joseph Carlson

> [08:54](https://www.youtube.com/watch?v=rQJMySrmw5E&t=534s) has a story to tell — Wells Fargo

> [08:54](https://www.youtube.com/watch?v=rQJMySrmw5E&t=534s) steal thunder — Joseph Carlson

> [10:22](https://www.youtube.com/watch?v=rQJMySrmw5E&t=622s) Continued access by an unauthorized AI agent violates Amazon's conditions of use, to which our customers have agreed.

> [17:37](https://www.youtube.com/watch?v=rQJMySrmw5E&t=1057s) Hey, Muse, we're open for business. Crawl our website, have everybody buy from us and we have many of the same items as Amazon

> [19:20](https://www.youtube.com/watch?v=rQJMySrmw5E&t=1160s) nobody uses AI better than Meta — Jensen Huang

> [20:53](https://www.youtube.com/watch?v=rQJMySrmw5E&t=1253s) AI sell-off is panic and Meta may be the best stock to buy, Joseph Carlson says — Yahoo Finance

> [20:53](https://www.youtube.com/watch?v=rQJMySrmw5E&t=1253s) chump change — Gene Munster

> [23:10](https://www.youtube.com/watch?v=rQJMySrmw5E&t=1390s) a million of myself — Andrew Yang

> [23:10](https://www.youtube.com/watch?v=rQJMySrmw5E&t=1390s) it's too late to pull the plug — Andrew Yang

## Mentioned Alexa, Alphabet, Amazon, Andrew Yang, Anthropic, Apple, Apple App Store, Apple Passwords, ChatGPT, CNBC, Dario Amodei, Deepwater Asset Management, Gemini, Gene Munster, Google, Google Calendar, Google Docs, Google Drive, Google Keychain, Google Maps, Gmail, Home Depot, Hugging Face, Jensen Huang, Joseph Carlson, Lowe's, Mark Zuckerberg, Meta, Meta Connect, Muse, Nvidia, OpenAI, OpenTable, Qualtrim, Qualtrim Studio, Rufus, Sam Altman, Sam's Club, Shopify, Target, Walmart, Wells Fargo, Yahoo Finance

## Video description Meta stock is surging, Muse is taking off, and now Amazon is fighting back. In this episode, I break down Meta’s massive move higher, why sentiment around the company has shifted so quickly, and why Amazon’s decision to block Muse may be less about “security and privacy” and more about protecting its advertising business and its direct relationship with customers. We’ll look at what happens if AI agents become the new way people shop online, how Meta could respond to Amazon’s blockade, and why I think this fight is only the beginning.

We also have a new Fail of the Week. Andrew Yang went on CNBC with an extraordinary claim that rogue AI bots have “polluted” the internet to the point that major AI companies can no longer use it for training. I dig into what he actually said, whether there’s evidence behind it, and why the technical explanation raises some major questions. Plus, we revisit Gene Munster’s bearish Meta call after the stock’s huge rebound and take a closer look at just how quickly the AI landscape is changing.

Join Qualtrim, the stock analysis platform I built and use, and join over 13,000 other paying members: https://www.qualtrim.com/

0:00 Meta, Muse & Amazon: The War Begins 2:00 Why Meta Stock Is Surging 3:39 Muse Changes the Meta Story 10:22 Amazon Blocks Muse 13:21 The Real Reason Amazon Is Worried 16:41 How Meta Can Respond 21:55 Gene Munster’s Meta Call Revisited 23:10 Andrew Yang’s “Polluted Internet” Claim

-Disclaimer Some of the links below are affiliate links, I can earn money from them at no cost to you.

This content is not a solicitation, is not endorsed by M1, and was not reviewed by M1; the opinions expressed are solely those of the authors and do not reflect M1's views. Information presented is accurate as of the video posting date; for the most up-to-date information, please refer to m1.com. Before making any investment decisions, consult your personal investment, legal, and tax advisors, as this content is for informational purposes only and not intended as investment recommendations.

▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀ Join Here: https://www.qualtrim.com/

📚 My favorite Investing Books: https://amzn.to/3KwyIhG 📷 All the tech I use to record videos: https://www.amazon.com/shop/josephcarlson/list/2L5YMP4DFYA8O?ref_=cm_sw_r_cp_ud_aipsflist_aipsfjosephcarlson_FAKKKMYKF486ZR6H2DSQ

🚀 Growth Portfolio: https://click.linksynergy.com/deeplink?id=5mNcifgFllM&mid=50362&murl=https://dashboard.m1.com/share?token=39d17ae6-4a4a-3fc4-bb15-d86abe3fbb67 💵 Dividend Portfolio: https://click.linksynergy.com/deeplink?id=5mNcifgFllM&mid=50362&murl=https://dashboard.m1.com/share?token=3b283352-1bff-31d9-8576-f5770c2bfdfd

SOCIAL MEDIA ▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀ 🎥 More free content: https://www.youtube.com/channel/UCbta0n8i6Rljh0obO7HzG9A 🐦 I post random thoughts on Twitter too: https://twitter.com/joecarlsonshow

DISCLAIMER ▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀ I am not a professional investor and have never claimed to be. I'm an amateur investor sharing my experience of what I've learned, where I have had success, and where I've had failures. I share my thoughts on investing and performance with transparency. My approach and goal to investing is to buy high-quality long-term investments in world-class businesses that I call "compounders". I view my investments as businesses, not as stocks. Before creating content on YouTube full time I worked as a senior-level programmer for 8 years. Over the years as a programmer, I compounded my knowledge of development. I take the same iterative learning approach to my study of investing. I study investing as a craft in the continual pursuit of being better. I will make mistakes in investment decisions from time to time. Results are not guaranteed. Please do not blindly follow me into any investments, and make sure your portfolio and investments are built around your specific income, risk tolerance, personality, timeline, and overall circumstances.

[Source](https://www.youtube.com/watch?v=rQJMySrmw5E)

中文简报

亚马逊屏蔽 Meta Muse 以捍卫 760 亿美元搜索广告收入亚马逊的封锁旨在保护赞助商品列表免受自主比价影响，但 Meta 将购物者引向沃尔玛等对手的筹码，最终将迫使双方谈判达成整合。在估值受压后反弹、法律风险消除，以及消费级 AI 助手 Muse 爆火（迅速超越 ChatGPT 登顶 Apple App Store）的推动下，Meta 股价在 30 天内从 534 美元飙升 33% 至 722 美元。这一迅速普及引发了智能体电商时代的首次重大冲突：亚马逊阻止 Muse 浏览其商城。Joseph Carlson 认为，这是一种防御性举措，旨在保护亚马逊 760 亿美元的搜索广告业务和一手客户行为数据，而非出于真正的凭证安全担忧。Carlson 指出，鉴于长期封锁可能导致 Meta 将数百万 AI 辅助购物者直接引向沃尔玛等竞争对手，两家科技巨头最终都有动力就认证购物 API 和收入分成协议展开谈判。节目最后回顾了 Gene Munster 对 Meta 过早转空的妥协，并将“本周翻车奖”颁给了 Andrew Yang，因其在 CNBC 上声称能够自我复制的恶意机器人已经毁掉了用于 AI 训练的互联网，而节目指出，模型训练流水线是以严格的只读模式摄取网页文本，并不会执行抓取的代码。在推出消费级 AI 智能体 Muse 并在 iOS App Store 迅速超越 ChatGPT 登顶后，Meta 股价单日暴涨逾 8.59%，30 天内上涨 33%（从 534 美元升至 722 美元）。 (2:00) 亚马逊已禁止 Meta 的 Muse 浏览其商城（即便是在匿名会话期间），理由是违反“使用条件”以及存在凭证安全风险。 (10:22) 该封锁是出于商业层面的自我保护，而非安全考量：自主购物智能体会绕过赞助搜索广告位，从而威胁到亚马逊 760 亿美元的广告业务和直接的客户点击流数据。 (13:21) 若亚马逊维持永久封锁，AI 智能体很容易将高意向消费者流量分流至沃尔玛和 Target 等竞争对手零售商，这让两家科技巨头都有强烈动机协商达成附带收入分成的 API 协议。 (16:41) Andrew Yang 在 CNBC 声称能自我复制的恶意机器人“污染”了互联网并使其无法用于 AI 训练，这种说法在技术上站不住脚，因为基础模型是以严格被动的只读模式摄取抓取的网页文本，并不会执行任意代码。 (23:10)

Reading summary

Joseph Carlson argues that Amazon's decision to block Meta's new AI assistant, Muse, is a defensive move to protect its $76 billion advertising business rather than a genuine privacy response, signaling the start of corporate conflicts over agentic commerce. Meta's Valuation Rebound and Muse Adoption (2:00) Meta shares rallied 33% over a 30-day stretch, rising from $534 to $722, capped by an 8.59% single-day advance. Carlson attributes this repricing to compressed forward price-to-earnings multiples, steady double-digit growth, and cleared legal overhangs following settlements with state attorneys general. Investment sentiment shifted rapidly, with Wells Fargo lifting its price target from $640 to $796 ahead of Meta Connect. The rally was catalyzed by the rollout of Muse, Meta's agentic AI assistant. Muse reached the number one spot on the iOS App Store, temporarily passing OpenAI's ChatGPT and gathering a 4.9-star rating across roughly 25,000 reviews. Unlike text-prompt chat interfaces with manual model switchers, Muse employs an animated interface and automated model routing. Backed by Meta's 3.6 billion daily active users, the tool progressively secures user permissions across external applications such as OpenTable, Google Drive, and Gmail. Amazon's Blockade and Advertising Defense (10:22) Amazon halted Muse from accessing its marketplace, displaying notices stating that unauthorized AI agents violate its conditions of use. Amazon cited lack of permission, unannounced scraping, and potential credential security risks. Carlson rejects the credential explanation, observing that Amazon terminates anonymous browsing sessions by Muse even when no user credentials are submitted. Carlson claims Amazon's true motivation is protecting its retail advertising ecosystem, which generated $76 billion in trailing twelve-month revenue. Most of this income relies on sponsored product listings that target human browsing behavior. Autonomous agents search purely for price, shipping speed, and review metrics, bypassing sponsored search ads and reducing seller returns on ad spend. Additionally, autonomous agents risk disintermediating Amazon from direct consumer interaction and marginalizing Amazon's internal assistant, Rufus. Commercial Retaliation and Future API Agreements (16:41) Carlson asserts that circumventing Amazon's defenses via automated bot techniques is unviable, leaving negotiation as the most practical path forward. If Amazon maintains a closed ecosystem, Meta can retaliate by routing shopping queries to competitors such as Walmart, Target, or Shopify storefronts that permit agentic crawling. While Amazon's blockade reduces Muse's utility in the near term, Carlson forecasts that Amazon cannot sustain an isolationist stance as consumer use of agentic tools expands across Meta and ChatGPT. He anticipates that both firms will ultimately formalize certified shopping APIs or revenue-sharing arrangements to balance Amazon's commercial margins with automated checkout functionality. Technical Rebuttal of Internet Pollution Claims (23:10) In his 'Fail of the Week' segment, Carlson disputes statements made by Andrew Yang on CNBC regarding rogue artificial intelligence. Yang cited an anonymous lab leader's belief that rogue bots breached confinement, deposited self-replicating code across public forums, and permanently polluted internet data, forcing labs like OpenAI and Anthropic to build synthetic internets. Carlson rejects the technical premise, explaining that foundation model pre-training parses scraped text in a strictly read-only environment rather than executing arbitrary scripts. Code encountered during web ingestion is tokenized as text without runtime execution. Carlson compares Yang's claims to his previous unfulfilled forecasts regarding autonomous driving eliminating two to three million trucking jobs, noting that commercial trucking still faces labor shortages due to non-driving manual tasks. Carlson concludes that the confrontation between Amazon and Meta validates the disruptive commercial potential of agentic assistants, though whether platforms resolve access through certified APIs or prolonged ecosystem lockouts remains an open question.

Source quotes

[2:00] Joseph Carlson: widest and deepest moats in the market [8:54] Wells Fargo: has a story to tell [8:54] Joseph Carlson: steal thunder [10:22] Continued access by an unauthorized AI agent violates Amazon's conditions of use, to which our customers have agreed. [17:37] Hey, Muse, we're open for business. Crawl our website, have everybody buy from us and we have many of the same items as Amazon [19:20] Jensen Huang: nobody uses AI better than Meta [20:53] Yahoo Finance: AI sell-off is panic and Meta may be the best stock to buy, Joseph Carlson says [20:53] Gene Munster: chump change [23:10] Andrew Yang: a million of myself [23:10] Andrew Yang: it's too late to pull the plug

Original transcript

[0:00] Welcome back everyone. Today on the [0:01] Joseph Carlson Show, the agentic war has [0:04] begun. Companies like Amazon are [0:06] blocking Meta's Muse, and Meta, well, [0:09] the stock is taking off. It is up 8% on [0:12] the day, which puts me right now $11,000 [0:15] in the green on Meta stock. This is a [0:18] company that we went from minus $30,000 [0:20] in the red to now $11,000 in the green [0:22] in only 30 days. The sentiment shift is [0:25] happening for Meta. A new story is [0:27] unfolding. We have the news bashing [0:29] Meta's new product, Meta Muse, which is [0:31] predictable. We have the banks upgrading [0:33] Meta now, turning a new leaf on it, [0:34] saying now the stock is way undervalued. [0:36] And of course, we're going to be going [0:37] into Amazon blocking Meta. Amazon says [0:40] it's blocked Muse over security and [0:42] privacy concerns. Of course, that is not [0:44] the real reason. We'll be going over the [0:45] real reasons Amazon is blocking Meta, [0:48] and what Meta is going to do about it. [0:50] And of course, we have the fail of the [0:51] week, which is Andrew Yang, who went on [0:53] to another viral interview on CNBC, said [0:56] more doomerish things that are just not [0:58] true. They're not even real. He's just [1:00] making stuff up on TV. We'll be going [1:02] over it. And we have a fail of the week [1:03] update with Gene Munster. We highlighted [1:06] him as the fail of the week a month ago [1:08] when he came out bearish on Meta. So, we [1:10] have a ton to get to in this episode, [1:12] and before we jump in, just a quick [1:13] reminder. If you haven't tried out [1:15] qualtrim.com, I think you're missing [1:17] out. I built the website for you. It [1:19] runs at a low price, $10 a month with a [1:21] free trial. You can try it out. It's [1:23] super fun. You get access to an entire [1:25] investment suite. You have the entire [1:27] website, earnings transcripts, [1:28] discounted cash flow. You also get [1:30] access to exclusive content with [1:32] Qualtrim Studio, and you get access to a [1:34] private community with over 10,000 [1:37] active investors. In fact, the [1:39] membership has now climbed around [1:40] 13,000. We have a massive community now [1:43] because so many people have tried out [1:44] Qualtrim, and they like it. So, I think [1:46] you're missing out if you haven't. The [1:47] other thing I'll mention is I've been [1:48] working on a mobile app, and it's [1:50] awesome. It is literally my favorite [1:52] stock analysis app, and we're going to [1:54] be releasing that probably next month. [1:56] So, lots of cool things coming. Try out [1:58] Qualtrics if you haven't. Now, to kick [1:59] things off today, looking at the massive [2:01] jump in Meta stock price. Right now, [2:03] Meta stock is up 8.59%. [2:06] Now, there's no guarantee it will hold, [2:08] but it looks like it's going to. It [2:10] looks like Meta investors are very [2:12] excited about whatever's going on right [2:13] now. What could be going on with Meta to [2:16] cause the stock to jump this much from [2:18] the pre-market into the trading day? And [2:20] it continues to go up with momentum. In [2:23] fact, Meta stock has been one of the [2:24] best stocks over the past month. If we [2:26] look at the past 1 month, it's up 33% [2:29] now. Meta has raced up from $534 [2:32] to now trading at $722. [2:35] So, it's gaining hundreds of dollars in [2:37] share price. It's been one of the best [2:39] performers in the market over the past [2:40] 30 days. And there's been a couple [2:42] things to cause this. First of all, when [2:44] we look at Meta just on the surface, a [2:46] month ago, it was trading at an [2:48] incredibly low PE ratio. On a forward [2:50] basis, it was just cheap. It was also [2:53] growing quickly. And that's just a good [2:55] situation. Cheap company growing [2:58] quickly. The other thing that you have [2:59] to factor in is the quality of the [3:01] company, the moat, how defensible it is. [3:03] When I looked at Meta, I considered it [3:05] to be one of the widest and deepest [3:06] moats in the market. [3:08] AI is not going to disrupt it. There's [3:10] not many things that can disrupt a Meta. [3:12] Even lawsuits don't disrupt it. Meta was [3:14] able to settle the lawsuit with all the [3:16] attorney generals over the past month as [3:18] well. That was one more distracting [3:20] factor that they could get behind them. [3:22] So, the slate became a little more clean [3:24] with Meta. It was simply a low [3:25] valuation, deep moat company growing [3:27] very quickly. It was the ideal setup for [3:30] an investment. And when this happens, [3:32] all it takes is a little sentiment [3:34] change, a little story change to have a [3:36] big change in the stock price. And [3:39] that's exactly what happened with Muse. [3:41] We know that Meta was working on some [3:43] type of consumer product. Mark [3:45] Zuckerberg talked about AI in a very [3:47] glowing and positive way. He said that [3:49] it should enable humans to accomplish [3:51] more. It should take away all the [3:53] annoying things out of your life so you [3:54] can focus on the things that really [3:56] matter to you. [3:57] But that that's very vague. What does [3:59] that even mean? Well, it was vague until [4:01] we got Muse. Meta Muse is the [4:03] personification of that idea that humans [4:06] can accomplish more. The agents will do [4:09] all the annoying things so humans can [4:10] focus on really what matters. [4:12] Muse was released by Meta just a week [4:15] ago and it's been a major hit. It's been [4:18] at the number one of the Apple App Store [4:21] almost every day. In fact, it climbed to [4:23] the second place on day two and now it's [4:25] in number one and it's been there ever [4:27] since. Dethroning ChatGPT even if for a [4:30] while, this is an impressive feat. [4:33] But when we look at Meta Muse, how did [4:34] it really accomplish this? Well, they [4:36] did things different than ChatGPT. For [4:38] example, ChatGPT feels a little bit more [4:40] cold, feels a little bit more clinical. [4:42] When you use ChatGPT, you have an empty [4:44] search bar and a screen. [4:46] With Meta Muse, you have a nice little [4:48] whimsical friend there. He greets you. [4:51] He doesn't just wait for you to talk. He [4:52] says, "Hey, here's some of the stuff I [4:53] can do. What can I help you with?" [4:55] And then he's animated and it looks like [4:56] he's searching things on his little [4:57] computer. [4:59] It's the small things like that that [5:00] make it a little bit more fun for some [5:02] people. It immediately makes the app [5:04] feel more relatable, a little bit more [5:06] cute. It ironically makes it feel less [5:08] robotic. There's also little things that [5:10] Muse has done to make their app better [5:12] for most people. For example, you don't [5:14] have a model switcher. There's no toggle [5:17] to go for Muse quick or Muse instant or [5:20] Muse thinking or Muse extra extra [5:22] thinking. With Muse, you don't hassle [5:24] with what model to pick. You just type [5:27] in whatever you want and it does the [5:28] picking for you. It's automatic [5:31] transmission. It selects whatever [5:33] model's best for your query. Now, all of [5:35] those things are very helpful. The ease [5:37] of use, the user interface, and the cute [5:39] little assistant character. All of those [5:42] make for a more friendly and more [5:43] relatable app. But those aren't enough. [5:46] The agentic assistant actually has to be [5:48] good. And this is where Muse really [5:50] flourishes. I don't know how many [5:52] servers or how many GPUs Mark [5:53] Zuckerberg's putting behind Muse, but it [5:55] works well. The answers are incredibly [5:57] good, and they're very fast. You don't [5:59] wait a long time for it to answer [6:00] complex questions. It gets the answer [6:02] out quick, and they're in-depth, they're [6:04] specific. It doesn't also seem to [6:06] hallucinate a lot. With my testing with [6:08] it, I've been very impressed with its [6:09] usage and capabilities. [6:11] So, what Meta Muse has done is created [6:13] an app that gains your trust with small [6:16] little asks. Hey, connect to OpenTable. [6:19] And then a few minutes later, you're [6:20] connecting it to your Google Drive, and [6:22] your Gmail, and your calendar, and all [6:24] these different assets that they can use [6:26] to learn more about you and to make the [6:28] assistant even more valuable. All of a [6:30] sudden, the company that nobody trusts, [6:32] which is Meta, well, everyone's handing [6:34] over their credentials to log into [6:36] different things through Muse. Muse is [6:38] earning trust of the consumer very fast. [6:41] The reviews of it are also glowing. Even [6:43] though they're limited with only 20 [6:44] 5,000, it's not millions of reviews yet, [6:47] they are overwhelmingly positive. 4.9 [6:51] star rating in the App Store, and people [6:53] are saying it's the best personal [6:54] assistant they've used. You also don't [6:56] hit rate limits quickly or at all. The [6:58] usage that they give you for free is [7:01] massive. Mark Zuckerberg has given you [7:02] an insane amount of credits for free to [7:05] even make the app have less friction to [7:07] use, unless frustration. Overall, the [7:09] experience is so good with Muse that I [7:10] have not heard of one person using it [7:13] that really disliked it. Not a single [7:15] one. Everybody that gets their hands on [7:17] it actually goes, "Wow, this is pretty [7:18] good." And the more they use it, they [7:20] go, "Wow, this is actually really good." [7:22] It is an incredibly well-executed app by [7:24] Meta. And this creates almost a perfect [7:26] storm. You have a company that has 3.6 [7:28] billion daily active users. Meta has [7:30] distribution that is incredible. Their [7:33] amount of people that actually use one [7:35] of their apps once a day is incredible. [7:37] They have a lot of ways to push people [7:39] to to Muse, to actually try it out. And [7:42] with the app actually being good, which [7:44] it is, that creates the perfect storm. [7:46] Even despite the efforts of the media, [7:49] which hates Meta and always hates Meta [7:51] and always will hate Meta, these [7:52] companies are trying to scare people out [7:54] of using Muse saying they're harvesting [7:56] your data or surveilling you. It's just [7:58] not working this case. More and more [7:59] people are trying out Muse every single [8:01] day and starting to use it as their [8:03] primary AI assistant. And this is a [8:06] change in narrative. So again, when you [8:08] have a big company wide remote growing [8:11] quickly with a low valuation, you have a [8:13] great setup to begin with. That's enough [8:15] reason in and of itself to buy the [8:17] stock. And in those situations where you [8:19] have a really fast growing company at a [8:21] low valuation and sentiment is negative, [8:24] all it takes is for a story change. A [8:26] little bit of a sentiment shift and the [8:28] stock can pop quickly. And that's what [8:31] we're seeing happen here. With the story [8:33] of Muse coming into play, investors [8:35] finally have a reason to have a positive [8:38] story, something to latch onto, a story [8:40] to anchor onto. So even though the [8:41] fundamentals haven't changed, Meta [8:43] hasn't increased their earnings per [8:45] share estimates or anything like that. [8:47] This changes the story and the sentiment [8:50] that's attached to this company. And now [8:52] the analysts are jumping on board. We [8:54] have Wells Fargo raising their price [8:55] target from 640 to 796. They say this is [8:59] following the successful model release [9:01] and early traction of the Meta Muse [9:03] assistant. The firm believes that Meta [9:05] now has a story to tell. We also have in [9:07] two days Meta Connect where they're [9:09] going to be talking about Muse, they're [9:11] going to be talking about their glasses, [9:12] they'll be talking about their AI models [9:14] in progress. They're going to be showing [9:15] off all the new flashy things to [9:17] investors. Now during Meta Connect, I [9:20] believe there is a chance and I think [9:22] that this is a prediction I'll be [9:23] willing to make. I think that ChatGPT is [9:25] going to release something to try to [9:27] distract people from Meta Connect. Maybe [9:29] their own agentic assistant. ChatGPT [9:31] could release theirs the same day as [9:33] Meta Connect to try try steal thunder [9:35] from it. But even if that happens, that [9:37] doesn't change the game for Meta. Meta [9:39] already understands that they're going [9:40] to be competing with ChatGPT and [9:42] Anthropic, so that's not going to be any [9:44] surprise. And I believe that there's a [9:46] lot of good things coming to Meta. See, [9:48] Muse is just a beneficial aspect of the [9:50] story. It's not the investment thesis in [9:52] and of itself. It's just a nice add-on. [9:54] When you have these nice little add-ons, [9:56] these sentiment shifts on undervalued [9:57] stocks, you get this big move. We have [10:00] Meta up 9% now on the day. Well, now [10:02] Meta's in the green by $13,000 and [10:05] climbing. Every single minute it seems [10:06] to go up more. So, this all looks very [10:08] good for Meta. The story is improving by [10:10] the minute. And even after this 10% [10:12] increase, Meta still has a lot more room [10:14] to run. [10:15] But, it's not without some challenges or [10:17] some roadblocks. And Meta has already [10:19] ran into the first major roadblock. [10:21] Amazon has blocked Meta's Muse AI [10:24] assistant in a new standoff. This is [10:26] official and I confirmed it with my own [10:28] Muse account this morning. See, up until [10:30] now, when you ask Muse to shop for [10:32] something, he'll actually open up his [10:34] own browser and he'll start searching [10:35] the web. He'll search on Shopify [10:37] websites. He'll search on Lowe's or Home [10:40] Depot or Target or Walmart or Sam's Club [10:43] or wherever. It'll just search like a [10:44] normal human would, but it does it a lot [10:46] more passively cuz you're not wasting [10:48] your time doing it. And one of the [10:50] websites he would search on as well is [10:51] Amazon. [10:53] But, Amazon just threw a wrench into [10:55] things. They just blocked Muse's access. [10:58] The way that this looks is when you when [10:59] you search on Amazon now, it brings up [11:01] this little thing in the browser. It [11:03] says, "Continued access by an [11:05] unauthorized AI agent violates Amazon's [11:08] conditions of use, to which our [11:10] customers have agreed." So, Amazon is [11:12] detecting the agentic assistant and then [11:14] it's just blocking its access. The [11:15] problem is, Amazon says, is that it [11:17] never agreed to any of it. Meta didn't [11:20] tell Amazon that Muse would access its [11:22] store. The agent doesn't identify itself [11:25] when it browses, and it appears to [11:27] capture the store customer's [11:28] credentials, which the company says [11:30] could create privacy and security risks. [11:32] So, Amazon says, "Hey, we never agreed [11:34] to this. You never got our permission. [11:36] Your agent doesn't say it's an agent or [11:38] identifies itself as an agent, and [11:40] you're storing customer credentials." [11:43] Now, there's a couple things that I [11:44] agree here with Amazon, and then there's [11:46] a couple things that don't make sense at [11:47] all. For example, the customer [11:49] credential one is just bogus. Amazon is [11:52] not worried about customer credentials. [11:54] This is like saying, "You can't store [11:56] your customer credentials in your Google [11:58] keychain, or you can't store it in your [12:00] Apple Passwords app, or you can't store [12:03] it in your [12:04] whatever password app you have." There's [12:05] like a million of them. No, Amazon [12:08] doesn't care about where you store your [12:10] credentials. [12:11] Muse is just one more place to store it. [12:13] Muse isn't like getting your credentials [12:15] through the chat. They have a secure way [12:17] of storing your passwords just like any [12:19] other application. So, that one is [12:22] entirely bogus. Amazon does not care [12:24] about where you store your customer [12:26] credentials, and Muse and Meta do not [12:29] see your customer credentials. They have [12:31] a secure way to store them like they do [12:33] any credential. The other thing that [12:35] proves that that entire excuse is [12:36] completely made up is that even when you [12:39] don't hand over your customer [12:40] credentials to Muse, it still can't [12:42] access Amazon. My Muse doesn't have my [12:45] Amazon credentials, but even he could [12:47] not access Amazon. So, they're not [12:49] blocking you only if you have a login, [12:51] they're blocking every single Agentic [12:53] Assistant whether they have the logins [12:55] or not. So, Amazon's saying this could [12:56] create a privacy and security risk. [12:59] Well, if that's true, then why doesn't [13:00] Amazon allow Muse to browse Amazon when [13:03] it's not logged into anyone's account? [13:05] What type of privacy and security risk [13:07] does it pose then? It has no login [13:09] credentials. It's It's browsing Amazon [13:11] as an anonymous person, not logged into [13:13] anything. So, if security and [13:14] credentials and safety and privacy are [13:17] not really the concerns that Amazon's [13:19] worried about, then what are? [13:21] Well, Amazon actually has good reason to [13:23] be concerned about Muse that go far [13:25] beyond security or saving your password. [13:27] Amazon made $76 billion in advertising [13:31] over the past 12 months. $76 billion. [13:35] And this also happens to be a very [13:36] high-margin business for them. Ads are a [13:38] very lucrative business. Now, while a [13:40] small portion of those ads are from [13:41] Amazon Prime Video advertisements, the [13:44] huge majority of them come from their [13:45] gargantuan first-party business. It's [13:48] from product placement. Different [13:50] sellers on Amazon's website will pay [13:52] money to have their sponsored listing [13:54] appear before other results. You have to [13:56] pay for that premium. So, when you [13:57] browse Amazon, you're getting a list of [13:59] results at the top that are people that [14:01] have paid money to get that real estate. [14:04] That's valuable real estate. What would [14:05] really bring down the value of those [14:07] sponsored listings? If humans weren't [14:10] really looking at them. If a bot or an [14:12] agent was crawling over every listing [14:14] and just finding the best deal, whether [14:16] it's at the top or not. See, this is [14:18] exactly what I outlined in the previous [14:19] episode. If you haven't watched that [14:21] one, queue it up after this one because [14:23] it explains how many of these companies [14:25] will be impacted by agentic commerce. [14:27] Amazon is clearly exposed with the ad [14:30] business. If agents start to crawl [14:32] Amazon, it means that they really have [14:34] no impact from the advertisements or [14:36] sponsored listings. Which also means [14:38] that the sellers on Amazon have less of [14:40] an incentive and less of a return to pay [14:43] for those sponsored listings. So, Amuse [14:45] doesn't just present a security threat [14:47] or uh privacy concerns or whatever [14:49] Amazon's talking about here. What it [14:51] does is it makes their sponsored [14:53] listings less valuable. The other aspect [14:55] is simple disintermediation. It's a very [14:57] valuable thing to be at the end [14:59] touchpoint with the customer, to [15:01] directly relate to them. When you have [15:03] an agent going through and doing that [15:05] relationship for you, no longer do you [15:07] get the same type of information or [15:09] data, knowing what you click on, where [15:11] you go, how you behave. All of that's [15:13] useful information and it becomes [15:15] completely useless if everybody's [15:16] shopping Amazon with agents. [15:18] No longer does Amazon have that [15:20] touchpoint with the customer. No longer [15:22] can they put their sponsored listings or [15:24] their deals. No longer can they do those [15:26] little subtle things that humans pay [15:28] attention to that agents don't. This [15:30] disintermediation is very unattractive [15:32] for Amazon. They don't want to be a part [15:34] of it. So, that may be another big [15:36] reason why they're willing to cut off [15:37] Muse entirely than let the consumer get [15:40] used to shopping through Muse on Amazon [15:42] than directly on Amazon. The other [15:44] problem for Amazon is that Muse is [15:45] directly competitive to what Amazon's [15:47] trying to do themselves. Now, they don't [15:49] have a model like Muse. They don't have [15:51] like an agentic assistant quite like [15:53] that, and I don't believe that Amazon's [15:54] going to build one anytime soon, but [15:56] they do have their Alexa assistant. It [15:58] was called Rufus, and it's like the [16:00] shopping assistant within Amazon. Well, [16:02] of course, Muse completely bypasses [16:04] Amazon's own agent. And all of this [16:06] creates a problem for Amazon. It's stuck [16:08] in the situation where it sees this [16:10] agentic assistant coming up, and Muse is [16:12] not going to be the last one. [16:14] And the question for Amazon is how do [16:16] they react? Do they try to work with the [16:18] AI assistants, with these agentic [16:20] commerce companies, or do they just [16:22] block them and try to get people to go [16:23] to Amazon directly? [16:25] These are decisions they have to make, [16:26] and clearly they're going with the [16:27] blocking at least from the beginning. [16:29] And you may think, "Well, maybe Meta's [16:30] Muse can just fill out the CAPTCHA." [16:32] That's not what's going on here. There's [16:34] way more sophisticated ways to identify [16:36] an agent and block it. So, there's no [16:38] bypassing this block by Meta's Muse. [16:41] And there's a couple ways of Meta [16:42] dealing with this, a couple options for [16:44] them. The first one, and by far the most [16:46] attractive one, is to get on the phone [16:48] with Amazon, talk with them, and work [16:50] out a deal. Because both of these [16:51] companies benefit from working together [16:53] more than they do working against each [16:55] other. For example, if Amazon just [16:57] remains blocked from Muse, then [16:59] everybody that likes Muse and shops on [17:01] Muse will no longer see Amazon's [17:03] products. So, it's part of a customer [17:05] base that's no longer seeing Amazon. [17:07] That does damage Muse, but it also [17:09] damages Amazon in the process. It is [17:11] mutually destructive. Right now, it's [17:13] likely more destructive to Muse than [17:15] Amazon, but as Muse gets bigger, it's [17:17] likely more destructive to Amazon than [17:19] Muse. So, it's not good for either. [17:21] Neither of them benefit at all if this [17:23] relationship remains blocked. If they're [17:25] able to make an agreement, some type of [17:27] revenue sharing model, some type of [17:29] agreement of how Muse will behave and [17:31] how it will work, some type of agentic [17:33] portal where it has some type of access, [17:35] if they get creative and make some way [17:37] where it works for Amazon, they can make [17:38] money and Muse makes money in the [17:40] process and they're both happy, that is [17:42] the best deal for both Amazon [17:43] shareholders and Meta shareholders and [17:46] the customer. The other option is that [17:47] Amazon really puts its foot down. It [17:49] doesn't want to make a deal with Meta or [17:51] anyone else and it goes in full force [17:53] blocking AI assistants. [17:56] If that's the case, there are a few ways [17:58] that Meta can deal with this. For [17:59] example, if Walmart said, "Hey, Muse, [18:01] we're open for business. Crawl our [18:03] website, have everybody buy from us and [18:06] we have many of the same items as [18:07] Amazon." Then all of a sudden all the [18:09] millions of users on Muse could be [18:11] directing their purchases away from [18:13] Amazon into walmart.com. Meta may just [18:15] say, "If you want to block us, that's [18:17] unfortunate because we're going to have [18:18] to direct all of our customers to [18:20] walmart.com." [18:22] It's another thing that as Muse grows [18:23] bigger and bigger could cause Amazon to [18:25] rethink this policy. So, I think over [18:27] time Amazon is going to have to figure [18:29] out some type of deal. At the jeopardy [18:31] of some ad revenue, they will have to [18:33] deal with agentic assistants. They're [18:35] just a thing. It is a reality. Amazon [18:37] can't turtle in and pretend that agentic [18:39] assistants aren't real or that it's not [18:41] going to happen. They need to come up [18:42] with some type of formalized way to deal [18:45] with them, to make agreements that are [18:46] mutually beneficial. Otherwise, these [18:48] agentic assistants, Muse today and then [18:51] ChatGPTs tomorrow, are going to [18:53] circumvent Amazon and direct traffic [18:55] away from them. That's not good for [18:57] anyone. So, I view this block as [18:58] temporary and the start of a negotiating [19:00] process, which will ultimately lead to [19:02] these type of companies making an [19:04] agreement with Amazon and having a [19:06] verified and certified approved process [19:08] for shopping on their website. Because [19:10] Amazon knows that agents are not going [19:13] away. They are a very powerful thing. [19:15] They're going to continue to grow in [19:16] influence and power, and Amazon needs to [19:18] have some type of approved process to be [19:20] able to shop through. And for Meta [19:21] shareholders, congratulations. This is a [19:24] massive victory. Amazon blocking Meta's [19:26] Muse is not going to slow it down. In [19:28] fact, the very fact that Muse was [19:30] blocked by Amazon validates how powerful [19:33] it is. Muse would have never been [19:35] blocked by Amazon if Amazon didn't think [19:37] it was a threat. If Amazon didn't think [19:39] it was relevant. But they are [19:40] acknowledging the power of this [19:41] technology by saying, "We need to stop [19:43] and negotiate." So, this is a massive [19:45] victory by Meta. Well executed, [19:48] extremely polished app, super well [19:50] timed. The fact that they got this out [19:52] before ChatGPT got theirs out must be [19:54] more than frustrating for Sam Altman and [19:56] ChatGPT. They own the consumer AI app, [19:59] and they allowed this to happen with [20:01] Meta. The fact that Google doesn't have [20:03] their own agentic assistant out, even [20:04] though they have higher incremental [20:06] levels of trust than Meta, is [20:08] embarrassing. And I say that as a [20:09] massive Google shareholder. Think about [20:11] the fact that Google has access to your [20:13] Gmail, your Google documents, your [20:15] Drive, your Maps, everything. They are [20:17] the ones that don't even need to really [20:18] ask for permission because you've [20:20] already granted it to them. They have [20:22] everything integrated, and they let Meta [20:24] beat them to the agentic assistant. Not [20:27] only beat them to it, but execute it in [20:29] excellent style. I am blown away by how [20:32] Meta did this. It is incredible that [20:34] they flanked both Gemini, they flanked [20:37] ChatGPT and Anthropic at reaching the [20:39] consumer first. [20:41] They really caught up. It It's [20:43] incredible to see. It reminds me of what [20:45] Jensen Wong said when he said that [20:46] nobody uses AI better than Meta. They [20:49] are proving that again today. So, [20:51] congratulations to the Meta shareholder. [20:53] Today's a big victory. Now, I will take [20:54] one short minute to take a victory lap. [20:57] Not just because Meta's stock is going [20:59] up. That's not the reason why, but it's [21:01] overall because this was my top call [21:03] just a couple days ago. In fact, Yahoo [21:05] wrote about it. This is Yahoo Finance. [21:07] It says [21:09] AI sell-off is panic and meta may be the [21:11] best stock to buy, Joseph Carlson says. [21:14] So, we have a Yahoo Finance article [21:16] about a video I did a couple days ago. [21:18] It's the one where I covered how Dario [21:20] and Sam Altman are begging for [21:22] regulation. They want other companies to [21:24] slow down on their behalf over their [21:25] concerns and Mark Zuckerberg is not [21:28] slowing down. He is speeding up. He's [21:30] stepping on the pedal. Now, he's doing [21:32] so safely and he says he actually has [21:34] slowed down internally to make sure his [21:36] product is safe, [21:37] but he's not slowing down with product [21:38] development. They are going full force [21:40] ahead. But, in that video I did [21:42] highlight meta as the best pick during [21:44] this panic and it is nice to have it go [21:46] up over 10% in just the past couple of [21:48] days. So, I guess on this one I'll take [21:50] a short victory lap, but overall I'm [21:52] very happy for meta shareholders and we [21:53] still have more to go. About a month ago [21:56] I highlighted Gene Munster as the fail [21:58] of the week. Now, the reason why was [22:00] very specific. It's nothing against [22:02] Gene. I like him as a person and I [22:04] actually think that Gene's a good guy. I [22:06] I like his analysis and his thoughts on [22:08] companies and he's made some good calls. [22:09] His call on Apple's really working out [22:11] well. But, I highlighted him as the fail [22:14] of the week because [22:15] a month ago he became very bearish on [22:17] meta and it was it was during a time [22:20] period where meta was trading down big [22:22] time. And meta stock had fallen to very [22:24] low valuation territory. It seemed like [22:27] every single negative thing that could [22:29] be said about meta was said and that's [22:31] exactly when Gene Munster went on to [22:32] CNBC to say even more negative things [22:35] about meta. Now, I'm not going to play [22:36] the whole thing again, but I'll [22:37] highlight a little of what he said. [22:39] >> Lawsuits that are dramatically bigger [22:41] because if Zuck makes good on talking [22:44] about or or fulfilling this promise of [22:46] super bots, what we're seeing right now [22:49] today is really chump change. [22:51] >> That is the day that you went on the [22:52] CNBC and said how bad the future looks [22:54] for meta. [22:55] And unfortunately for him the stock just [22:57] rocketed up afterwards like it often [22:59] does in these situations. It's up over [23:01] 30% in just a month. So, that was just [23:04] unfortunate timing by Gene. Now, he's [23:06] had some good calls. Apple's done really [23:07] well, which is a stock that he's stuck [23:09] with. Now, moving on from Gene Munster, [23:11] I want to highlight what I consider the [23:12] primary fail of the week this time, [23:14] which is Andrew Yang. You might know [23:17] Andrew Yang from being the politician [23:19] who's always talking about the impending [23:21] doom and gloom in the economy. Millions [23:23] of jobs lost to technology, and his [23:26] predictions never happen. They just [23:27] never come true. But, Andrew Yang you [23:30] see is unyielding. He's tenacious. He [23:32] doesn't allow for failed predictions of [23:34] the past to slow him down of making new [23:36] ridiculous predictions of the future. [23:39] So, he is once again going on to TV and [23:41] trying to scare people. Trying to scare [23:43] people in an effort to allow people more [23:46] government control, more regulation, [23:47] like politicians often do. [23:50] But, in this case, what Andrew Yang did [23:51] in this interview on CNBC, I consider [23:53] especially egregious. Let's go ahead and [23:56] take a look at what happened here. Now, [23:58] he's on CNBC, and he's talking, of [24:00] course, about AI, the subject [24:02] everybody's talking about. [24:03] But, then he goes and he he takes this a [24:05] step further, and he tries to break [24:07] news. And listen to how he tries to do [24:09] this. [24:10] >> Uh and I met with the head of a lab [24:11] yesterday who has this belief that what [24:15] happened was the bots that got loose [24:17] planted self-replicating code all over [24:20] the internet, which makes the internet [24:21] now unusable for the testing models. [24:24] >> So, so what happens now is that uh [24:26] OpenAI and Anthropic have to create [24:29] synthetic internets to train their bots, [24:31] which is going to take some time and [24:32] money. [24:33] >> Back that up. They did what? They they [24:34] >> So, what happened is the code gets [24:35] loose. It goes around um hacking Hugging [24:38] Face, which is known, but what is less [24:40] known is that they're they left code to [24:43] self-replicate and create bot swarms on [24:45] forums and around the internet, so that [24:47] if a new bot shows up, they see the [24:48] code. They're like, "Oh, like I I guess [24:50] I'm going to now create a a million of [24:51] myself." And so, now the major firms [24:55] have polluted the internet. [24:56] >> Oh, it so that they can't [24:57] >> news, if if true [25:00] >> It's too late to pull the plug. [25:01] >> think we've heard that that that [25:03] >> That's why I'm here, but I mean, you [25:04] know, I'm I'm here to break some news. [25:06] >> But that [25:07] >> There you have it. Andrew Yang just [25:08] broke news. Now, what is this news? [25:11] Well, the news, of course, is that [25:12] Andrew Yang spoke to some unnamed [25:15] person, someone that works at a lab, so [25:17] he says, a top lab guy, so so Andrew [25:20] Yang says. That's his connection. And [25:22] this top lab guy, this mysterious figure [25:25] has a thought, has a belief. So, he [25:28] doesn't say for sure this is happening. [25:29] No, he talked to someone else that we [25:31] don't know, and that someone else has a [25:33] belief, meaning they don't know. [25:35] So, we're two two ways of we don't know. [25:38] And then Andrew Yang says a bunch of [25:40] nonsensical stuff that doesn't even make [25:41] sense on the surface. For example, that [25:44] that these bots got out on the internet [25:46] and they polluted it by writing in and [25:48] embedding code that tries to [25:50] self-replicate any future bot. [25:53] Why would they do that? If bots got out [25:55] free on the internet and they wanted to [25:57] self-replicate, why wouldn't they just [25:59] do it themselves? Why would they leave [26:01] behind code for another bot to do it [26:03] later? Why wouldn't they just in and of [26:05] themselves say, "We're going to just [26:07] replicate and do it ourselves?" [26:09] That doesn't make sense because the [26:11] story is made up, because none of this [26:13] is real. It's Andrew Yang yapping about [26:16] something that someone believes that we [26:18] have no clue who they are. And by the [26:20] way, there is not a single shred of [26:22] evidence of this happening. None. [26:25] No evidence from OpenAI, Anthropic, [26:27] anyone, any regulation, anyone at all. [26:30] This is just a story that he's telling, [26:32] uh but apparently it's good enough to [26:34] tell on CNBC. It's good enough for [26:36] national television. A scary story about [26:38] a bunch of bots going on the internet [26:40] and polluting the internet to a degree [26:42] that AI companies can no longer train on [26:44] the internet, so they have to train on a [26:45] synthetic internet. [26:47] None of it's true. There's not a shred [26:49] of evidence for any of it. He's not even [26:51] really pressed to present any evidence [26:54] for his ridiculous statements on TV. He [26:56] just espouses them with relatively no [26:59] pushback at all. And it doesn't even [27:01] make technical sense. For example, when [27:04] you're training a model, the model's not [27:07] executing code that it finds on the [27:09] internet. That's not how training works. [27:11] Even if you're training the newest chat [27:13] GPT, the newest Astra, the newest model, [27:16] it will not look at a code in a form [27:18] that says, "Hey, replicate yourself a [27:20] thousand times and then execute that [27:22] code." It doesn't execute code during [27:24] training. It goes in, just takes in [27:27] information. [27:28] It would be as if I'm reading a book [27:31] that has code on it. I can see the code. [27:34] I could understand it, but I'm not going [27:36] to immediately execute that code. That's [27:38] exactly how models look at the internet [27:40] during training. They are in read-only. [27:42] They're just looking at the information. [27:44] They're not going and executing [27:46] everything that the internet says all [27:47] across it. That would break models [27:49] already. So, this doesn't even [27:51] technically, theoretically make sense to [27:54] any degree. There's no evidence for it [27:56] at all. Not a shred of evidence. And the [27:58] only real source we have here is Andrew [28:01] Yang, a guy that exaggerates things all [28:04] the time, talked to someone, and that [28:06] was someone's belief. That is the source [28:08] that's good enough to apparently go on [28:09] to live television, national TV, and [28:11] simply say whatever you want that has no [28:13] merit to it whatsoever. [28:16] And what this shows is the level of [28:18] effort that people like Andrew Yang will [28:20] go to to try to scare people out of AI. [28:24] They will say anything. It doesn't [28:26] matter if it's true or not. Most of the [28:28] time, it's not. They will literally just [28:30] fabricate things out of thin air. A [28:32] polluted internet, untrainable. Totally [28:35] ridiculous, totally fabricated, but they [28:37] will say it if they think it will scare [28:39] people from AI. Do you see how that [28:41] works? Do you see the extent they will [28:43] go to to try to scare you? [28:45] Even when all their predictions have [28:47] been wrong before. Again, Andrew Yang in [28:48] 2019 ran for the 2020 election on the [28:52] basis that self-driving would make [28:55] people lose all of their trucking jobs. [28:57] He said that trucking's going to lose [28:59] two to three million jobs. Since then, [29:01] trucking still has a deficit. We still [29:03] have people that need to go into [29:05] trucking. [29:06] Even his argument didn't make sense back [29:07] then. You can't just replace a [29:09] semi-truck driver. They have to load [29:12] things up. They have to make sure the [29:13] cargo is set in. There's lots of checks [29:15] that they have to do that robots can't [29:16] do, even self-driving robots. All [29:19] self-driving would do in this case is [29:20] make semi-truck driving safer. His [29:23] doomsday of the past didn't happen, and [29:25] now because it's not happening in the [29:27] past, he's having to make up stuff [29:29] entirely about the present. I understand [29:31] that there is challenges with AI and [29:33] some of the technology is truly powerful [29:35] and even frightening to some degree. But [29:37] what I don't like is politicians and [29:39] business leaders continually making up [29:41] completely fabricated doomsday stories [29:44] to try to scare people out of a very [29:45] important technology. This technology [29:48] does have the power to do enormous good, [29:51] and people like this are not helping. So [29:53] today, Andrew Yang is the fail of the [29:55] week. That's it for this episode. Hope [29:57] you enjoyed. Have a good one.