Musk’s SpaceX Agrees to Acquire AI Coding Startup Cursor for $60B
All-In Podcast
SpaceX is leveraging its massive xAI compute infrastructure to absorb Cursor’s revenue growth ahead of an anticipated $2 trillion IPO, even as surging token consumption and private equity debt defaults threaten legacy SaaS business models.
## Background
This episode of the All-In Podcast was published by the All-In Podcast channel on YouTube on April 24, 2026. According to the show notes, the episode covers a potential SpaceX-Cursor deal, a SaaS debt crisis, Apple appointing John Ternus as its new CEO, a federal indictment of the Southern Poverty Law Center, and a Science Corner segment on the rise of colon cancer in young people.
## TL;DR
The besties analyze SpaceX's structured acquisition of AI coding platform Cursor for $60B and its strategic alignment with xAI's Colossus cluster. They transition to the SaaS market correction, examining Thoma Bravo's collapse on Medallia, the shift toward headless agentic software, and the hidden risks of venture debt. After evaluating Tim Cook's legacy and John Ternus's appointment as Apple CEO, the panel dives into a federal grand jury indictment of the SPLC over covert informant funding, closing with a Science Corner breakdown linking a spike in early-onset colon cancer to the herbicide picloram.
## Atmosphere & style
The episode captures the show's signature mix of insider tech analysis, high-stakes finance commentary, political debate, and playful banter. Starting with lighthearted ribbing over David Sacks delaying the recording for an impromptu White House meeting, the hosts maintain an energetic, highly opinionated dynamic. Sharp, analytical breakdowns of private equity and enterprise SaaS give way to fierce libertarian frustration during the SPLC segment, before concluding on an engaging, intellectually curious note during Friedberg's scientific deep dive.
## Key takeaways
- [00:00](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=0s) David Sacks delayed recording after an impromptu White House meeting with President Trump, characterizing him as sharp, genial, and highly focused on accelerating AI infrastructure and behind-the-meter energy deployment.
- [04:55](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=295s) SpaceX agreed to acquire AI coding platform Cursor at a $60B valuation by late 2026, backed by a $10B breakup fee that leverages xAI's Colossus GPU cluster for compute.
- [04:55](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=295s) Cursor's ARR rapidly grew from $2B in February to a projected $6B by year-end, aligning with SpaceX's target of $22B–$24B revenue ahead of its planned $2T IPO.
- [07:14](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=434s) The developer interface (IDE) has become the strategic battleground in AI, with specialized tools acquiring massive model compute to prevent being squeezed out by vertically integrated foundation labs.
- [18:32](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=1112s) Private equity firm Thoma Bravo handed SaaS company Medallia to creditors after debt servicing surged from $100M to $300M, exposing deep vulnerabilities in levered enterprise software.
- [24:00](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=1440s) Public enterprise SaaS stocks face steep drawdowns (e.g., Salesforce, ServiceNow) as seat-based licensing models erode and enterprise spend shifts to API tokens and headless AI agent architectures.
- [35:37](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=2137s) Venture debt poses severe operational risks to high-growth startups during structural market pivots due to strict debt servicing terms and lack of equity-aligned downside protection.
- [46:20](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=2780s) Tim Cook's 15-year tenure at Apple established massive capital efficiency and a 10x market cap expansion, setting up new CEO John Ternus to navigate the transition into an ambient, integrated AI era.
- [51:15](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=3075s) Apple risks hardware stagnation unless it expands beyond the iPhone ecosystem into AI search, wearables, smart home technology, or strategic platform acquisitions.
- [01:00:32](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=3632s) A federal grand jury indicted the Southern Poverty Law Center (SPLC) on 11 counts of wire fraud and money laundering related to hidden offshore accounts and informant payments.
- [01:01:14](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=3674s) Indictment evidence revealed the SPLC funneled over $3M to extremist group leaders—including $270k to an organizer involved in the 2017 Charlottesville event—which subsequently boosted its fundraising revenues to $136M.
- [01:06:58](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=4018s) The hosts criticize the non-profit industrial complex, calling for structural tax reform around 501(c)(3) designations to eliminate perverse incentives that reward manufactured societal crises over problem-solving.
- [01:19:03](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=4743s) A breakthrough Barcelona epigenomic study linked an 80% surge in early-onset colorectal cancer among adults under 50 to long-term exposure to the persistent agricultural herbicide picloram.
- [01:22:27](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=4947s) Spatial epidemiology shows counties with high picloram application exhibit a ~3x higher risk ratio for early-onset colon cancer, prompting calls for an immediate EPA safety review.
## Detailed walkthrough
### [00:00](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=0s) Bestie Intros and Sacks' White House Meeting
The episode opens with banter about Jason Calacanis being caught in the crosshairs of recent political controversies, including the Southern Poverty Law Center (SPLC) indictments. Jason jokingly reveals he is carrying a stiletto and a P35 for protection before the show launches into its signature intro sequence.
The hosts welcome back David Sacks, who delayed the podcast recording by an hour due to an impromptu meeting at the White House with President Donald Trump. Sacks pushes back on mainstream media portrayals of the President as moody or hot-tempered, characterizing him instead as genial, charming, sharp, and genuinely interested in business and technology policy.
The discussion pivots to the administration's stance on artificial intelligence and how critical leadership is during this technological inflection point. Sacks argues that an alternative administration under Kamala Harris would have throttled progress through over-regulation, DEI mandates, and climate-focused restrictions on data centers.
Instead, President Trump has championed letting AI companies build off-grid, behind-the-meter power generation to keep pace with demand. Sacks notes that Trump respects high-IQ builders and maintains a remarkably balanced view of industry leaders—acknowledging the intelligence and product execution of companies like Anthropic, even while navigating differences over Pentagon contracts and left-leaning corporate cultures.
* David Sacks pushed back the recording of Episode 270 by one hour after receiving an impromptu meeting with the President at the White House.
* Sacks highlights Trump's early support for letting AI companies deploy behind-the-meter power generation rather than restricting data center buildouts.
---
[04:55](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=295s) SpaceX Acquires Cursor: The $60B AI Infrastructure Deal
Jason introduces the headline story of the week: SpaceX’s landmark deal to acquire Cursor, the leading AI-assisted coding platform. Under the structured agreement, SpaceX will either acquire Cursor by the end of 2026 at a valuation of $60 billion—$10 billion above its previous rumored raising target—or pay a $10 billion breakup fee to collaborate on building next-generation models.
<Image alt="SpaceX rocket launch juxtaposed with AI code interface" caption="SpaceX & Cursor: AI Compute Meets Code" src="image_agent_tag_1"/>
Cursor's financial trajectory is accelerating at an unprecedented pace, growing from a $2 billion annual run rate at the end of February to a projected $6 billion by the end of 2026. This massive top-line influx is highly accretive to SpaceX, which is projecting $22–24 billion in 2026 revenue as it targets a $2 trillion valuation ahead of its anticipated IPO.
The match brings together Cursor’s proprietary software environment—including its Composer 2 model—and xAI’s colossal compute infrastructure centered at the Colossus cluster in Memphis, which is scaling from 550,000 GPUs up to 1 million units alongside future space-based compute initiatives.
* **SpaceX 2026 Revenue:** Projected at $22B–$24B prior to the Cursor integration.
* **Target IPO Valuation:** SpaceX is aiming for a $2 trillion public market valuation.
* **Colossus GPU Scale:** Currently operating at 550,000 GPUs, expanding to 1,000,000 GPUs.
* **Benchmark Position:** Cursor’s Composer 2 model currently ranks between GPT-4.5/4.6 and Claude Opus on coding leaderboards.
---
[07:14](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=434s) Strategic Value, Compute Leverage, and the IDE War
Chamath breakdown the financial logic behind the transaction, noting that structuring the deal with a future close prevents SpaceX from having to constantly rewrite and delay its S-1 filing with the SEC. By using stock-for-stock terms pegged to SpaceX’s future valuation, Elon Musk effectively secures Cursor at an intrinsic 50% discount while issuing equity at peak valuation multiples.
Chamath points out that xAI had excess capacity at its Colossus cluster following aggressive token discounting for Grok. By acquiring Cursor, Musk repurposes underutilized compute into the most valuable third-party application wrapper in the entire AI ecosystem.
The conversation explores why the developer interface (IDE) has become the decisive battleground in AI. As foundation model generalists like OpenAI (Codex) and Anthropic (Claude Code) vertically integrate into developer tooling, standalone app wrappers face an existential existential squeeze. Cursor gains compute independence and a permanent model partner, while xAI secures millions of real-time developer workflows and proprietary reinforcement learning data.
Chamath warns that while enterprises are currently burning millions on unmanaged AI "agents," they will soon face a centralizing reckoning. Redundant API calls, duplicated data stores, and uncontrolled token usage will force companies back to standard software engineering hygiene managed through powerful IDEs.
* Enterprise token consumption is surging month-over-month without cost controls or optimization routines.
* Mid-tier tasks are routinely routed to hyper-expensive frontier models when open-source or specialized smaller models (like Moonshot or Kimmy K2.6) could execute them at a fraction of the cost.
---
[14:53](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=893s) Cyber Warfare Models and Compute Constraints
David Sacks highlights what he views as the "white-hot center" of AI development: domain-specific cybersecurity models. The release of ultra-large models like Anthropic's Mythos—an estimated 10-trillion parameter model—demonstrated the power of frontier AI as both an offensive exploit generator and an automated defensive shield.
However, massive parameter sizes make models like Mythos prohibitively expensive and compute-bound to serve at scale. Sacks predicts a fierce race among frontier labs to train specialized, distilled cyber models that match Mythos-level vulnerability detection at a fraction of the token cost to satisfy panicking corporate CISOs and IT departments.
<ElicitationsGroup message="Explore related topics from this episode:">
<Elicitation label="Analyze the SaaS debt crisis and market correction" query="Walk me through the All-In Podcast discussion on the SaaS bloodbath and private equity software debt bomb."/>
<Elicitation label="Review the Apple CEO transition to John Ternus" query="Summarize what the All-In Besties said about John Ternus replacing Tim Cook as CEO of Apple."/>
</ElicitationsGroup>
[15:51](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=951s) Cursor Acquisition Structure and Elon's Strategy
Polymarket odds heavily signal that SpaceX's acquisition of Cursor is real, giving it a 74% chance of happening alongside an 80% chance of a SpaceX IPO by the end of August 2026. The structure of the deal is being praised as clever: it keeps the impending IPO on schedule while giving Cursor’s leadership a massive incentive to execute over the next six months. Even though there is a $10 million break-up fee, the hosts point out that the money would likely flow right back to SpaceX anyway to cover the massive compute costs Cursor incurs.
A key concern in tech M&A is maintaining talent momentum after a big payout. Following SpaceX’s earlier acquisition of xAI, several co-founders departed after finding themselves sitting on liquid SpaceX stock. Elon Musk recently addressed these organizational shifts directly, admitting on X that xAI was "not built right the first time around" and is currently being rebuilt from the foundations up—a candid admission that drew over 50 million views.
- **Polymarket Predictions:** 74% chance SpaceX acquires Cursor; 80% chance of a SpaceX IPO by late August 2026.
- **Acquisition Term:** Features a $10 million break-up fee, primarily covering heavy underlying compute usage.
- **Elon Musk Quote:** *"xAI was not built right the first time around, so is being rebuilt from the foundations up."*
- **Consolidation:** Bringing engineering teams into the same physical building eliminates constant task-switching between Tesla, SpaceX, and xAI.
---
[18:32](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=1112s) The SaaS Debt Bomb and Thoma Bravo's Medallia Handback
Transitioning to broader software market troubles, the conversation turns to private equity giant Thoma Bravo handing customer experience SaaS platform Medallia over to its creditors. Acquired in 2021 for $6.4 billion ($5.1 billion in equity and $3 billion in debt), Medallia saw its annual debt servicing costs spike from $100 million to $300 million. When key lenders like Blackstone refused to extend a lifeline, Thoma Bravo walked away.
```
+-----------------------------------------------------------------------+
| MEDALLIA DEAL AT A GLANCE |
+--------------------------+--------------------------------------------+
| 2021 Acquisition Price | $6.4 Billion (All-Cash at Market Peak) |
+--------------------------+--------------------------------------------+
| Initial Debt Load | $3.0 Billion |
+--------------------------+--------------------------------------------+
| Debt Servicing Spike | $100M/year ==> $300M/year |
+--------------------------+--------------------------------------------+
| Wiped-Out Equity | $5.1 Billion |
+--------------------------+--------------------------------------------+
| Sales Performance | Internal reports show team at ~18% of goal |
+--------------------------+--------------------------------------------+
```
While walking away wipes out billions in equity on paper, the besties speculate that Thoma Bravo likely recouped significant capital through prior dividend recapitalizations. However, the core issue remains structural: legacy enterprise software is losing its moat. Instead of purchasing rigid customer feedback software, companies can now use AI agents to spin up custom survey and analytics pipelines in-house for a fraction of the cost.
- **Sales Collapse:** Reports indicate Medallia’s sales team was tracking at just 18% of its target, highlighting severe churn and top-line pressure.
- **Private Equity Playbook:** Leveraged buyouts rely on predictable cash flows to service high debt loads; when net revenue retention breaks, the capital structure implodes.
---
[24:00](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=1440s) Economic Deflation, Fed Policy, and the SaaS Index Crash
During Fed Chair confirmation hearings, Kevin Warsh highlighted the powerful deflationary force of AI. Rather than viewing the "SaaS apocalypse" as an isolated tech crisis, it represents a macro shift: the cost of running an enterprise is dropping rapidly. While deflation can impair levered businesses whose revenues shrink, it frees up corporate budgets to reinvest in growth elsewhere, driving broader productivity.
The public markets reflect this structural shift. Major enterprise software stocks have taken a massive beating over the past six months as tokens replace seats. While category leaders once commanded multiples of 13x ARR, many solid businesses doing $1 billion in ARR now trade around 3x ARR.
- **Kevin Warsh on Inflation:** Stated that standard Fed CPI surveys fail to reflect what everyday Americans experience, while AI will drive unprecedented, deflationary productivity gains.
- **The Value Shift:** Startups and enterprises are moving spending away from legacy SaaS seats toward raw API tokens and custom internal tools.
- **The PE Paradox:** Low multiples present attractive rollup opportunities, but private equity cannot safely leverage assets if net dollar retention unpredictably swings from 120% down to 80%.
Here is a detailed, faithful walkthrough of this section of the transcript, capturing the arguments, data points, and dynamic between the hosts.
[31:07](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=1867s) The SaaS Pricing Crisis and Headless Architectures
Chamath Palihapitiya and Jason Calacanis dissect the core structural issues facing legacy Software-as-a-Service (SaaS) vendors. Beyond balance-sheet debt, these companies are trapped by inflated unit pricing. While many legacy SaaS vendors generate $1 billion in annual recurring revenue (ARR), their unit costs relative to value have spun out of control—historically accounting for roughly 10% of value, but swelling to nearly 30%. As software contracts come up for renewal, enterprise buyers are looking to slash spending by 50%, two-thirds, or even 75%.
The shift toward "headless" software architectures—where AI agents interact directly with underlying enterprise software via protocols like Model Context Protocol (MCP)—is fundamentally breaking traditional per-seat pricing models. As an example, a company like "Freebook" no longer needs 50 seat licenses for Workday; it only needs two seats to serve as the integration bridge for autonomous agents writing into and reading from the database. Multiplied across millions of enterprise customers, this shift creates a structural price compression. Unless incumbent vendors reset their pricing or new entrants deliver lower-cost alternatives, the market cannot complete its necessary clearing process.
- **Unit Cost Inflation:** Historical SaaS cost-to-value sat around 10%; current enterprise contracts sit closer to 30%, driving buyers to demand steep renewal cuts.
- **Seat-Model Collapse:** Headless agentic integration reduces seat requirements drastically (e.g., dropping from 50 Workday seats to 2).
[32:31](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=1951s) Salesforce, Bargain Hunting, and Founder-Led Maneuverability
The hosts evaluate Salesforce as a prime case study in legacy SaaS repricing. Shares of Salesforce fell 9% on the day of recording, leaving it at a $140 billion enterprise value against $15 billion in free cash flow—trading at less than 10 times free cash flow. Calacanis and David Sacks note that this valuation looks like a historical bargain, pointing out Marc Benioff's strategic pivot toward headless product architectures and aggressive share repurchases. Sacks highlights the stark tactical contrast between incumbents: while Workday reacted by attempting to charge a "toll" whenever an AI agent interacts with its system, Benioff embraced headless capabilities across Salesforce's platform.
Chamath cautions that standard valuation metrics may be misleading. If market dynamics compress software cash flows by cutting off terminal value beyond years five to seven, valuations will naturally compress down to 3 to 5 times free cash flow. However, massive free cash flow reserves provide strategic optionality—a playbook previously executed by Apple, Meta, Google, and Uber. Sacks and Calacanis conclude that the defining metric for surviving the AI transition may be whether a company remains founder-led. Founders have the authority to "burn the boats" and radically reinvent business models, whereas hired professional managers tend to retreat to defensive maneuvers, such as introducing new fee structures to protect legacy revenues.
- **Salesforce Valuation:** Trading at under 10x free cash flow ($140 billion enterprise value on $15 billion free cash flow after a 9% single-day drop).
- **Strategy Split:** Workday is attempting to toll AI agent access, while Salesforce under Marc Benioff is moving toward a fully headless stack.
- **Valuation Compression Risk:** If terminal cash flows past years 5–7 are discounted to zero, SaaS cash-flow multiples could fall to 3x–5x regardless of underlying business quality.
[35:37](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=2137s) The Fragility and Pitfalls of Venture Debt
Sacks shifts the discussion to venture debt, expressing strong opposition to venture-backed founders taking on debt instruments. While private equity firms utilize leverage effectively, venture-backed founders frequently misinterpret venture debt as non-dilutive equity, forgetting that it carries strict repayment obligations and operational covenants. Bank oversight limits a founder's ability to make abrupt strategic pivots during market shifts. Calacanis and Sacks emphasize that equity financing aligns long-term incentives and preserves strategic optionality, whereas debt instruments make hot startups brittle.
Chamath shares a personal story from his own past, recounting a period when he managed a $420 million credit line. During a sudden market downturn, the underlying assets collateralizing the loan shrank rapidly in value, triggering a terrifying inward collapse of the credit facility. Concurrently, over a single weekend, hundreds of millions of dollars he held at Credit Suisse were endangered as the bank teetered on the edge of collapse. He describes it as "the worst moment of my professional working life," admitting he violated his core rule against using debt in pursuit of higher returns. He cautions founders that banks will not allow startups to burn through debt as their final 3 to 6 months of runway, as debt lenders cannot absorb total losses in exchange for modest 8% yields.
- **VC vs. Debt Returns:** Venture capital investors accept high loss rates because top performers yield 10x to 1,000x returns; debt lenders earning ~8% yields cannot tolerate default losses and will aggressively seize assets or extract heavy concessions (doubling interest rates, demanding warrants).
- **Personal Anecdote:** Chamath detailed almost losing everything while managing a $420M credit line alongside counterparty risk with a couple hundred million dollars parked at a failing Credit Suisse.
- **Warren Buffett Principle:** Highly intelligent executives routinely go bankrupt by taking on leverage.
[40:18](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=2418s) Government Debt, Pension Crises, and Public Waste
The conversation turns to macro-level debt and government spending. Sacks draws a historical parallel to the corporate pension crises of the 1950s. During that era, American corporations realized that defined-benefit pension plans were inherently unstable—either severely underfunded, risking bankruptcy, or overfunded, locking up unproductive capital. Consequently, private industry systematically transitioned to defined-contribution models like 401(k)s. However, public sector employee unions blocked similar reforms in government, leaving state and local governments burdened with massive unfunded pension liabilities that threaten fiscal solvency 70 years later.
Calacanis notes public figures like Spencer Pratt bringing transparency to public employee compensation in Southern California, where government workers frequently double their overtime pay immediately prior to retirement to inflate their lifelong pension payouts. The hosts frame government-run programs as susceptible to systemic inefficiency and fraud, citing California's spending of $12 billion annually on the "homeless industrial complex" despite worsening homelessness metrics.
- **Corporate vs. Public Pension Evolution:** Private enterprise phased out defined-benefit pensions in favor of 401(k)s starting in the 1950s; public sector unions preserved defined-benefit pensions, accumulating severe long-term liabilities.
- **California Homelessness Budget:** $12 billion allocated annually, yet public homelessness metrics continue to rise.
[42:10](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=2530s) Citizen Journalism vs. Legacy Media on Public Fraud
The hosts discuss independent citizen journalist Nick Shirley, who has been uncovering billions of dollars in public fraud across California. Sacks notes that Shirley could generate substantial personal wealth by acting as a private whistleblower to collect statutory whistleblower awards, though Calacanis argues his public reporting delivers far greater public utility. The hosts criticize legacy media outlets for abandoning traditional investigative reporting, such as scrutinizing government waste and mismanagement, in favor of commentary and institutional defense.
Calacanis points out that legacy outlets are being forced to adapt to citizen journalism, referencing Bari Weiss at CBS tasking reporters to adopt investigative playbooks on public waste. Sacks laments that mainstream media historically exposed government waste—such as infamous reports on $900 defense hammers or Tyco CEO Dennis Kozlowski’s expensive corporate extravagances—but now routinely vilifies independent auditors and tech leaders who attempt to eliminate waste, fraud, and abuse from public budgets.
- **Whistleblower Potential:** Shirley's public-interest investigative work has uncovered billions in state-level fraud.
- **Media Shift:** Legacy media has largely transitioned from investigative reporting into opinion-driven coverage, leaving a vacuum filled by independent street-level citizen journalists.
<ElicitationsGroup message="Where would you like to go next?">
<Elicitation label="Analyze the next transcript section" query="Please provide the walkthrough for the next part of this transcript."/>
<Elicitation label="Summarize SaaS debt & pricing dynamics" query="Summarize the key arguments around SaaS overpricing, seat-model collapse, and venture debt from this transcript section."/>
</ElicitationsGroup>
Jason Calacanis reminds listeners that while general tickets for the fifth annual All-In Summit in Los Angeles (September 13–15) sold out almost immediately, applications remain open at `[allin.com/events](https://allin.com/events)`. He urges attendees to secure their passes early rather than asking for favors late in the process. David Sacks makes a major announcement regarding the summit's programming: they will host a single high-profile political panel featuring Pennsylvania’s two sitting U.S. Senators, Democrat John Fetterman and Republican Dave McCormick.
The besties joke about the contrasting personal styles of the two senators on stage. Fetterman's trademark casual aesthetic of hoodies, shorts, and Timberland boots will share the stage with McCormick's tailored Brioni suits.
- **Summit Details:** 5th Edition of the All-In Summit in Los Angeles, CA (September 13–15, 2026).
- **Featured Panel:** A bipartisan discussion with U.S. Senators Dave McCormick (R-PA) and John Fetterman (D-PA).
---
[46:20](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=2780s) New Apple CEO: John Ternus Succeeds Tim Cook
The conversation turns to the news of Tim Cook’s resignation and the appointment of 25-year Apple veteran John Ternus as the new CEO. Ternus, who previously led hardware engineering for products like the iPad and AirPods, had been the long-time betting favorite on Polymarket. Jason notes that Apple’s core lineup (Siri, AirPods, iOS ecosystem) has grown stagnant under a purely profit-driven model and asks what direction the new CEO should take to reignite innovation.
David Friedberg argues that the computing software paradigm is fundamentally shifting and that Apple’s priority must be creating a unified, ubiquitous AI layer integrated across all devices. This system should seamlessly access personal emails, calendar entries, home automations, and media preferences using natural language. Jason suggests acquiring specialized voice-AI companies like WhisperFlow to replace the existing Siri team.
- **Leadership Transition:** John Ternus succeeds Tim Cook as Apple CEO after leading hardware teams for iPad and AirPods.
- **Market Prediction:** Ternus remained the favorite on Polymarket leading up to the announcement.
- **AI Integration Goal:** Friedberg highlights the urgent need for a context-aware, ubiquitous AI layer across all Apple hardware.
---
[48:50](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=2930s) Tim Cook’s Legacy: Capital Allocation and Privacy Focus
David Sacks defends Tim Cook’s 15-year tenure as CEO, emphasizing his operational efficiency and massive financial results. Under Cook's leadership, Apple’s market capitalization surged by more than 10x, while annual revenue expanded from roughly $100 billion to over $400 billion. Cook also shifted the revenue mix toward higher-margin services and protected the brand's public reputation through a strong commitment to user privacy and strategic relationships in Washington.
Chamath Palihapitiya frames Cook’s role as an "impossible job"—succeeding Steve Jobs is akin to playing basketball alongside Michael Jordan and then being tasked with *being* Michael Jordan. He categorizes CEOs into two buckets: the visionary innovator and the master steward. Cook stands at the top of the steward category, driving massive capital efficiency by replacing Intel chips with custom Apple Silicon, maintaining lightweight M&A, and shrinking Apple's total share count by nearly 40–50% through aggressive buybacks.
```
+-------------------------------------------------------------------------+
| TIM COOK'S 15-YEAR TRACK RECORD |
+--------------------------------+----------------------------------------+
| Metric / Strategy | Outcome under Tim Cook |
+--------------------------------+----------------------------------------+
| Annual Revenue Growth | $100 Billion → $400+ Billion |
| Market Capitalization | Increased > 10x |
| Capital Return Strategy | Reduced share count by ~44–50% |
| Key Engineering Milestone | Transitioned from Intel to Apple Silicon|
+--------------------------------+----------------------------------------+
```
- **Financial Growth:** Revenue scaled from ~$100B to $400B+ annually; market cap grew by over 10x.
- **Share Repurchases:** Apple reduced its total share count by almost 50% under Cook, contrasting sharply with the Steve Jobs era where zero capital was returned to shareholders.
- **Silicon Strategy:** In-house R&D successfully eliminated reliance on Intel chips, setting up Apple’s on-device processing capabilities.
---
[51:15](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=3075s) Missed Opportunities & Challenges Facing John Ternus
When Sacks asks what product categories Apple missed, Jason lists four major opportunities: lightweight smart glasses, an AI-first search/answer engine, a self-driving automobile, and a standalone integrated Apple television set. He argues that after the loss of Steve Jobs and chief designer Jony Ive, Apple lost its creative soul and shifted into iteration and margin extraction. While products like the Apple Vision Pro were pushed to market, they have yet to achieve widespread consumer adoption.
Chamath warns of the long-term risk facing John Ternus: over-reliance on a single high-margin product line like the iPhone. As AI open-source protocols, Model Context Protocol (MCP), and multi-modal agents expand, users will interact with knowledge through a heterogeneous mix of hardware—such as smart glasses, pens, and ambient pins—rather than just two dominant mobile operating systems.
- **Missed Product Frontiers:** Smart glasses, self-driving cars, dedicated home TVs, and native AI search/answer engines.
- **Wearables & Robotics:** The besties point out that future growth requires deeper AI integration across wearables (AirPods, smart pins, watches) and potentially consumer robotics.
- **Hardware Heterogeneity Risk:** Rising per-unit pricing strategies in hardware and SaaS face structural disruption as ambient AI devices decouple interactions from traditional smartphones.
---
[57:40](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=3460s) Historical Parallels: Disney’s Succession Lessons for Apple
Sacks draws a direct historical parallel between Apple's current transition and the corporate history of the Walt Disney Company. Before his death, Steve Jobs explicitly advised Tim Cook not to repeat Disney’s mistakes by becoming trapped in the founder's historical vision. When Walt Disney passed away, his brother Roy Disney—acting as the operational, business-focused co-founder much like Tim Cook—maintained the company's momentum for five years before passing away in 1971. Afterwards, Disney experienced a stagnant period throughout the 1970s until Michael Eisner took over in 1984 and revitalized the enterprise.
Jason expands on the Disney comparison, noting how Michael Eisner instituted the "vault strategy"—re-releasing classic animated movies to theaters every seven years—to unlock value from existing IP. Later, Bob Iger leveraged Disney's pristine balance sheet and theme park distribution to execute bold strategic acquisitions like Pixar, Marvel, and Star Wars. Jason suggests that John Ternus could follow Iger's playbook by using Apple’s massive cash reserves for bold acquisitions rather than relying strictly on share buybacks.
- **Roy Disney Parallel:** Tim Cook served as the master operational steward following the visionary founder, expanding Apple's enterprise value tenfold over 15 years.
- **1970s Stagnation vs. 1980s Revitalization:** Sacks notes the key question for John Ternus is whether Apple will drift into an uninspired period or immediately execute a new growth playbook.
- **Acquisition Strategy:** Instead of relying entirely on capital returns, Apple could use its cash flow to acquire strategic technology platforms and IP.
---
[01:00:32](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=3632s) Southern Poverty Law Center (SPLC) Indictment
The besties shift to a news segment covering a federal grand jury indictment of the Southern Poverty Law Center (SPLC). The panel discusses allegations surrounding the organization's financial practices, specifically highlighting how the entity accumulated over $800 million in offshore bank accounts while operating under its non-profit status.
Jason outlines the core charges, stating that the SPLC has been indicted on 11 counts of wire fraud and money laundering covering activities between 2014 and 2023.
- **Offshore Holdings:** The discussion highlights $822 million accumulated by the non-profit organization in offshore accounts.
- **Federal Grand Jury Indictment:** 11 counts involving alleged wire fraud and money laundering schemes spanning 2014 through 2023.
[01:01:14](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=3674s) The SPLC Indictment: Informants, Fictional Entities, and Charlottesville
Jason reads through the explosive details of the federal grand jury indictment against the Southern Poverty Law Center (SPLC). The indictment alleges that the SPLC used hidden bank accounts to funnel $3 million in donor money to paid confidential informants to infiltrate groups like the KKK, the Aryan Nation, neo-Nazi organizations, and the Unite the Right organizers. While the SPLC’s stated mission is "to be a catalyst for racial justice in the South and beyond," the podcasters point out how wildly the organization expanded its targets over the years. Jason notes how the SPLC began loosely labeling moderate figures as "hate-adjacent" in their "Hate Watch" headlines—including their mutual friend Sam Harris—in order to fuel cancellation campaigns.
The discussion turns to the financial engine driving the SPLC. Following the 2017 Charlottesville event—and the heavily spun media narrative surrounding President Trump's comments—the SPLC saw its fundraising skyrocket. In 2016, their revenue stood at $58 million; post-Charlottesville, it more than doubled to $136 million and has remained elevated ever since. According to the indictment, a key confidential informant designated as "F-37" was an active member of the online leadership chat group that planned the 2017 Unite the Right event. Acting under SPLC supervision, F-37 posted racist content online, helped coordinate transportation for attendees, and received over $270,000 from the SPLC between 2015 and 2023.
- The SPLC's mission statement claims it works to "dismantle white supremacy, strengthen intersectional movements, and advance the human rights of all people."
- Targets included the KKK, Aryan Nation, neo-Nazi groups, Unite the Right organizers, the Proud Boys (labeled a hate group), and the Oath Keepers (labeled part of the militia movement).
- SPLC annual revenue spiked from $58 million in 2016 to $136 million post-Charlottesville.
- Informant F-37 was paid over $270,000 by the SPLC while actively helping organize transportation and posting racist content leading up to Charlottesville.
[01:03:36](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=3816s) Fomenting Extremism for Fundraisers: A $270k Investment for an $81M Return
David Sacks steps in to add critical details, emphasizing that the $270,000 paid to informant F-37 is only a fraction of the story. The indictment alleges that the SPLC secretly funneled over $3 million to a wide swath of violent, racist extremist groups, including the Ku Klux Klan, the American Nazi Party, Aryan Nation, and the United Klans of America. Sacks argues that an organization supposedly dedicated to fighting racism was actually funding and fomenting it. By paying these groups to organize high-profile events and racial clashes, the SPLC created the very specter of widespread hate they could point to when soliciting money from wealthy donors.
Sacks calls it a classic "grift," noting the absurd return on investment: a $270,000 spend to help orchestrate events around Charlottesville yielded an $81 million surge in immediate fundraising. To keep donors—ranging from Hollywood celebrities to corporate givers—in the dark, the SPLC set up shell bank accounts under fictitious entities. While Jason points out the SPLC's legal defense—that non-profits naturally hide informant payment origins so as not to blow their cover, much like the FBI—Sacks rejects the comparison. He notes that the SPLC is not a law enforcement agency, they were paying the actual leaders of these hate groups rather than passive moles, and they were instructing those leaders to provoke activity rather than merely report on it.
> "So, this group that was supposed to be fighting racism in fact was fomenting racism by paying these groups to basically organize protests that SPLC could then point to and say that America has a huge racism problem, donate to us... That $270,000 investment led to an $81 million return, pretty good." — **David Sacks**
[01:06:58](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=4018s) NGO Overlords, Tax-Exempt Abuse, and the Libertarian Reset
Chamath and Friedberg react with fury over how the non-profit industrial complex operates. Chamath argues that modern NGOs are "cosplaying as overlords and power brokers" and should be entirely dismantled. He calls on donors to sue the SPLC and demand their money back, highlighting that the organization is sitting on an $822 million offshore war chest. He outlines what he sees as the standard NGO playbook: secretly provoke the problem, feed the chaos to friendly media outlets to blow it out of proportion, raise tens of millions of dollars off the moral panic, and consolidate political power.
Friedberg zeroes in on the IRS tax code, reading the legal definition of 501(c)(3) exempt activities—charitable, religious, educational, scientific, literary, public safety, or preventing cruelty to children or animals. He demands to know how 90% of modern non-profits qualify under this standard. As a libertarian, Friedberg insists that while people should be free to fund whatever activism they want, these groups must lose their tax-exempt status, stop receiving taxpayer money, and lose the ability to offer tax deductions to donors. Jason and the besties joke about their own encounters with cancellation culture—including Jason recalling a 2014 Vice article labeling him with the "most offensive tweet ever" over a blog post about tech journalism.
- **SPLC Cash Reserves:** Chamath notes the SPLC currently has $822 million sitting in offshore accounts.
- **IRS 501(c)(3) Exempt Purpose Definition:** Must be strictly limited to charitable, religious, educational, scientific, literary, testing for public safety, fostering amateur sports, or preventing cruelty to children/animals.
- **The Non-Profit Playbook:** 1) Provoke or create societal tension, 2) Leak to aligned media allies, 3) Amplify the crisis, 4) Execute massive fundraising drives off the outrage.
[01:12:35](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=4355s) Grand Jury Realities and the Structural Flaws of Non-Profits
The conversation turns to whether traditional human rights organizations have completely lost their way. Chamath asks Jason—who previously worked at Amnesty International—what the odds are that groups like Human Rights Watch or Amnesty are actively manufacturing human rights abuses to justify their operational budgets. Jason admits that while Amnesty had a noble original mandate based on the UN Universal Declaration of Human Rights (focusing on political prisoners and torture), many of these groups went completely off-track to chase new funding streams, giving it a 50/50 chance that modern NGOs engage in sketchy tactics to whip up panic.
Sacks points out that a federal grand jury indictment is a massive hurdle, meaning prosecutors have already presented hard evidence of wire fraud and concealed accounts rather than just launching baseless political "lawfare." The group concludes by contrasting private business with non-profit incentives. In free-market business, a company must deliver real value or face bankruptcy. In the NGO world, there is no natural market feedback loop; in fact, the incentives are inverted—the worse the societal problem appears, the more money the non-profit collects.
- **Grand Jury Standard:** Sacks highlights that grand jury indictments require presenting concrete physical and financial evidence to citizens, making the concealed bank accounts hard for SPLC to explain away.
- **Structural Inversion:** Businesses fail when they don't solve problems; NGOs risk losing their funding if the problem they claim to fight is actually solved.
[01:15:27](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=4527s) The NGO Fundraising Loop and the Evolution of Civil Rights Goals
David Sacks and the group dive into the structural incentives that govern non-governmental organizations (NGOs) and non-profits, outlining how these entities prioritize institutional survival over solving the core problems they were established to address. Sacks details how NGOs raise funds from donors to engage in a specific mission, but over time, the real-world impact of their activity stops mattering as much as maintaining the fundraising apparatus itself. The primary objective shifts toward inventing justifications to repeatedly ask donors for more capital.
The incentives become even more distorted when NGOs receive federal funding. In those instances, the organization’s primary focus pivots entirely toward lobbying the government for larger budget allocations, irrespective of whether their underlying programs are functional or effective. The only requirement is that the organization can successfully spin the program as a success to secure its continued funding and preserve jobs.
Addressing the Southern Poverty Law Center (SPLC) specifically, Sacks lays out his theory on how civil rights organizations transitioned from their original historic mandates into modern ideological movements. He acknowledges that civil rights was once a noble and vital cause in the United States, established to dismantle the legacy of segregation and Jim Crow laws. However, Sacks asserts that non-profits never declare victory or disband when a goal is reached; no executive voluntarily offers to be fired because their job is complete. He argues that following the election of Barack Obama in 2008, it became evident to most people that systemic legal barriers based on skin color had been overcome, presenting an opportunity for these groups to wrap up their initial missions.
Instead of declaring victory, the goalposts were shifted during Obama's second term toward "anti-racism." Sacks explains that anti-racism redefined success from equality of opportunity to equality of outcomes, demanding that demographics across all societal positions match overall population distributions. He argues that because the public would have rejected a direct push for identity socialism or outcome-level engineering, new academic and institutional terminology was created to institutionalize the effort—a shift that has taken years for the public to unpack and recognize. Jason Calacanis adds a slight defense of historical contexts, noting that while groups like the SPLC previously played a crucial role alongside agencies like the FBI in infiltrating the Ku Klux Klan and neo-Nazi organizations, that degree of continuous domestic monitoring isn't operating under the same necessities in 2026.
- Sacks argues that NGOs funded by government grants prioritize lobbying for budget expansion over real-world program efficacy.
- The shift from traditional civil rights to "anti-racism" represents a fundamental movement from equal opportunity to equal outcomes.
- Historically, groups like the SPLC partnered with federal law enforcement to dismantle organized hate groups, though the hosts question the scale of their modern mandate.
[01:19:03](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=4743s) Science Corner: Investigating the Spike in Youth Colorectal Cancer
Transitioning into a surprise installment of "Science Corner," High-Tech / Science host Division Lead (referred to as Friedberg) presents a newly published, landmark research paper investigating the underlying environmental causes of the dramatic rise in early-onset colorectal cancer. Friedberg notes that colon cancer has now climbed to become the third leading cause of cancer overall, highlighting a terrifying demographic shift over the last two decades. Historically considered an age-related disease primarily affecting individuals over 70, cases among young adults under the age of 50 have surged by more than 80% over the last 20 years.
To isolate the underlying trigger, a research team out of Barcelona, Spain, conducted an epigenomic study analyzing tumor tissue samples from colon adenocarcinoma patients. By comparing the gene expression profiles (RNA output) of cancer cells in patients under 50 against those over 70, the researchers sought to trace environmental exposures that alter how genes are switched on or off. The cellular samples for the study were sourced directly from The Cancer Genome Atlas (TCGA), a federally funded database managed by the National Cancer Institute.
- Early-onset colorectal cancer (diagnosed in individuals under 50 years old) has increased by over 80% in the last two decades.
- The Barcelona study analyzed epigenomic data and RNA expression to detect environmental triggers causing cellular malfunction.
- Tissue samples were retrieved from The Cancer Genome Atlas (TCGA), a multi-million-dollar federal scientific repository.
[01:22:27](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=4947s) Picloram Exposure and Epigenomic Risk Factors
After controlling for lifestyle factors, caloric intake, obesity rates, alcohol consumption, birth weight, and gender, the research team identified a single overwhelmingly consistent variable linked to early-onset colon cancer: exposure to an agricultural herbicide named **picloram**.
Friedberg outlines the history of picloram, noting it was originally developed by The Dow Chemical Company in 1963. Designed as a synthetic auxin (a mimic of natural plant growth hormones), the chemical causes targeted weeds to experience rapid, uncontrolled growth spurts that quickly lead to plant death. It has been extensively utilized to clear broadleaf weeds on pastureland and rangeland where cattle graze, as well as along railway lines, highways, and utility corridors.
The primary danger of picloram lies in its environmental persistence. Unlike rapidly degrading compounds, picloram resists biodegradation, remaining active in soil and surface water for well over a year while leaching into local groundwater supplies. Friedberg highlights a major regulatory oversight: the last formal safety evaluation conducted on picloram by the U.S. Environmental Protection Agency (EPA) occurred in 1995. Thirty years ago, regulatory testing lacked the modern epigenomic sequencing tools required to detect how long-term, low-dose exposureal shifts turn critical regulatory genes on or off over time, eventually causing tissue cells to go haywire and develop into malignant tumors.
- **Developer & History:** Picloram was created by Dow Chemical in 1963 as a synthetic auxin herbicide.
- **Common Applications:** Used for broadleaf weed control in cattle pastures, rangelands, roadside maintenance, and power line utility rights-of-way.
- **Environmental Behavior:** Highly persistent; resists biodegradation for over 12 months and easily leaches into groundwater.
- **Regulatory Gap:** The U.S. EPA last performed a comprehensive safety study on picloram in 1995, long before epigenomic RNA sequencing was available.
[01:25:31](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=5131s) Geographic Correlation, Patent Status, and Federal Science Funding
When asked whether picloram is manufactured by a single corporate entity like Monsanto, Friedberg explains that the chemical is long off-patent. It is manufactured generically at scale—primarily in China—and distributed globally under hundreds of different commercial brand names.
To validate their molecular findings, the research team correlated picloram application data from the U.S. Geological Survey's Pesticide National Synthesis Project against county-level cancer rates in states with robust reporting data, including California, Connecticut, Georgia, Iowa, New Mexico, Utah, and Washington. The spatial epidemiology yielded a striking correlation: counties featuring higher environmental applications of picloram exhibited a significantly higher frequency of early-onset colon cancer cases, returning a strong odds ratio of approximately 3x.
Friedberg uses the study to highlight the critical role of federal funding for fundamental science. The National Cancer Institute established The Cancer Genome Atlas with a $100 million upfront investment, maintaining it for just a few million dollars annually. That baseline infrastructure enabled an independent foreign university to access human tissue samples and uncover a widespread carcinogen risk without relying on corporate research funding.
Friedberg argues that the EPA should immediately initiate a formal regulatory review of picloram to determine if it should be banned from the food supply and industrial applications. Furthermore, regulatory agencies should mandate epigenomic screening protocols for all industrial and agricultural chemicals to detect chronic toxicity before products enter the environment.
The segment closes with lighthearted banter as Jason, Chamath, and Sacks joke about Friedberg's upcoming role on the President's Council of Advisors on Science and Technology (PCAST) and his meticulous deep dive into oncology research, before wrapping up the segment.
- **Epidemiological Signal:** Counties with high historical picloram application showed up to a 3x higher risk factor (odds ratio) for early-onset colon cancer.
- **Government Role in Research:** The discovery relied directly on publicly available tissue samples maintained by the federally funded Cancer Genome Atlas.
- **Policy Call:** Friedberg urges the EPA to conduct an immediate review to remove persistent epigenomic toxins like picloram from agricultural and industrial supply chains.
[01:29:25](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=5365s) Podcast Outro and Sign-Off
Jason Calacanis wraps up episode 270 of the "world's greatest podcast," delivering his signature closing remarks as the host and moderator. He thanks his fellow Besties—Chamath Palihapitiya, David Sacks, and David Friedberg—for their contributions to the discussion. With a cheerful farewell to the listeners, their friends, and neighbors, Jason signs off until the next episode.
Chamath adds a final "Love you boys. Bye-bye," initiating the podcast's iconic outro audio montage. The clip highlights classic show banter, catchphrases, and signature soundbites, including references to "Rain Man David Sacks," the "Dictator," "Queen of Quinoa," and the recurring motto, "Let your winners ride." The episode comes to a humorous and lighthearted end with a classic clip of banter about a dog taking a dump in Sacks's driveway, culminating in the familiar "I'm going all in" musical theme.
- Episode Number: 270
- Present Besties: Jason Calacanis, Chamath Palihapitiya, David Sacks, David Friedberg
## Notable quotes
> [15:51](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=951s) xAI was not built right the first time around, so is being rebuilt from the foundations up. — Elon Musk
> [01:13:36](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=4416s) So, this group that was supposed to be fighting racism in fact was fomenting racism by paying these groups to basically organize protests that SPLC could then point to and say that America has a huge racism problem, donate to us... That $270,000 investment led to an $81 million return, pretty good. — David Sacks
> [01:29:25](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=5365s) Love you boys. Bye-bye. — Chamath Palihapitiya
## Mentioned
SpaceX, Cursor, Apple, Southern Poverty Law Center (SPLC), Donald Trump, Kamala Harris, David Sacks, Jason Calacanis, Chamath Palihapitiya, David Friedberg, John Ternus, Tim Cook, Steve Jobs, Marc Benioff, Salesforce, Workday, Thoma Bravo, Medallia, Blackstone, Kevin Warsh, Nick Shirley, John Fetterman, Dave McCormick, Anthropic, OpenAI, Claude, Grok, xAI, Colossus, Picloram, Dow Chemical, EPA, National Cancer Institute, The Cancer Genome Atlas (TCGA)
马斯克旗下SpaceX同意以600亿美元收购AI编程初创公司Cursor
在预期的2万亿美元IPO前夕，SpaceX正利用其庞大的xAI算力基础设施吸纳Cursor的营收增长，而与此同时，激增的Token消耗量和私募股权债务违约正在威胁传统SaaS商业模式。
几位主持人深入分析了SpaceX以600亿美元结构化收购AI编程平台Cursor的交易，以及该交易与xAI的Colossus集群的战略协同。随后他们转向SaaS市场回调话题，探讨了Thoma Bravo在Medallia项目上的失利、向无头代理软件的转变以及风险债务的潜在风险。在评估了Tim Cook的功绩遗产以及John Ternus被任命为Apple首席执行官之后，嘉宾们剖析了联邦大陪审团因秘密线人资金问题对SPLC提起的起诉，最后在“科学角”环节解读了早发性结肠癌病例激增与除草剂毒莠定（picloram）之间的关联。
David Sacks 因在白宫与特朗普总统举行临时会面而推迟播客录制，他透露特朗普思维敏锐并专注于推动 AI 基础设施建设，极力支持科技公司采用表后独立发电以打破电网对数据中心的限制。 (0:00)
SpaceX 达成协议计划在 2026 年底前以 600 亿美元估值收购 AI 编程平台 Cursor，并附带 100 亿美元解约费条款；Cursor 的 ARR 预计将在年底冲至 60 亿美元，此交易将整合 xAI 的 Colossus 算力集群并助力 SpaceX 冲刺 2 万亿美元估值 IPO。 (4:55)
私募巨头 Thoma Bravo 因年度偿债成本从 1 亿美元飙升至 3 亿美元，将软件公司 Medallia 拱手移交债权人；随着传统按席位付费模式被 Headless AI 架构和 API Token 颠覆，杠杆过高的企业级 SaaS 行业正面临普遍的债务与估值危机。 (18:32)
Tim Cook 正式卸任苹果 CEO，其 15 年任期实现了市值逾 10 倍扩张并注销了近半数流通股；长期负责硬件工程的资深老将 John Ternus 接棒上任，面临如何在下一代硬件阵列中重构跨平台无缝 AI 体验的艰巨挑战。 (46:20)
Reading summary
The All-In hosts evaluate artificial intelligence's disruption of enterprise technology, examining SpaceX’s proposed $60 billion acquisition of Cursor, structural repricing across SaaS models, and Apple’s CEO transition, alongside discussions on non-profit accountability and new environmental cancer research.
SpaceX's Cursor Deal and Developer Platform Consolidation (4:55)
SpaceX structured an agreement to acquire AI coding platform Cursor by late 2026 at a $60 billion valuation, with a $10 billion breakup fee, as Cursor projects annual recurring revenue to grow from $2 billion to $6 billion. Chamath Palihapitiya noted the delayed structure protects SpaceX’s $2 trillion IPO schedule while absorbing excess compute from xAI’s Memphis Colossus cluster, which is expanding from 550,000 toward 1 million GPUs. As foundation model generalists like OpenAI and Anthropic build native coding tools, standalone software wrappers face margin compression. Acquiring Cursor secures proprietary reinforcement learning data and millions of developer workflows for xAI's models like Composer 2 and Grok, while providing Cursor with compute independence.
SaaS Margin Compression and Enterprise Debt Vulnerabilities (18:32)
Enterprise software is experiencing severe margin pressure as AI agents and headless architectures displace per-seat licensing. Enterprise buyers are shifting to direct API connections like Model Context Protocol (MCP), slashing required seat counts—such as dropping dozens of Workday seats to two integration bridges—and pushing for contract renewal cuts of 50% to 75%. Public enterprise software has dropped substantially over six months, including ServiceNow down 54%, Snowflake down 43%, and Salesforce down 32%.
This price compression exposes heavily levered balance sheets. Private equity firm Thoma Bravo handed Medallia to creditors after debt servicing costs tripled from $100 million to $300 million annually amid an 82% sales target shortfall. David Sacks and Chamath cautioned that venture debt introduces severe fragility for startups, as low-yield debt lenders cannot absorb total losses and will restrict operational flexibility during market shifts.
Apple's CEO Transition and Strategic Challenges (46:20)
With Tim Cook stepping down, Apple appointed 25-year hardware veteran John Ternus as CEO. Sacks and Chamath defended Cook’s 15-year tenure as an operational success, noting Apple expanded revenue from $100 billion to over $400 billion, grew market capitalization tenfold, and reduced its share count by nearly 50% through aggressive repurchases while transitioning to custom Apple Silicon.
However, the hosts argued Apple has fallen behind in emerging categories, missing opportunities in smart glasses, autonomous driving, and conversational AI search. David Friedberg emphasized that Ternus must prioritize a unified, context-aware AI operating layer spanning all Apple hardware. Chamath warned that relying on the high-margin iPhone ecosystem creates vulnerability as multimodal AI agents and open protocols decouple computing from smartphones.
SPLC Federal Indictment and Non-Profit Incentives (60:32)
A federal grand jury indicted the Southern Poverty Law Center (SPLC) on 11 counts of wire fraud and money laundering regarding $822 million in offshore accounts. The indictment alleges the SPLC directed over $3 million via shell entities to extremist group leaders and confidential informants, including paying informant F-37 over $270,000 while they actively organized logistics and posted racist material around the 2017 Charlottesville rally. Sacks argued the organization manufactured racial conflict to drive fundraising, which jumped from $58 million in 2016 to $136 million post-Charlottesville. Chamath and Friedberg criticized modern non-profits, arguing that 501(c)(3) tax exemptions sustain entities whose survival relies on perpetuating and publicizing crises rather than solving them.
Epigenomic Evidence Links Picloram to Colon Cancer (79:03)
David Friedberg highlighted a Barcelona study investigating the 80%-plus rise in early-onset colorectal cancer among individuals under 50 over the past 20 years. Analyzing tumor gene expression profiles from the National Cancer Institute's Cancer Genome Atlas, researchers isolated exposure to picloram—a persistent synthetic auxin herbicide developed in 1963 for weed control in pastures and utility corridors—after controlling for diet, obesity, and alcohol intake. The chemical resists biodegradation and leaches into groundwater, yet the EPA has not conducted a comprehensive safety evaluation on it since 1995. County application data correlated with up to a 3x higher incidence of early-onset colon cancer, leading Friedberg to call for an immediate EPA review and mandatory epigenomic toxicity screening.
Across enterprise software, Big Tech leadership, non-profit institutions, and chemical regulation, the discussions highlight how existing business and regulatory models are strained by artificial intelligence, structural leverage, and newly surfaced empirical data.
## Background
本期节目由知名商业与科技投资播客《All-In Podcast》于 2026 年 4 月 24 日发布在 YouTube 平台。节目简介显示，四位主持人本期重点探讨了 SpaceX 拟以 600 亿美元收购 Cursor 的重磅交易、SaaS 行业的杠杆债务风暴与估值崩塌、苹果新 CEO John Ternus 接棒 Tim Cook、南方贫困法律中心（SPLC）遭刑事起诉，以及青年结直肠癌激增潜在诱因的最新科研成果。
## TL;DR
四位主持人首先深入拆解了 SpaceX 拟以 600 亿美元收购 Cursor 的交易结构，指出马斯克正借助高估值预期与闲置算力整合应用层资产，以此赋能 SpaceX 的万亿美元 IPO 叙事。随后，针对 Thoma Bravo 放弃 Medallia 控制权的爆仓事件，节目深度剖析了 AI 智能体与无头架构对传统 SaaS 按席位付费模式的毁灭性冲击，并警告初创公司警惕风险债陷阱。在政治与社会议题中，主持团队反思了库克时代的资本运作与新任 CEO Ternus 面临的硬件挑战，严厉声讨了 SPLC 涉嫌制造仇恨收割捐款的 NGO 乱象。最后，科学角重点解读了除草剂毒莠定（Picloram）通过表观遗传学途径诱发青年结直肠癌的开创性研究。
## Atmosphere & style
整期节目展现出顶尖投资人圈子独特的对话张力，将冷峻严谨的商业财务解构与毫无顾忌的死党互损完美融合。四位主持人一边拿白宫会晤、防身武器以及“天王星”谐音梗大开玩笑，一边在探讨杠杆违约与加州体制腐败时流露出强烈的愤怒与警惕。从资本市场的残酷算计切换到对癌症病理的硬核科普，全场节奏紧凑、信息量极大，在辛辣讽刺与智性交锋中带给听众兼具洞察力与娱乐性的现场听感。
## Bulletin
- [04:55](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=295s) SpaceX 计划以 600 亿美元估值收购 AI 编程独角兽 Cursor，若交易未果将支付 100 亿美元合作/分手费；此举可将 Cursor 预计 2026 年底达 60 亿美元的 ARR 现金流与马斯克的超级算力集群结合，同时规避打乱 SpaceX 推进 2 万亿美元估值 IPO 的 S-1 申报。
- [18:32](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=1112s) 私募股权巨头 Thoma Bravo 拟将 2021 年以 64 亿美元私有化的客户体验 SaaS 龙头 Medallia 移交给债权人，51 亿美元股权价值面临清零；高额偿债成本与企业转向自建 AI 智能体导致其销售完成率暴跌至 18%，标志着传统杠杆收购 SaaS 模式出现严重裂痕。
- [46:19](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=2779s) 苹果 CEO 库克宣布卸任并由掌舵硬件工程 25 年的约翰·特努斯（John Ternus）接任；库克任内带领苹果市值膨胀逾 10 倍并将股本回购注销近半，但新管理层正面临 iPhone 护城河在跨设备 AI 智能体时代被瓦解的转型大考。
- [01:00:32](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=3632s) 南方贫困法律中心（SPLC）遭联邦大陪审团以电汇欺诈和洗钱起诉，指控其在 2014 至 2023 年间挪用超 300 万美元秘密资助包括三K党与夏洛茨维尔集会策划者在内的极端分子，涉嫌自导自演激化社会矛盾以拉动数以亿计的捐款。
- [01:19:03](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=4743s) 西班牙科研团队基于癌症基因组图谱（TCGA）与 RNA 测序研究发现，陶氏化学于 1963 年研发的难降解除草剂毒莠定（Picloram）在表观遗传学层面是驱动 50 岁以下人群结直肠癌发病率激增超 80% 的潜在元凶，高暴露地区患癌优势比约为 3 倍。
## Key takeaways
- [07:22](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=442s) SpaceX 拟以 600 亿美元全资收购 Cursor，通过高溢价兜底协议锁定顶尖代码强化学习数据与商业化现金流，同时避免直接并购审计导致 SpaceX 的 S-1 招股书失效重写。
- [07:22](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=442s) 马斯克将 Colossus 集群上闲置的 GPU 算力注入给极度缺乏底层基础设施的 Cursor，用未来高估值股票支付对价，完成了一次极具算力与资本协同效应的兼并。
- [09:41](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=581s) 底层大模型厂商向下渗透蚕食工具链（如 Claude Code），倒逼独立应用层企业必须向算力基础设施巨头深度捆绑以寻求生存。
- [11:43](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=703s) 企业内部泛滥的低效 AI 智能体引发数据冗余与 Token 账单膨胀，迫使 IT 部门依赖具备完备工程与治理能力的成熟前端 IDE 进行统合管控。
- [21:12](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=1272s) Medallia 暴雷揭示了垂直 SaaS 行业危机：企业无需再采购昂贵的通用软件，转而让内部团队指示 AI 生成定制化轻量 Agent，直接瓦解了传统 SaaS 高净收入留存率（NRR）的神话。
- [23:55](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=1435s) AI 带来的生产力爆发正在实体经济中兑现通缩红利，企业砍掉高额外部软件预算释放出再投资资本，但依赖高负债运转的传统 SaaS 杠杆部门却因此面临集中违约风险。
- [29:54](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=1794s) 模型上下文协议（MCP）与自主智能体推动软件交付走向“无头化”，企业只需极少数数据库读写接口席位即可运作，彻底摧毁了传统 SaaS 赖以生存的人头订阅（Per-seat）扩张模式。
- [32:31](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=1951s) 如果市场共识认为 AI 将在数年后颠覆现有 SaaS 商业模式，折现现金流（DCF）模型中的终值将被强制归零，哪怕公司现金流充沛，其估值乘数也将面临折半压缩。
- [36:43](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=2203s) 风险借贷（Venture Debt）是初创公司的致命毒药，其严苛的财务契约会剥夺团队面对技术颠覆时的机动调整空间，在下行期更易遭遇债权机构的掠夺式抽贷。
- [39:06](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=2346s) 传统公共部门的确定给付型养老金计划（Defined-benefit）在公会推动下固守旧制，员工临退休虚报加班拉高基数，正演化为各级政府财政的定时炸弹。
- [41:37](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=2497s) 主流媒体对体制内腐败放弃了深入调查，而独立公民记者凭借实地走访揭露加州游民工业复合体等系统性财政浪费，治理核心应转向清除公共系统 20%-30% 的损耗。
- [46:19](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=2779s) 库克在苹果任内展现了极致的守成与资本配置能力，成功去除了对 Intel 芯片的依赖并常态化回购股票，但产品线日益沦为利润收割机器，错失了智能眼镜等前沿硬件爆发点。
- [57:46](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=3466s) 特努斯接任苹果 CEO 面临类似华特·迪士尼去世后的代际考验，未来异构化 AI 硬件百花齐放，苹果亟需走出过度依赖单一 iPhone 盈利的舒适区，建立统一感知的个人 AI 体验层。
- [01:04:23](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=3863s) 联邦大陪审团指控 SPLC 设立秘密账户向极端仇恨组织高层输送资金自导自演激化社会矛盾，以此在夏洛茨维尔事件后制造恐慌并换取高达 8100 万美元的筹款激增。
- [01:15:01](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=4501s) 非营利组织（NGO）缺乏商业公司的市场倒闭出清机制，为维持生存与募资不断“移动球门”，从追求宪法权利与机会平等异化为追求结果均等，甚至演化为反向制造社会危机的吸金机器。
- [01:21:10](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=4870s) 表观遗传学分析表明，环境化学物质低剂量长期暴露会篡改细胞转录开关；EPA 自 1995 年后未再系统评估除草剂毒莠定，突显出现行化学品毒理审查体系与前沿分子生物学研究的严重脱节。
- [01:25:31](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=5131s) 青年早发性癌症诱因的突破性确认，主要归功于联邦政府长期资助维护的 TCGA 基因组生物样本库与 RNA 测序底座，印证了国家级基础科研基础设施不可替代的公共价值。
## Detailed walkthrough
### [00:00](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=0s) 节目开场与“全副武装”的 JCal
节目录制伊始，几位主持人就互相打趣调侃。大家拿 Jason Calacanis（JCal）开涮，称他恰好处于当前热议话题与南方贫困法律中心（SPLC）卷入风波的交汇点上，“紧邻副总统，又紧挨着 SPLC”。JCal 笑着反唇相讥，感谢大家把他推到了舆论的风口浪尖。面对兄弟们的起哄，JCal 更是当场放狠话，戏称自己随时备好了弹药，直接掏出防身家伙，直言“这就是为什么我带着细头弹簧刀（stiletto）和 P35 手枪”，放话“要是谁想翻过围栏找麻烦，尽管来试试，JCal 早就准备好了”，引得全场爆笑。
### [01:02](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=62s) Sacks 白宫会晤特朗普与两党 AI 路线之争
在正式进入第 270 期节目的常规介绍后，话题迅速转向刚刚让全剧组推迟录制一小时的 David Sacks。作为华盛顿特区近期的常客，Sacks 仅凭一条“正与总统在一起，推迟录制”的短信就“耍了大牌”。Sacks 解释称，自己当天在白宫办事，顺便询问总统是否有空，特朗普欣然抽出了时间进行简短闭门会晤。
谈及与特朗普的接触，Sacks 驳斥了主流媒体将特朗普刻画为喜怒无常、动辄对下属咆哮的形象，坚称特朗普在私下交流中始终非常和蔼、充满魅力（"charming AF"），且对前沿议题充满好奇并乐于提问。
Sacks 深入对比了两届政府在人工智能领域的政策分歧：
- **拜登政府路线**：被指充满“末日论者的神经质”（doomer neuroses），通过行政命令强推审查制度与 DEI（多元、平等、包容）价值观，过度监管导致数据中心审批滞后甚至面临停滞。
- **特朗普政府路线**：重在确保美国在科技竞争中取胜。特朗普早在一年多前、在数据中心成为政治焦点之前，就率先提出允许 AI 公司直接自建“表后发电”（behind-the-meter）设施，而非采取伯尼·桑德斯（Bernie Sanders）式的全面关停与严监管。
- **关于 Anthropic 的插曲**：在谈及近期风波不断的 Anthropic 团队时，特朗普私下评价他们是“绝顶聪明的家伙”，研发出了出色的产品，虽然政治立场极度偏左、甚至曾试图指挥五角大楼并遭到军方反弹，但这些分歧依然属于“可以协作解决”的范畴。Sacks 强调，特朗普骨子里极度看重并偏爱高智商人群（high-IQ people）。
### [04:55](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=295s) 重磅交易：SpaceX 拟以 600 亿美元收购 Cursor
随后，主持人切入本期核心议题：SpaceX 与 AI 编程领军独角兽 Cursor 达成的轰动性重磅交易。双方不仅计划联合打造号称“全球最强编程与知识工作 AI”的全新模型，交易结构更是引发市场高度关注：
- **交易条款细节**：SpaceX 承诺将在 2026 年底前以高达 600 亿美元的估值全资收购 Cursor（相较 Cursor 此前 500 亿美元的融资传闻溢价 100 亿美元）；若收购未成，SpaceX 将直接向 Cursor 支付 100 亿美元的合作费用，彭博社等媒体普遍将这 100 亿美元视作变相的“分手费”，意味着收购几乎已成定局。
- **Cursor 惊人的吸金速度**：截至 2026 年 2 月底，Cursor 的年化经常性收入（ARR）已达 20 亿美元，预计到 2026 年底将翻两番飙升至 60 亿美元，成为名副其实的“印钞机”。虽然起步阶段主要调用 Anthropic 的底层大模型，但 Cursor 在 3 月推出的自研专有模型 Composer 2 性能强劲，在榜单上已稳稳居于 GPT-4.5 与 Claude Opus 之间。
- **赋能 SpaceX IPO 叙事**：SpaceX 预计 2026 年自身营收在 220 亿至 240 亿美元之间，并正锚定 2 万亿美元的 IPO 估值上市。Cursor 的超高毛利与强劲现金流并表，将极大丰富 SpaceX 的营收增长故事。
- **算力互补优势**：埃隆·马斯克旗下的 Colossus 超级计算集群已拥有 55 万张 GPU，并正在向 100 万张扩容，未来还将探索太空数据中心。Cursor 长期受制于算力瓶颈，而这次联手恰似“花生酱碰上巧克力”，JCal 大胆预测，SpaceX/xAI 与 Cursor 的联合体将在未来 12 个月内登顶全球 AI 编程排行榜榜首。
### [07:22](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=442s) Chamath 拆解交易结构：S-1 规避重写与算力资本运作
作为 SpaceX 的老股东，Chamath 从资本运作与底层架构的视角对这笔交易进行了深入解构：
```mermaid
flowchart TD
accTitle: SpaceX 与 Cursor 交易与算力协同结构图
accDescr: 展示 SpaceX 利用 2 万亿美元 IPO 估值股票、结余算力与 Cursor 团队及商业化现金流的互补整合模式。
subgraph SpaceX_xAI["SpaceX 与 xAI 资源"]
A["2 万亿美元 IPO 估值预期<br/>(股票作为高溢价支付对价)"]
B["Colossus 集群 55 万-100 万 GPU<br/>(Grok 早期降价潮退去后的闲置算力)"]
end
subgraph Cursor_Assets["Cursor 核心资产"]
C["20 亿-60 亿美元 ARR 强劲现金流"]
D["顶尖团队 + 强化学习代码数据闭环"]
end
A -->|"以未来高估值股票支付<br/>实际成本折半 (相当于现值 300 亿)"| C
B -->|"算力注入解渴"| D
C & D --> E["重塑全球第一 AI 编程与知识引擎"]
```
Chamath 指出，之所以采取“收购意向结合 100 亿美元兜底分手费”的独特模式，核心是为了**保护 SpaceX 正在推进的 S-1 招股书不至于过期**。如果直接启动复杂的跨行业并购审计，公司必须全盘重写披露文件、修改风险声明，这会彻底拖垮 IPO 节奏。在事实层面，这笔交易已板上钉钉。
从财务杠杆来看，这堪称马斯克的又一次“商业柔术”神作：
- **估值折价套利**：SpaceX 目前一级市场估值约 1 万亿美元，若以上市后预期的 2 万亿美元估值发行股票，用“明天的美元”支付 600 亿美元对价，折算回现值仅相当于 300 亿美元，实际相当于打了五折。
- **闲置算力盘活**：此前 Grok 曾在模型输出端推行极其激进的代币（token）价格补贴，吸引了大量试水用户，但随着补贴效应衰减，Colossus 集群内出现了利用率偏低的多余算力。马斯克顺势将这些闲置算力注入给极度缺乏基础设施的 Cursor，一举将 AI 领域最火爆、变现路径最直接的应用层“包装器”（wrapper）收入麾下，并锁定了最具价值的代码强化学习数据与顶尖算法工程团队。
### [09:41](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=581s) 政治围剿与太空算力：SpaceX 与 Cursor 的互补整合
谈话首先回顾了拜登政府时期对马斯克的法律围剿（lawfare）。司法部曾起诉 SpaceX，指控其未雇佣庇护寻求者，但由于国际武器贸易条例（ITAR）的严格限制，SpaceX 在法律上根本不能雇用非美国公民。如今 SpaceX、xAI 与 Tesla 已经站在人类工程的最前沿，正在构建太空算力集群与 Optimus 人形机器人等颠覆性产品。
在这一背景下，SpaceX 计划以 600 亿美元收购 Cursor 的逻辑变得尤为清晰。Cursor 在 AI 编程领域拥有无可匹敌的统治力，而编程正是当下最炙手可热的赛道。但 Cursor 此前面临严重瓶颈：底层极度依赖通用模型厂商（如 OpenAI 与 Anthropic），而这些厂商已开始向下垂直整合，推出 Claude Code 与 OpenAI Codex 等竞品。通过融入马斯克的技术生态，Cursor 获得了顶级算力底座，而 xAI 则借此补齐了海量的工程训练数据、企业级客户以及顶尖的代码交互经验。
```mermaid
flowchart TD
accTitle: SpaceX 与 Cursor 整合架构
accDescr: 展示 SpaceX/xAI 与 Cursor 结合前后的资源与能力互补结构
subgraph xAI_SpaceX["xAI / SpaceX 算力底座"]
A["超级算力集群"]
B["Grok 基础大模型"]
C["数据与网络基础设施"]
end
subgraph Cursor_Layer["Cursor 开发者入口"]
D["原生 IDE 交互层与 UX"]
E["高粘性开发者与企业客户"]
F["海量真实代码训练数据"]
end
Cursor_Layer -->|"提供工程数据与企业分发"| xAI_SpaceX
xAI_SpaceX -->|"注入庞大算力与自研模型支持"| Cursor_Layer
```
- **司法诉讼回顾**：美国司法部曾指控 SpaceX 歧视庇护寻求者，但 ITAR 法规严格禁止向非美国公民透露敏感火箭发射技术。
- **竞争环境恶化**：通用大模型提供商正在加速吞噬上层工具链，促使独立应用层厂商寻求与算力基础设施巨头深度捆绑。
---
### [11:43](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=703s) IDE 护城河：智能体泛滥时代的软件工程治理
针对 Cursor 是否会放弃诸如 Kimi K2.6 或 Moonshot 等第三方底层模型的疑问，几位主持人指出，维持模型选择权是 Cursor 用户体验的核心竞争力。Cursor 真正的壁垒不仅是调用何种底层模型，而是其针对开发者打磨的 IDE 交互层与应用层框架，在整体用户体验上远胜于单纯的模型集成。
在过去的 120 天里，科技行业逐渐意识到，AI 最核心的价值依然是由“写软件”来实现的。虽然市场追捧所谓的“智能体（Agents）”，但智能体本质上就是快速生成出来的轻量应用。这种狂热正在引发严重反噬：企业内部生成了成千上万个低效智能体，导致冗余的数据创建、重复的存储开销与大量的 API 资源浪费（例如亚马逊内部曾被曝光类似乱象）。企业迫切需要具备扎实软件工程能力的人员来构建安全架构与管理框架，而集成 AI 能力的标准前端 IDE 正是这场架构治理的核心载体。
- **Token 账单膨胀**：企业内部由于代码生成量激增，Token 支出呈指数级上升，亟需高效的模型分流中间件。
- **模型分级路由**：未来仅有最前沿的复杂推理任务才需要调用昂贵的前沿模型，日常基础代码将由开源或小模型承接。
---
### [14:53](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=893s) 网络安全大模型竞赛与收购交易细节
网络安全正成为 AI 编程领域中最关键的杀手级应用。Mythos 等前沿模型的出现向外界展示了 AI 在网络攻防两端的破坏力与防御潜力。然而，像拥有约 10 万亿参数的大模型不仅运行极其昂贵，就连 Anthropic 等巨头也缺乏足够的算力进行全面商用托管。行业未来的发力点将集中在具有类似能力、但更专精且推理成本更低的网络安全垂直模型上，以满足首席信息安全官（CISO）和企业 IT 部门应对 AI 黑客攻击的紧迫需求。
转向收购细节，Polymarket 数据显示 SpaceX 收购 Cursor 的概率高达 74%，且 SpaceX 在 8 月底前完成 IPO 的概率达 80%。这笔交易不仅能防止 IPO 进程被打乱，还设置了 1000 万美元的分手费机制，促使 Cursor 团队全力交付。此前马斯克曾在社交平台上直言：“xAI 第一次构建得并不正确，因此目前正在从底层基础进行全面重构（xAI was not built right first time around, so is being rebuilt from the foundations up）。”收购 Cursor 正是这一人才与技术架构重组的重要一步。与此同时，随着 Walter Isaacson 等人公开讨论 Tesla 与 SpaceX 合并的可能性，马斯克麾下各业务线的整合协同趋势愈发明显。
- **预测市场数据**：Polymarket 预测 SpaceX 收购 Cursor 的胜率为 74%，SpaceX 8 月底前 IPO 概率为 80%。
- **算力成本绑定**：Cursor 若脱离 SpaceX 独立运行，将不得不按天价自费购买大模型算力，反向强化了并入生态的动力。
- **推文影响力**：马斯克坦诚重构 xAI 的单条动态获得了超过 5080 万次浏览量。
---
### [18:32](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=1112s) 私募股权的 SaaS 债务炸弹：Medallia 爆仓先兆
节目随后切入下一个核心重磅议题：私募股权行业是否即将迎来 SaaS 债务炸弹？知名私募巨头 Thoma Bravo 正接近达成协议，将其旗下的客户体验管理软件公司 Medallia 正式移交给债权人。
这一事件被视作私有化 SaaS 泡沫破裂的标志性里程碑。Thoma Bravo 曾于 2021 年市场估值最高峰时期，以 64 亿美元全现金将 Medallia 私有化。随着高利率环境持续、企业软件支出紧缩以及生成式 AI 对传统 SaaS 商业模式的冲击，依靠高杠杆运转的软件并购基金开始面临前所未有的偿债危机与违约风暴。
- **标的背景**：Medallia 为客户体验（CX）领域的 SaaS 龙头企业。
- **交易规模**：Thoma Bravo 于 2021 年高点以 64 亿美元全现金收购，目前由于无法支撑庞大债务即将交出控制权。
### [19:00](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=1140s) Thoma Bravo 与 Medallia 暴雷：SaaS 债务炸弹引爆
私募巨头 Thoma Bravo 在收购客户体验管理平台 Medallia 的交易中背负了高达 30 亿美元的债务。回溯至 2021 年，Medallia 当年的年收入尚有 4.7 亿美元，并保持着 20% 的年化增速。然而，彭博社近日披露，Thoma Bravo 为 Medallia 承担的债务偿付成本即将翻倍暴增，从原先的每年 1 亿美元直接飙升至 3 亿美元。面对这一困境，包括 Blackstone（黑石集团）在内的债权方均拒绝向这家步履维艰的 SaaS 公司施以援手。最终，Thoma Bravo 疑似直接“交出钥匙”撒手不管，导致 51 亿美元的股权价值彻底清零，沦为一场公开的资本惨败。
面对主持人抛出关于 SaaS 行业逆风的提问，Chamath 提出了耐人寻味的逆向视角。他认为 Thoma Bravo 是一家运营极佳的机构，联合创始人 Orlando Bravo 也是极其顶尖的投资人，其基金回报率长期处于“狂暴”状态：
- **股息资本重组（Dividend Recap）**：Chamath 推测 Thoma Bravo 在过去持有该资产的五年中，极有可能通过至少一到两次的杠杆分红拿回了全部或绝大部分本金，甚至已经实现了正回报。
- **私募股权（PE）的不良资产逻辑**：与风险投资寻找“白纸一张、不惜代价高增长”的初创模式不同，PE 的标准打法原本就是低价接盘陷入困境的资产，并派驻强大的运营合伙人团队进驻拯救。
- **核心溃烂与断臂求生**：一旦选择直接交钥匙认赔出局，往往说明 Medallia 内部存在无法修补的“核心溃烂”，加上基金自身的下行风险早已通过套现得到充分对冲，因此斩仓走人对其影响有限。真正的受害者是蒙受巨大账面损失的债券持有人，这虽可能推高 Thoma Bravo 未来交易的借贷成本，但绝非外界想象的血本无归。
```mermaid
flowchart TD
accTitle: Medallia债务违约与资本结构崩溃流程
accDescr: 展示了从高杠杆并购、利息飙升到股权清零、债权受损的传导过程
subgraph Acquisition["2021 杠杆收购"]
A["Medallia 资产<br/>4.7亿美元营收 / 20%增速"] --> B["Thoma Bravo 私有化<br/>承担 30 亿美元债务"]
end
subgraph Crisis["2026 债务成本飙升"]
B --> C["年偿债利息激增<br/>1亿美元/年 ➜ 3亿美元/年"]
C --> D["黑石等债权人<br/>拒绝提供流动性救助"]
end
subgraph Outcome["资本爆仓与资产重组"]
D --> E["Thoma Bravo 放弃控制权<br/>抹去 51 亿美元股权价值"]
E --> F["债券持有人 & CLO 遭受重创<br/>企业控制权移交债权方"]
end
```
---
### [21:12](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=1272s) AI 智能体背刺垂直 SaaS：定制化替代买软件
尽管 Chamath 对 PE 的防守策略抱有信心，但 Friedberg 与其他几位 Bestie 指出，Medallia 乃至全行业 SaaS 公司的底层基本面正在发生根本性坍塌。据社交平台披露的内部运营数据显示，Medallia 的销售团队业绩惨淡，销售完成率仅达到预设目标的 18% 左右。
Medallia 赖以生存的核心业务是“客户体验管理”（Customer Experience Management），典型应用场景例如在加勒比海游轮之旅结束后，向游客发送详尽的满意度调研，再通过算法将反馈数据转化为运营见解提供给管理层，帮助其改进服务质量。然而，这种传统的“反馈闭环”在生成式 AI 时代遭遇了致命打击：
- **从采买软件转向自建智能体**：面对同样的工作流，企业如今无需再为昂贵的通用 SaaS 支付持续订阅费，只需让内部团队指示 AI 快速搭建一个轻量级 Agent。不仅成本微乎其微，还能完美契合自身业务的定制化流程。
- **高确定性现金流的神话破灭**：以往 SaaS 公司能够承受高额杠杆融资，核心逻辑在于其可预测的财务模型——高达 118% 的净收入留存率（NRR）意味着无需大幅拉新即可通过老客增购增长 18%，进而轻松撬动杠杆借贷。
- **拉新冻结与高流失率共振**：由于企业采购决策发生根本转变，传统 SaaS 在新客获取上寸步难行，老客流失率更远超财务模型预测。债务违约和 CLO（贷款抵押凭证）在二级市场的折价交易，正是这批高杠杆软件公司造血能力中断的直接反映。
---
### [23:55](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=1435s) 凯文·沃什听证会与 AI 通缩：杠杆破裂与经济扩张的两面性
Friedberg 将 SaaS 行业的“血洗惨状”与美联储主席提名人选凯文·沃什（Kevin Warsh）的国会听证会紧密联系在一起。市场往往习惯将“SaaS 末日”视作科技圈内部孤立的商业灾难，但从更宏观的经济学视角来看，这正是 AI 兑现其“通缩红利”的现实投射——企业运营所必须支付的技术溢价正在以惊人的速度被击穿。
根据沃什在听证会上的阐述，AI 驱动的通缩演进正在带来前所未见的生产力大爆发，但其对实体经济和金融体系将带来复杂的两极分化影响：
- **生产力飙升与劳动力错配**：AI 大幅提升了单位生产效率，但在生产力转化为实际收益与劳动力市场适应吸收之间，将存在不可忽视的断层与阵痛。
- **高杠杆部门的连锁违约**：通缩的残酷一面在于它直接导致特定行业的收入萎缩。当一个长期建立在“永远高增长”假设并背负高杠杆的经济板块（如 SaaS 行业）突然遭遇营收负增长，其债务将迅速恶化并爆仓，可能诱发不良的金融外溢效应。
- **降本释放的资本再投资**：与萧条时期的经济收缩不同，由技术进步推动的通缩最终将促进宏观经济扩张。当企业能够砍掉 50% 的外部 SaaS 预算，节省下来的资本即可被重新配置到推动企业真实增长的核心领域，从而带动其他经济部门繁荣。
- **物价衡量方式的彻底失真**：沃什在听证会期间还批评了现行美联储衡量通胀（CPI/PCE）的指标体系，指出无论各类官方指数如何粉饰，普通家庭的实际感受依然是“所有东西都贵得离谱”。这表明旧有的通胀计量方式完全脱离了民众现实，亟待美联储在制定利率政策时予以重构。
---
### [26:28](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=1588s) Token 吞噬软件时代：二级市场暴跌与 PE 并购新变局
Jason 顺势总结了当下软件投资的残酷转折点：**“SaaS 的失血正是算力 Token 贩子的狂欢”**（"SaaS's loss is the token dealer's gain"）。如今最具敏锐度的初创公司早已成为排头兵，不再采购外部工具，而是利用大模型自主编写开发工具与业务看板。在二级市场上，昔日不可一世的头部云软件公司全线崩盘，呈现出触目惊心的跌幅：
- **Salesforce**：近 6 个月内累计暴跌 32%；
- **ServiceNow**：同期腰斩，重挫 54%；
- **Snowflake**：市值大幅回撤 43%；
- **Adobe**：股价跌去 33%；
- **Figma**：在经历了轰动一时的 IPO 爆炒后，距离高点已暴跌 67%。
Sacks 随后从风险投资（VC）与私募股权（PE）的资本循环逻辑切入，深度拆解了软件行业的退出生态。在历史上，软件企业原本仅有两种常规退出渠道：首次公开募股（IPO）与被大公司兼并收购（M&A）。直到过去十余年间，大型 PE 机构大举入场，才为创投圈开辟了“第三条退出路径”。这套通常由三分之一股权搭配三分之二高杠杆债务的 LBO（杠杆收购）模式，此前长期运转于传统非科技实体领域，然而当这套依赖高额债务托底的金融工具全盘套用在如今正被 AI 剧烈颠覆的 SaaS 行业时，其赖以运转的底层资产正面临前所未有的范式震荡。
### [27:55](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=1675s) 私募股权的 SaaS 杠杆困境与估值崩塌
Sacks 剖析了私募股权（PE）在科技投资中的基本逻辑。进行杠杆收购（LBO）的核心前提在于标的必须具备高度稳定且可预测的现金流；一旦现金流不及预期导致无法支付债务利息，债权人便会行使质权接管资产，股权价值将瞬间归零。在过去相当长一段时间里，市场普遍坚信成熟期软件公司（哪怕是具备 IPO 潜力的成熟标的）天然具备这种铁板钉钉的现金流稳定性。然而，当前公开市场与一级市场的 SaaS 估值体系已经发生剧烈分化：虽然出现了一批年经常性收入（ARR）达到 10 亿美元、年增速 20%、毛利率高达 80% 的成熟 SaaS 龙头，估值倍数却从历史上的 13 倍 ARR 暴跌至仅剩 3 倍，甚至被戏称为“用五毛钱买一块钱资产”。
面对 Jason 关于“这是否意味着 SaaS 已经见底、可以展开兼并整合（Roll-up）”的提问，Sacks 直言自己内心极为矛盾。单纯从定价角度看，当前估值确实极具吸引力，但在私募股权巨头 Thoma Bravo 将陷入困境的 Medallia 拱手转让给债权人的重磅消息曝出后，市场恐慌蔓延，Salesforce 单日暴跌 9%。Sacks 强调，PE 商业模式能够成立的前提是现金流可预测，但当下 SaaS 的底座正被 AI 彻底动摇——企业软件过去的净收入留存率（NDR）能维持在 120%，如今短短半年到一年内就可能跌落至 80%，因为海量客户正在流失，转向按 Token 计费的新型方案，或是利用 AI 自动生成定制软件。
```mermaid
flowchart TD
accTitle: PE杠杆收购在AI时代的逻辑断裂
accDescr: 展现传统PE依赖稳定现金流加杠杆的闭环被AI原生定制与Token按需消费打破的过程
subgraph 传统闭环["传统 SaaS LBO 模式"]
A["高毛利 + 高粘性"] --> B["高净收入留存 (NDR > 120%)"]
B --> C["高度可预测现金流"]
C --> D["加注巨额杠杆收购 (PE)"]
end
subgraph AI冲击["AI 时代的断裂冲击"]
E["Token 计费与定制软件"] --> F["客户流失与 NDR 骤降 (< 80%)"]
F --> G["现金流不及预期"]
G --> H["债务违约与股权清零"]
end
D -.->|遭遇技术拐点| H
```
---
### [29:54](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=1794s) 定价模式的扭曲与 MCP/Agent 对席位制的毁灭打击
Chamath 深入剖析了 SaaS 公司如何逐步落入自己编织的定价陷阱。初创公司起步时通常采用破坏性定价打破垄断，秉持“提供 10 美元价值只收取 1 美元费用”的 10% 价值比原则；然而，随着企业逐轮堆叠风险投资（VC）和成长型股权（Growth Equity），清算优先权（Liquidation Preference）层层累加，公司背负着 3 亿、5 亿甚至 10 亿美元的优先清偿要求，还必须额外向投资人交付 15% 到 20% 的回报门槛。为了交差，管理层唯一的手段就是不断提价。当私募股权机构作为接盘的最后一棒进场，注资数十亿美元股权与杠杆债时，为了偿债他们绝不可能降价抢占市场，只能继续推高单价。这导致大量所谓的十亿美元级 SaaS 巨头其产品单位成本严重失控，定价占客户所得价值的比重已从原本的 10% 飙升至 30% 以上，不仅催生出无数竞争对手，更在自己背后画上了巨大的靶子，迫使客户在合同到期时扬言砍掉 50%、甚至 75% 的预算。
更具毁灭性的是软件交付形态的质变。随着智能体（Agent）和模型上下文协议（MCP, Model Context Protocol）的普及，产品正在走向“无头化”（Headless）。以企业人力系统 Workday 为例，以往企业需要为 50 名员工购买 50 个软件席位，但未来可能只需要购买 2 个席位作为 AI 智能体读写 Workday 数据库的接口，其余所有日常操作均由智能体代劳。如果将这一席位锐减效应放大到全市场上百万家企业，传统 SaaS 赖以生存的按人头订阅模式（Per-seat license）将遭遇断崖式坍塌。Chamath 断言，当前的行业暴跌正是因为单位成本与客户真实价值严重脱节，如果不彻底重置定价体系或被廉价新品替代，软件行业就无法迎来真正的出清。
- **优先权累积的恶性循环**：巨额融资金额带来刚性回报门槛，倒逼企业不断提价，丧失降价防御的价格弹性。
- **价值与定价失衡**：SaaS 软件抽取价值的比例从初创期的 10% 飙升至 30% 以上，成为企业削减预算的首要目标。
- **人头订阅模式瓦解**：无头架构和自主智能体使企业只需购买极少数用于系统读写的“接口席位”，摧毁席位规模扩张逻辑。
---
### [32:31](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=1951s) Salesforce 的“低估值”诱惑与无头架构自救
针对当天的市场异动，Sacks 与 Chamath 梳理了 Salesforce 的核心数据：公司单日大跌 9%，对应 150 亿美元自由现金流（FCF），企业价值（EV）已被压缩至 1400 亿美元左右，估值倍数已不足 10 倍自由现金流。Jason 惊呼这已经进入了典型的“捡便宜货”（Bargain hunting）区间，并询问兼并之王 Marc Benioff 是否会借此机会大举回购自家股票。随后讨论转向了各家 SaaS 巨头应对 AI 冲击的分野：相较于 Workday 试图通过对 AI 交互强行设立“买路钱”收费来固守阵地，Benioff 反其道而行之，火速推出了支持无头化交互的全新产品形态，主动敞开系统拥抱 AI 协议。
然而，Sacks 和 Chamath 警告称，10 倍现金流看似极具吸引力，但其合理性完全取决于传统估值模型的终值假设。如果市场形成共识，认为 AI 将在第 5、6 或 7 年后彻底颠覆现有商业模式，那么在现金流折现（DCF）模型中，第 7 年及以后的终值（Terminal Value）就必须被数学性地直接归零。一旦远期现金流归零，哪怕当前盈利能力再强，公司的估值倍数也理应从 10 倍进一步腰斩至 3 到 5 倍现金流区间。不过，Jason 依然指出，拥有充裕现金流和资金储备的大型巨头——无论是过去的苹果、Meta、谷歌还是 Salesforce，依然握有巨大的战略主动权来抵御周期风暴。
---
### [34:42](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=2082s) 创始人模式与自由现金流的敏捷优势：为何坚决反对创业公司碰风险债
面对 SaaS 行业的大洗牌，讨论聚焦于哪类管理者才能带领公司穿过周期。大家一致推崇由创始人掌舵的敏捷企业：作为自创立起执掌数十年的创始人，Marc Benioff 具备下注所有筹码、“破釜沉舟（Burn the boats）”实施自我革命的魄力；他过去先后借力社交、移动互联网和大数据浪潮完成了业务进化，在 AI 浪潮中成功突围的概率依然极高。相比之下，那些聘请职业经理人打理的成熟企业，往往只会做出试图维持旧体制、对外部 Agent 强征过路费的短视决策。Chamath 甚至指出，只要去翻阅各家公司的 10-K 年报，拆解其人头许可单价的通胀轨迹与成本结构，就能一眼预判出哪些传统软件将最先走向消亡。
在话题收尾阶段，Sacks 对创业公司的资本结构提出了严肃警告，坚决反对初创企业创始人使用风险债（Venture Debt）。他指出，私募股权机构了解金融工程规则并具备成熟的重组手段，但创业者往往错误地将风险债当成普通风险投资，忘记了这笔钱必须按期刚性兑付。更致命的是，风险债包含大量严苛的财务与业务契约（Covenants），一旦背上债务，银行将时刻监督公司财务账目，这会彻底剥夺企业的战略灵活性，使公司在面对类似 AI 这种重大技术变局时根本无法果断调整业务方向。在当下的颠覆性周期中，唯有手握充沛自由现金流、无债务枷锁绑架的公司，才拥有在废墟中生存并完成调头的最高自由度。
- **创始人抗周期优势**：创始人拥有打破旧利益格局、推翻既有商业模式并重新下注的威望与决断力。
- **职业经理人的避险弊端**：倾向于依赖存量收费、阻挠技术整合，加剧产品被替代的风险。
- **风险债的致命脆弱性**：债务契约限制业务转型，把本应灵活自如的初创公司绑死在刚性偿还周期上。
### [36:43](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=2203s) 风险借贷的致命陷阱：为何债务是初创企业的“毒药”
几位主持人深入探讨了债务对企业和个人的毁灭性影响。债务最大的隐患在于剥夺了创始人的机动回旋余地，强制绑定固定的还款周期，使企业在遭遇市场剧烈动荡时变得极度脆弱。在零利率政策（ZIRP）或市场狂热的高峰期，风险借贷（venture debt）机构往往会主动向创始人推销巨额现金。许多创始人误以为这是在不稀释股权的前提下“延长跑道”（extending runway），但萨克斯（David Sacks）和卡拉卡尼斯（Jason Calacanis）指出，对于优质创业公司而言，股权融资永远是更好的选择——股权带来了灵活性，吸纳的是同舟共济、追求长期上涨空间的伙伴；而债务机构的目标完全不同，他们只在乎下行保护。
弗里德伯格（David Friedberg）直言不讳地将风险借贷斥为“硅谷最恶劣的秃鹰业务”（the worst vulture-like business in Silicon Valley）。他表示自己从未见过风险借贷能实质性提升企业质量，反而屡屡看到它给公司留下不可逆的创伤；那些能从风险借贷中赚到钱的机构，往往是因为被投企业根本没去动用那笔钱。银行或债权人绝不愿意成为初创企业账上最后 3 到 6 个月的救命钱，因为这意味着他们极大概率会血本无归。风险投资人（VC）追求的是 10 倍、100 倍乃至 1000 倍的超级回报，因此能够承受大量归零；但借贷机构仅赚取 8% 左右的固定利息，根本无法承担本金全损的风险。一旦企业账面资金见底、只剩下银行贷款，债权人就会毫不留情地“抽毯子”（rug pull）。当债权人感到恐慌时，他们会开出极为苛刻的掠夺性条件：将利率翻倍、索要巨额股权认股权证（warrants），让创始人彻底陷入如同坐牢般的债务困境。
```mermaid
flowchart TD
accTitle: 股权融资与风险借贷在下行期的机制对比
accDescr: 对比展示股权投资人在下行期能够承担风险，而借贷机构为保本会采取抽贷、加息与稀释股权等严苛手段
subgraph Equity["股权融资 (Venture Capital)"]
E1["高风险偏好：博取 10x-1000x 超额收益"] --> E2["允许项目归零，不设固定还款日"]
E2 --> E3["利益一致，赋予创始人机动回旋余地"]
end
subgraph Debt["风险借贷 (Venture Debt)"]
D1["低收益模式：仅获取约 8% 息差"] --> D2["极度厌恶本金全损风险"]
D2 --> D3{"企业资金仅剩 3-6 个月"}
D3 -->|"触发风控抽贷"| D4["翻倍利率 / 索取 Warrants / 强制清算"]
end
```
- **收益与风险错配**：风险投资人通过高倍数回报吸收大量项目失败的成本，而借贷方仅收取约 8% 的票息，无法承受任何归零风险。
- **危机时的价值榨取**：当债权人感知到违约风险时，通常会通过惩罚性加息和索要廉价期权认股权证实施掠夺式自保。
---
### [39:06](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=2346s) 查马斯的4.2亿美金追缴噩梦与政府养老金炸弹
谈到陷入债务的痛苦体验，查马斯（Chamath Palihapitiya）分享了一段险些令他倾家荡产的惨痛亲身经历。他曾违反自己“绝不借债”的铁律，开立了一笔高达 4.2 亿美元的信用额度（credit line）以追求更高的账面收益。然而在一次市场动荡中，作为抵押物的资产市值大幅缩水，引发了可怕的反身性内爆。更要命的是，当时他账上还有数亿美元资金存放在正处于崩溃边缘的瑞信银行（Credit Suisse）。在那个周末，他一边拼命四处腾挪补仓，一边担忧银行破产导致存款归零，这成了他整个职业生涯中最黑暗、最绝望的时刻。查马斯直言“债务等同于坐牢”，并告诫听众巴菲特的名言——聪明人破产的唯一途径就是借债；他更当场发誓如果自己未来再碰杠杆，大家可以直接冲上来“朝我的脸狠狠揍一拳”。
随后，讨论延伸到了公共部门如何将债务“社会化”。弗里德伯格回顾了历史：在 20 世纪 50 年代，美国各大私营企业普遍实行确定给付型养老金计划（Defined-benefit pension plans），但这类计划非旱即涝——要么严重注资不足导致企业破产，要么资金过度充裕无法盘活。因此，美国企业界后来全面转向了 401(k) 等确定缴费型计划（Defined-contribution plans）。然而，政府部门却在公共雇员公会的强力推动下保留了传统养老金模式，这一遗留问题在 70 年后的今天正逐渐演化为引爆全美各级政府财政的定时炸弹。在加州等地，甚至出现了体制内人员在临退休前通过翻倍加班，将薪酬推高至 40 万到 50 万美元，从而锁定极高退休金的荒唐现象，几位主持人一致认为现行公共养老金体系不可持续，亟需向强制性储蓄或 401(k) 模式彻底转型。
- **查马斯的杠杆危机**：动用 4.2 亿美元信用额度加杠杆，遭遇抵押品暴跌与瑞信爆雷的双重绞杀，险些全盘覆没。
- **企业与政府养老金分歧**：私企早在数十年前便摒弃了兜底型养老金并转向 401(k)，而政府在公会游说下固守旧制，形成沉重的财政黑洞。
---
### [41:37](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=2497s) 调查新闻的衰落与尼克·雪莉揭露的加州腐败骗局
萨克斯进一步将矛头对准加州政府运作中充斥的庞大腐败与欺诈。他指出，加州每年在解决游民问题上挥霍高达 120 亿美元，但街头的游民数量却不降反升，形成了尾大不掉的“游民工业复合体”（homeless industrial complex）。为了掩盖这些弊端，加州政客甚至试图通过立法把公众的眼睛和耳朵蒙起来，假装欺诈并不存在。此时，独立公民记者尼克·雪莉（Nick Shirley）凭借其在街头的实地调查，单枪匹马揭露了加州政府内部数十亿美元的贪腐与骗局。主持人探讨了雪莉的运作模式：弗里德伯格等人认为如果走吹哨人诉讼（whistleblower awards）路线，雪莉本可以赚取数十亿美元的合法赏金，但他选择公开发布视频、追求影响力和曝光度，客观上为公众利益作出了巨大贡献。
更重要的是，尼克·雪莉的街头调查彻底让传统主流媒体蒙羞。曾经敢于敲门质询、深挖五角大楼“900美元一把锤子”或泰科国际（Tyco）CEO 丹尼斯·科兹洛夫斯基（Dennis Kozlowski）“6000美元象牙雨伞架”丑闻的深度调查记者，如今在传统大媒体中几乎绝迹；CNN、《纽约时报》和《洛杉矶时报》彻底沦为了兜售立场观点的宣传工具，甚至在面对体制内荒唐的贪腐公然护短。主流媒体不仅不再监督政府浪费，反而动用舆论机器将试图协助政府削减开支的商业领袖妖魔化并排挤出局。最后，几位主持人总结道，2028 年美国大选最有力的获胜方案绝不是向中产阶层加税，而是拿出切实有效的产品化方案，彻底根除公共系统中至少 20% 到 30% 的浪费、欺诈与滥用（waste, fraud, and abuse）。随后话题转向蒂姆·库克（Tim Cook）卸任与苹果新任 CEO 的任命。
- **加州游民工业复合体**：每年耗费 120 亿美元财政资金，游民规模却持续膨胀。
- **调查新闻的断代**：主流媒体放弃了最基础的穿鞋跑街、上门质询，独立公民记者凭借一部手机揭开数十亿美元级别的黑幕。
- **2028 年政治施政核心**：通过技术与管理手段消灭公共财政中 20%-30% 的系统性损耗，取代无节制的加税政策。
### [45:04](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=2704s) All-In 峰会门票告急与宾夕法尼亚两党参议员辩论预告
Jason 率先提醒听众，第五届 All-In 峰会将于 9 月 13 日至 15 日在洛杉矶举行，即便此前临时补录了少量名额，也转瞬间被抢购一空。他催促想要参会的人赶紧去官网申请，免得临近开幕前 60 天再跑来求情说“我是老粉，快放我进去”。紧接着，Sacks 抛出了峰会的一个重磅政治环节：来自宾夕法尼亚州的两位现任联邦参议员——共和党人 Dave McCormick 与民主党人 John Fetterman 将同台亮相，共同接受主持并从左右两派视角展开交锋。
现场随即陷入对两人反差画风的插科打诨。Jason 调侃 Fetterman 的到来意味着现场着装要求变成了凉鞋、短裤配 T 恤的“工装流浪汉风”，甚至笑称大家得把 Timberland 大黄靴翻出来；而另一边的 McCormick 则身形健硕、英俊干练。几人打趣说，这一场对决的看点完全可以包装成“谁穿得更好”（Who wore it better?）——一边是身着顶级 Brioni 定制西装的 McCormick，另一边则是穿着 Old Navy 平价休闲服的 Fetterman。
### [46:19](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=2779s) 苹果 CEO 换届：硬件老兵 John Ternus 接棒与创新停滞的批判
话题迅速切换到重磅商业新闻：Tim Cook 宣布卸任苹果 CEO，由在苹果效力长达 25 年的硬件工程高级副总裁 John Ternus 接任。Ternus 在预测市场 Polymarket 上从第一天起就是最大热门，他深耕 iPad、AirPods 等硬件管线，被报道普遍评价为作风果断的决策者。Jason 直言，苹果终于重新迎回了一位懂产品的人坐镇帅位，因为现有的产品线正变得日渐陈旧。他犀利地批评道，库克这十余年极其出色地“榨干了乔布斯产品遗产的每一分钱”（包括 iPhone、Apple TV、Apple Watch），但整套体系早已不再以创新为基石，而是完全沦为利润驱动机器。
在谈及苹果痛失的发展契机时，大家提到了诸多遗憾：Siri 体验糟糕透顶（Jason 抱怨自己给苹果贡献了上万美元买手机，结果 Siri 过了快二十年依然拼不对常见人名）；苹果错过了像 Meta 与 Ray-Ban 联名智能眼镜那样的突破口；投入巨资的自动驾驶汽车项目（Project Titan）最终惨遭腰斩；头显 Vision Pro 又重达数磅且远离主流市场。当被问及如果坐上苹果董事会席位会给新 CEO 提何建议时，Friedberg 指出，未来的软件底层架构与以往有着本质不同。苹果当务之急并不是再去死守旧软件，而是必须打造一个跨越所有设备的无所不在的个人 AI 层（AI layer）：它必须极其深度地理解用户本身，无缝打通邮件、日程、音乐偏好与智能家居，让用户在任何一块苹果屏幕前都能通过自然语言进行交互。
### [48:50](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=2930s) Sacks 与 Chamath 评库克功过：极致的资本管理者与商业操盘手
Sacks 对接下来的产品策略提出建议，他认为苹果不必在底层大模型领域盲目死磕前沿军备竞赛，更合理的解法是把 AI 赋能的 Siri 做成开放架构，允许用户在系统设置里自由绑定模型供应商（无论是 ChatGPT、Grok 还是 Claude），同时把好端侧存储与隐私的关口。回顾库克执掌苹果的 15 年，Sacks 给予了极高的历史评价：库克带领公司市值实现了超过 10 倍的飞跃，年营收从约 1000 亿美元膨胀至 4000 多亿美元，并通过大幅拉升软件服务收入的占比显著改善了盈利质量。更难能可贵的是，在科技巨头频陷泥潭的十几年里，苹果几乎没有任何公关丑闻或重大翻车，库克对用户隐私的执着捍卫赢得了公众口碑，甚至在政商关系上也游刃有余、极具分寸。
```mermaid
flowchart TD
accTitle: 苹果CEO角色分类与库克模式
accDescr: 对比开拓型创新者与守成型资源分配者的不同定位与表现
subgraph Types["CEO 的两大维度"]
A["开拓创新型 CEO<br/>（如 Steve Jobs）<br/>定义颠覆性硬件与体验"]
B["极致管家型 CEO<br/>（如 Tim Cook）<br/>精细化资本与人才分配"]
end
subgraph Execution["库克任内核心资本运作"]
B --> C["自研 Apple Silicon 芯片<br/>（摆脱 Intel，赋能端侧 AI）"]
B --> D["极度克制的外部并购"]
B --> E["天量现金分红与股票回购"]
end
```
Jason 则依然为苹果丧失锐气感到惋惜，认为自乔布斯离世、Jony Ive 出走后，苹果丢掉了那种敢为人先的“灵魂”，完全进入了迭代与收割利润的守成模式。Chamath 随即加入讨论，指出库克承担的是一项“近乎不可能的任务”——“这就像你刚跟迈克尔·乔丹做完队友，转头就被要求去成为迈克尔·乔丹”。Chamath 将顶级 CEO 划分为两类：一类是不断拓展边界的“创新者”（Innovator），另一类则是极致审慎的“守成管理者/管家”（Steward）。库克在第二类角色中做到了当之无愧的巅峰。
- **无可挑剔的资源统筹**：在掌舵期间，库克牢牢把控着“资本”与“人才”两大核心资产，不仅在研发端平稳去除了对 Intel 的依赖，全力推出了业界领先的自研芯片（Apple Silicon），这一代架构在今天本地化跑端侧 AI 模型的场景中展现出极其亮眼的能效优势。
- **克制并购与超额回报**：与盲目大肆收购不同，库克保持了极度克制的收购频率，同时在资本回报层面走出了与乔布斯完全相反的路线——通过常态化的大规模股息派发与天量股票回购，向股东输送了惊人的现金回报。
### [54:43](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=3283s) 苹果的资本配置：从乔布斯到库克，以及特努斯面临的多样化硬件挑战
在回顾苹果的资本配置历史时，几位主持人对比了史蒂夫·乔布斯与蒂姆·库克的鲜明反差。在乔布斯掌舵期间，公司对回购和分红毫无兴趣，次数“很容易数，就是零”。乔布斯极度喜欢把现金囤积在资产负债表上，时刻准备为某些颠覆性的终极目标（big shot on goal）发起全力冲刺。而库克时代则彻底倒向了股东友好路线——在库克任期内，苹果的总股本被缩减了将近 50%（约 44%）。查马斯（Chamath Palihapitiya）直言，库克是商业史上罕见的、极具成效且坚定回馈股东的 CEO，他找到了源源不断给投资者退钱的途径，每一位持有苹果股票的人都对此心怀感激。
然而谈及新任 CEO 约翰·特努斯（John Ternus）的未来，查马斯指出了潜藏在苹果财报深处的危机。当前 iPhone 的单机售价虽然水涨船高，但随着 AI 的爆发，人机交互的画布正被彻底撕开。未来世界绝不可能仅仅被两台终端和两套操作系统垄断，而将演进成一个高度异构化的硬件生态——智能笔、发光球体（orbs）、智能眼镜等各类形态将百花齐放。如果苹果过度沉迷于 iPhone 这款利润极度丰厚、粘性极强且拥有提价特权的单一“硬性毒品”，一旦底层生态转向开放的 MCP（Model Context Protocol）协议和跨设备自主智能体（Agents），单一王牌产品的护城河便会开始衰变。特努斯必须带领苹果跳出单一硬件的舒适区。
杰森（Jason Calacanis）对此表示赞同，并进一步指出苹果在可穿戴设备领域（如 AirPods 和 Apple Watch）已经打下了极佳的基础。未来的真正爆发点在于“AI 同步协同”——让用户的眼睛（眼镜）、耳朵（耳机）、手腕（手表）、手机和桌面终端通过统一的 AI 模型无缝感知与同步。杰森还现场展示了自己用来录音的便携智能硬件，并提出第五大潜力方向：消费级机器人。他认为如果乔布斯今天还活着，在看到扫地机器人 Roomba 和特斯拉的人形机器人 Optimus 时，一定会敏锐地切入消费级机器人和智能汽车领域并高水准落地。
---
### [57:46](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=3466s) 苹果接班人与迪士尼历史镜像：特努斯会迎来“70年代低迷”还是“80年代复兴”？
大卫·萨克斯（David Sacks）提出了一个引人深思的历史类比：当前特努斯的接班，与当年华特·迪士尼去世后的迪士尼帝国如出一辙。据传乔布斯生前曾特意叮嘱过库克：“千万不要重蹈迪士尼的覆辙。”当年华特·迪士尼去世后，整个公司因过度执念于华特的遗志、不敢越雷池一步而陷入长期停滞。当时接手迪士尼的是华特的哥哥罗伊·迪士尼（Roy Disney），罗伊是一位卓越的首席运营官（COO）式商业合伙人，其角色就像库克之于乔布斯。罗伊在华特死后让迪士尼的神迹延续了大约 5 年，直至 1971 年罗伊本人去世，随后迪士尼陷入了长达十几年的平庸管理层泥潭，直到 1984 年迈克尔·艾斯纳（Michael Eisner）入主才重振雄风。
```mermaid
flowchart TD
accTitle: 迪士尼与苹果领导层代际演变对比
accDescr: 对比华特·迪士尼到艾斯纳的发展脉络与乔布斯到特努斯的传承路径
subgraph Disney["迪士尼演变路径"]
W["创作者核心<br/>华特·迪士尼 (Walt Disney)"] --> R["运营型合伙人<br/>罗伊·迪士尼 (Roy Disney, ~5年)"]
R --> D70["1970年代<br/>缺乏远见的管理停滞期"]
D70 --> E80["1984 艾斯纳 (Michael Eisner)<br/>激活片库与业务创新"]
end
subgraph Apple["苹果演变路径"]
S["创作者核心<br/>史蒂夫·乔布斯 (Steve Jobs)"] --> T["运营型合伙人<br/>蒂姆·库克 (Tim Cook, 15年增值10倍)"]
T --> J["新任 CEO<br/>约翰·特努斯 (John Ternus)"]
J -.-> Q{"重蹈70年代停滞<br/>抑或开启80年代跃升？"}
end
```
萨克斯分析认为，库克作为商业伙伴获得了长达 15 年的黄金任期，并让苹果市值飙升了 10 倍以上（在市值后硬生生加了一个“0”），成绩堪称梦幻。但如今特努斯接棒，苹果已正式走完了自己的“华特与罗伊时代”。摆在特努斯面前的核心拷问是：苹果接下来会陷入类似迪士尼 70 年代的沉沦低迷，还是能迅速找到属于自己的 80 年代复兴路线？
杰森随即补充了艾斯纳与鲍勃·艾格（Bob Iger）在迪士尼商业史上的破局打法：
- **艾斯纳的“片库策略”（The Vault Strategy）**：他发现迪士尼拥有无与伦比的历史 IP 资产，于是规定经典动画电影（如《白雪公主》《小鹿斑比》）每隔 7 年才重新在院线放映一次，人为制造稀缺性并形成极高周期的发行业务。
- **艾格的“资产负债表并购扩张”**：艾格接手后充分利用母公司的资产负债表与主题乐园分发网络，大举重金收购皮克斯（Pixar）、漫威（Marvel）和卢卡斯影业（星球大战 IP）。
杰森指出，这为特努斯提供了极佳的借鉴：苹果与其单纯依靠账面巨资不断回购销毁股票、向股东返还现金，不如利用其坚不可摧的资产负债表发起极具野心的大型战略并购，这或许对股东长期价值更具增值意义。
---
### [01:00:32](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=3632s) 南方贫困法律中心（SPLC）遭大陪审团起诉：制造仇恨的捐款收割机
随后，播客切入当周引发全美轰动的政治与社会热点：南方贫困法律中心（Southern Poverty Law Center, 简称 SPLC）被联邦大陪审团正式起诉。查马斯首先抛出一个令人瞠目结舌的数据质问：作为一个口口声声为弱势群体维权的非营利组织，SPLC 是如何通过运作在海外离岸银行账户中囤积了高达 8.22 亿美元巨资的？几位主持人直呼这是“有史以来最大的政治骗局之一”。
杰森梳理了案件的核心指控：SPLC 被联邦控以 11 项电汇欺诈（wire fraud）和洗钱（money laundering）罪名。大陪审团指控在 2014 年至 2023 年长达九年的时间里，该机构利用隐秘的秘密银行账户，挪用了至少 300 万美元的善款资金，专门雇佣并操纵多名“有偿线人”（paid confidential informants）。该机构表面上的法定宗旨是“充当美国南部及其他地区种族正义的催化剂，消除白人至上主义，促进人权”，暗地里却指挥线人深度渗透 3K党（Ku Klux Klan）、雅利安国度（Aryan Nation）、新纳粹组织以及 2017 年夏洛茨维尔“团结右翼”（Unite the Right）集会的策划团队。杰森特别指出，该机构长期滥用“仇恨监测”（Hatewatch）的标签武器，不仅针对真正的极端组织，甚至将神经科学家兼作家萨姆·哈里斯（Sam Harris）这种温和理性派贴上“仇恨相关”的标签进行政治追杀与取消。
更为恶劣的是其背后的商业利益链条。夏洛茨维尔冲突事件发生后，主流媒体断章取义所谓特朗普“双方都有好人”的发言，直接激化了全美对立情绪，甚至成为了拜登参选的导火索。而 SPLC 借此恐慌大发横财，其机构营收从 2016 年的 5800 万美元垂直暴增逾一倍至 1.36 亿美元，此后多年持续维持在高位。
起诉书曝光的具体案情更为骇人：
- **核心线人代号 F-37**：在 2015 年至 2023 年期间，SPLC 秘密向其支付了超过 27 万美元。
- **亲自策划暴乱**：起诉书显示，F-37 实际上正是 2017 年夏洛茨维尔集会线上领导策划群的核心成员，是在 SPLC 的直接授意下亲赴现场，甚至在机构严密监督下在网上公开发表种族仇恨言论，并协助为暴乱集会协调与会人员的交通出行。
萨克斯总结道，真相令人震惊且讽刺：这家打着反种族主义旗号、每年向善良民众募集数以亿计捐款的左翼 NGO，实际上秘密将逾 300 万美元倒灌给全美最暴力的极端种族主义组织，包括 3K党和美洲纳粹党。他们通过暗中出资催生、煽动乃至组织极右翼暴力抗议活动，随后回过头来向公众宣称“美国的种族主义危机已迫在眉睫”，以此裹挟公众恐惧、源源不断地收割巨额捐款。这种自导自演制造仇恨的恶劣行径，让该机构彻底沦为法律审判的对象。
### [01:04:23](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=3863s) SPLC 诉讼细节：27万美元“投资”换来8100万美元筹款与假账户掩护
大卫·萨克斯（David Sacks）继续剖析南方贫困法律中心（SPLC）在夏洛茨维尔（Charlottesville）事件后的运作内幕。他指出，在事件发生后，SPLC 筹集到的资金激增了 8100 万美元，这意味着他们最初投入的 27 万美元带来了高达 8100 万美元的回报，这堪称一笔回报丰厚却性质极其恶劣的“投资”。萨克斯强调，整个事件的核心就是一出精心编排的诈骗（grift）。起诉书中最具杀伤力的证据之一在于，SPLC 涉嫌使用虚构实体开设银行账户，专门用来向自己的捐款人隐瞒这些资金往来。因为一旦好莱坞明星和富豪捐赠者得知自己的善款实际上流向了三K党（KKK）等极端仇恨组织，源源不断的捐款就会立刻终止。
面对这番指控，杰森·卡拉卡尼斯（Jason Calacanis）试图稍作中立澄清，提醒大家目前这些内容仍属于刑事起诉书中的指控阶段。SPLC 方面的辩解是，他们并没有策划这些活动，而是在执行“情报监控”，给出的款项是支付给线人的报酬。杰森指出，SPLC 的抗辩逻辑在于，如果使用 SPLC 的官方账户给线人打款，就如同联邦调查局（FBI）直接用公开机构账户向卧底线人开支票一样荒谬。
然而，萨克斯对这种“线人掩护说”嗤之以鼻，并指出其中存在两个致命漏洞：
- 第一，SPLC 支付资金的对象并不是打入极端组织内部的卧底或“鼹鼠”，而是这些组织的实际头目；
- 第二，这些极端组织头目收钱不是为了提供情报，而是直接受雇去煽动并组织极端活动；
- 萨克斯质问道，SPLC 根本不是执法机构，无权行使此类特权；如果他们真的问心无愧，为何要煞费苦心对捐赠人隐瞒全部资金轨迹。
```mermaid
flowchart TD
accTitle: SPLC 争议资金运作模式
accDescr: 展示指控中 SPLC 如何通过虚构实体付款给极端组织并最终通过制造社会恐慌获取巨额捐款的闭环流程
subgraph FundFlow["隐秘资金链条"]
A["SPLC 募捐资金"] --> B["开设虚构实体银行账户"]
B --> C["向白人至上/极端组织头目打款 ($27万)"]
end
subgraph EventAndReturn["事态演变与回报"]
C --> D["煽动夏洛茨维尔等极端暴力事件"]
D --> E["媒体头条渲染仇恨与恐慌叙事"]
E --> F["社会捐款激增 ($8100万)"]
F --> G["沉淀至离岸账户 ($8.22亿储备)"]
end
```
---
### [01:06:58](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=4018s) 非政府组织失控：8.22亿美元离岸资金与“逆向操作”叙事套路
杰森随后将问题抛给查马斯·帕里哈皮蒂亚（Chamath Palihapitiya）：一家非营利机构动用秘密线人渗透并资助这些组织，究竟是为了什么目标？杰森直言，“只要给我看激励机制，我就能告诉你结果”——SPLC 的现实激励就是：社会上的种族矛盾越激烈，他们能收到的捐款就越多。
查马斯情绪十分激动，痛斥当今的非政府组织（NGO）已经彻底失控。他指出，大量 NGO 正在“角色扮演”（cosplaying）成凌驾于大众之上的社会主宰者和权力掮客，必须立即予以拆解和清算。查马斯公开向听众呼吁，任何曾向 SPLC 捐过款的人都应该起诉他们并要求全额退款，把所有账目底册彻底撕开审查。
- **离岸巨额资金**：查马斯透露，目前有高达 8.22 亿美元的捐赠款正趴在 SPLC 的离岸银行账户中，这些资金本该属于被欺骗的捐赠者。
- **逆向炒作套路**：查马斯揭露了部分现代政治倡导团体的标准操作剧本（playbook）——口头上反对什么，暗地里就煽动相反的极端事态。表面上反种族主义，实际上在助长种族对立；声称维护少数群体权益，实际上通过制造极端事件反噬该群体的社会声誉。通过激化矛盾炮制耸人听闻的叙事，勾结主流媒体盟友选择性报道并无限放大，随后借机狂揽政治权力和巨额捐款。
---
### [01:08:39](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=4119s) 501(c)(3) 制度漏洞与全面审计呼声
杰森转向大卫·弗里德伯格（David Friedberg），询问他认为这是典型的“纵火犯冒充消防员”（自己放火再去灭火以求赞誉），还是一场针对左翼机构的政治司法战（lawfare）。弗里德伯格从美国国税局（IRS）的底层制度设计切入，痛批了整个非营利体系的虚伪性。
弗里德伯格指出，根据 IRS 501(c)(3) 的法定定义，免税组织必须从事纯粹的免税豁免活动，具体涵盖：慈善、宗教、教育、科学、扫盲、公共安全、促进业余体育竞技，以及防止虐待儿童或动物。弗里德伯格愤怒地反问：“现在市面上 90% 的所谓非营利组织，究竟凭什么能套进这些定义里？”他认为整个社会长期以来都在闭目塞听，放任无论左翼还是右翼的组织打着非营利旗号，行商业谋利和政治操弄之实。
作为一个坚定的自由意志主义者（libertarian），弗里德伯格明确表达了自己的立场：
- 任何组织若想从事其追求的政治活动或社会议程，完全可以自由去做，但绝不应享有 501(c)(3) 免税地位；
- 捐赠者向这些机构出资时不应享受慈善捐款税收抵免（tax deduction），政府更不能拨付公帑资助；
- 当前正是彻底“清仓”（reset the decks）的大好时机，应当全面整肃非营利生态，斩断利用免税外衣洗钱、吸金并逍遥法外的灰色链条。
---
### [01:10:45](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=4245s) 取消文化往事与“非营利机构造孽论”的调侃
接续对 NGO 利益驱动的讨论，杰森提炼出本期节目的核心主题——“全面审计一切”（audit everything）。无论是政府浪费、NGO 账目，还是陶氏化学（Dow）等巨头数十年前生产的化学品与癌症发病率的潜在关联，社会都亟需一场公开透明的彻底审查。杰森强调这绝不是非红即蓝的党派之争，而是纯粹的“绿色”（美元利益驱动）。当社会大众无法得知种族现实真相时，其破坏力是巨大的。
随后杰森回忆起自己在 2014 年遭遇“取消文化”（cancel culture）围攻的往事。他打趣提到自己职业生涯曾获得 Vice 杂志评选的“2014 年度最冒犯推文奖”。当时杰森在推特上发文称，在科技新闻和博客行业根本不存在种族歧视，任何人只需坚持写博客几年展示实力，谁也拦不住你，会有大把工作向你敞开。结果这一言论被扣上“无视种族特权”的帽子，各方势力试图切断他的媒体品牌和天使投资业务，对其事业造成了大约一年的负面冲击。
这段自嘲随即引来兄弟团的激烈互怼：
- 萨克斯咳嗽了一声反问：“慢着，杰森，你是说 SPLC 把你放进了抵制黑名单？”杰森承认不是 SPLC，而是 Vice 和山姆·哈里斯（Sam Harris）式的风波，但他切身体会过那种狂热。
- 萨克斯嘲讽杰森：“你平时做道德宣示（virtue signaling）那么小心翼翼，居然还会有人想取消你，真是不可思议。”
- 查马斯则狂妄宣称自己“根本不可被取消”（uncancelable），因为他打心底里根本不在乎这些蠢人在想什么，而杰森太在乎外界的看法。
在欢笑与调侃声中，查马斯突然向杰森抛出一记灵魂拷问：“杰森，你扪心自问，在你的潜意识深处，有多大比例觉得你那高尚小巧的人权观察（Human Rights Watch）组织，其实也在自己制造人权侵犯事件来拉赞助？”杰森不得不苦笑承认，回过头看很多当年的国际人权机构，确实也陷入了类似的激励陷阱之中。
### [01:12:50](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=4370s) 非政府组织的使命偏移：从反酷刑到审查与挑动对立
Jason 回忆起自己早年在国际特赦组织（Amnesty International）工作的经历，指出该组织最初拥有一项非常崇高的法定使命——依据埃莉诺·罗斯福（Eleanor Roosevelt）主持起草的《世界人权宣言》，专注于揭露和制止酷刑、非法监禁以及因言论自由受到的迫害。然而，随着时间推移，包括国际特赦组织和人权观察（Human Rights Watch）在内的机构为了筹集资金而彻底偏离了方向。它们开始插手跨性别权利等各种泛化的议题，甚至反过来审查他人，比如因为神经科学家 Sam Harris 邀请了《钟形曲线》（*The Bell Curve*）的作者 Charles Murray 做客播客，就对 Sam Harris 发起口诛笔伐。
```mermaid
flowchart TD
accTitle: NGO使命演变与异化逻辑
accDescr: 展示非政府组织从早期崇高人权使命演变至为筹款制造议题并推动结果平等的路径闭环。
subgraph EarlyStage["早期初衷（合法性确立）"]
A["明确具体使命<br/>（反酷刑/反囚禁/言论自由）"] --> B["赢得公信力与社会捐款"]
end
subgraph MissionCreep["使命偏移（维持生存与预算扩张）"]
B --> C["原有历史目标基本达成"]
C --> D["绝不宣布胜利与解散机构"]
D --> E["拓展议题与制造危机感<br/>（引入跨性别/身份政治）"]
E --> F["移动球门：从机会平等转向结果平等"]
F --> G["持续向捐赠人/政府游说要钱"]
G --> D
end
```
面对 Chamath 的直接追问——这些所谓的人权组织究竟有多大几率在暗中煽风点火、制造并不存在的人权危机以骗取信任，Jason 坦言这取决于具体机构：对于南方贫困法律中心（SPLC），他认为有 95% 的概率存在这种恶劣行为；而即便是国际特赦组织，今天在他看来也有 50% 的几率沦为翻硬币一样的存在。Jason 感慨，这些组织如今都已被彻底收编并变质，即便 SPLC 有一个伟大的创立初衷，现在也已经面目全非。
---
### [01:15:01](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=4501s) 无法宣告胜利的NGO系统性顽疾与“移动的球门”
谈及针对 SPLC 的联邦大陪审团起诉，Chamath 强调，能够走到大陪审团起诉这一步，说明检方已经完成了核心证据的预审，绝非普通人在掀起司法战（lawfare）。Jason 则引用法律界名言调侃道“大陪审团连一块火腿三明治都能起诉”，但他也承认，如果 SPLC 资助三K党（KKK）和新纳粹组织作为线人的指控属实，那简直匪夷所思，不过理应让其在法庭上接受检验。
Sacks 随即剖析了非营利组织（NGO）与商业公司在底层激励机制上的根本区别：
* **商业公司拥有市场反馈机制**：企业必须创造人们愿意购买的产品，依靠营业收入和利润维持运营，一旦持续亏损就会倒闭淘汰。
* **NGO 是纯粹的筹款驱动**：它们不售卖产品，一切依赖向捐赠人募资。随着时间推移，实际活动本身的有效性退居其次，能否编造理由持续向金主募资成了唯一的目标；一旦获得政府资助，情况则更糟，它们会将全部精力放在游说政府拨款上，项目是否奏效变得无关紧要，关键只在于如何粉饰包装。
Sacks 进一步提出了自己的理论来解释为何南方贫困法律中心不再关注美国南方依旧客观存在的贫困问题，反而沉迷于挑起种族对立。民权运动曾是一场针对种族隔离和吉姆·克劳法案（Jim Crow）的伟大抗争，并且取得了历史性胜利。但 NGO 内部没有任何人会主动宣布胜利并卷铺盖走人——“没有人会说‘我们解决了问题，把我的岗位撤了吧’”。在 2008 年奥巴马当选总统时，绝大多数人都已认识到美国不再是一个将最高职位因肤色而拒之门外的种族主义国家，这本该是宣告使命完成的时刻。
然而，这些组织非但没有功成身退，反而从奥巴马的第二任期开始彻底“移动了球门”，炮制出所谓的“反种族主义”（anti-racism）理论。这套理论将评判标准从最初的“机会平等”（equality of opportunity）硬生生替换为“结果平等”（equality of results / equality of outcomes），要求社会各项分布必须完全对齐人口比例，本质上是一种身份社会主义或共产主义包装。Jason 赞叹道，这种逻辑一旦被看穿就无法忽视，它完全是组织为了自保而重新将种族定义为永恒社会痼疾的手段。Chamath 也插话表示，客观事实是“美国社会的种族歧视程度远远低于大众想象”。
---
### [01:19:03](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=4743s) 科学角：年轻人群结直肠癌暴增80%背后的表观遗传学线索
在严肃的政治与制度讨论告一段落后，主持人将聚光灯交给了准备带来“惊喜科学角”的 Friedberg。Friedberg 介绍了一篇发表在权威期刊上的重要最新论文，旨在揭示青年人结直肠癌（Colorectal cancer）激增背后的潜在诱因与分子标记物。
结直肠癌目前已攀升至人类第三大致死癌症。在过去大约二十年里，50岁以下年轻人群的患病率出现了令人触目惊心的激增，增幅已超过 80%。历史上，结直肠癌一直被普遍认为是一种老年退行性疾病——通常在 70 岁以上人群中发病率才会急剧上升，因此年轻群体患病率的暴涨成为了困扰医学界的一大未解之谜，外界对于引发该病的根本环境诱因一直缺乏定论。
* **研究团队与研究设计**：来自西班牙巴塞罗那的研究团队开展了一项开创性研究，专门对比了 50 岁以下早发性患者与 70 岁以上晚发性患者肿瘤细胞的表观基因组（Epigenome）与基因表达差异。
* **分子机制解析**：当人体暴露于某种特定外部环境（无论是某些特定食物、饮品还是环境化学物质）时，受影响的细胞会触发特定基因的开启或关闭。研究人员正是通过检测肿瘤细胞中的 RNA 水平，观察哪些基因正在活跃表达合成蛋白质，哪些基因被异常沉默，从而顺藤摸瓜锁定驱动年轻人患癌的特异性分子机制与环境诱因。
### [01:21:10](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=4870s) 青年结肠癌激增的真凶：毒莠定（Picloram）浮出水面
Friedberg 继续深入剖析这篇重磅科研论文的研究方法。研究团队从由联邦政府资助的“癌症基因组图谱”（The Cancer Genome Atlas, TCGA）中获取了结肠腺癌（colon adenocarcinomas）组织样本，并将 50 岁以下的年轻患者与 70 岁以上的常规高发年龄患者进行基因表达谱（gene expression profile）对比。常规情况下，70 岁以上人群患结肠癌的风险极高，而 50 岁以下极为罕见。研究人员全面控制了各项混杂变量，包括生活方式、饮食评分、摄入量、肥胖程度、饮酒史、出生体重及性别等。在海量环境暴露因素的筛选分析图表中，有一整行格外刺眼地呈现出全然的高警示橙色——这正是除草剂毒莠定（picloram）。
```mermaid
flowchart TD
accTitle: 毒莠定暴露与早发性结直肠癌关联研究流程
accDescr: 展示从TCGA样本对比到表观遗传学分析及郡县流行病学印证的完整研究链条
subgraph TCGA["TCGA 组织样本对比"]
A["50 岁以下早发患者<br/>组织样本"]
B["70 岁以上常规患者<br/>组织样本"]
end
subgraph Analysis["表观遗传学与转录组分析"]
C["RNA 测序与基因表达分析"]
D["控制饮食/肥胖/饮酒/性别等变量"]
E{"筛查关键环境诱因"}
F["锁定农药：毒莠定 (Picloram)"]
end
subgraph Field["环境流行病学验证"]
G["美国多州郡县农药用量数据"]
H["高使用量地区患癌几率激增 (OR ≈ 3x)"]
end
A & B --> C
C --> D --> E --> F
F --> G --> H
```
毒莠定是陶氏化学（Dow Chemical）于 1963 年研发的一种人工合成植物激素类农药，其化学结构与植物天然分泌的生长素（auxin）高度相关。在 20 世纪 60 年代，工业界流行合成此类激素并喷洒于植物表面，诱发其失控疯长后快速枯死。随后数十年间，毒莠定被广泛投放于生态环境中：用于牛群放牧的牧场和草场杂草治理、清除公路及铁路沿线的灌木，以及清理工业用地与高压输电走廊的植被。然而，毒莠定最致命的特征在于极难发生生物降解（biodegradable），它会在自然环境中存留超过一年，并随地表水渗入地下水系统，形成长期的持久性污染。
令人震惊的是，美国环境保护署（EPA）上一次针对毒莠定展开系统的安全评估还要追溯至 1995 年。在那个时代，科学界尚未掌握表观基因组学（epigenomics）技术，过去的传统毒理测试往往只是直接将化学品施加在小鼠或人体细胞上，若短时间内没有引发急性细胞病变或突变，便简单判定为“无毒合格”。但新的表观遗传学研究表明，环境化学物质的长期低剂量暴露会篡改表观基因组，像开关一样错误地激活或关闭特定基因，导致组织细胞功能失控并最终恶化为肿瘤。这篇论文不仅为青年群体结肠癌高发提供了强有力的环境归因，也理应促使 EPA 重新评估该化学品的合法性，并在全美乃至全球范围内建立基于表观遗传学的新型工业化学品安全评估机制。
- **化学品背景**：毒莠定（Picloram）由陶氏化学于 1963 年研发，属于合成生长素类除草剂。
- **环境持久性**：难以降解，残存期超过一年，易随降雨进入水体和地下水网络。
- **监管严重脱节**：EPA 上一次安全审查发生在 1995 年，无法检测由表观遗传异常引起的远期致癌效应。
---
### [01:25:31](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=5131s) 全美郡县数据印证与基础科研的不可替代性
面对 Chamath 关于“患者是否仅仅因为居住地而接触到毒莠定”的追问，Friedberg 进一步列举了论文中的流行病学证据。研究团队调取了来自美国国家农药综合监测项目（Pesticide National Synthesis Project）的用药量评估，并将数据覆盖至加利福尼亚、康涅狄格、佐治亚、爱荷华、新墨西哥、犹他及华盛顿等州的大量郡县。横向对比显示，在毒莠定高用量的郡县中，年轻群体的结肠癌发病率呈现出极强的统计学正相关，其优势比（odds ratio）达到了惊人的约 3 倍，证实了环境暴露与患癌几率之间的明确联系。
针对该化合物目前的供应链现状，Friedberg 指出其核心专利早已到期，目前绝大多数原料药很可能是在中国以通用名药物的形式合成制造，随后在全球贴上各类品牌分装销售，导致其在农业与市政维护中彻底泛滥。Jason 认为，这项研究展现了大数据检测与 AI 算力结合的巨大威力，未来完全可以利用 AI 逆向筛查所有历史化合物。但 Friedberg 随后以兼任总统科技顾问委员会（PCAST）成员的视角纠正道，真正的功臣并非花哨的 AI，而是**政府对基础科研的持续托底**。
美国国家癌症研究所（NCI）多年前斥资 1 亿美元搭建了癌症基因组图谱数据库，如今每年仅需数百万美元维护样本库与数据接口，便能向全球科学家敞开大门，再结合近几年飞速成熟的 RNA 测序技术（RNA-seq），人类才得以直接探测基因的“开关状态”。这一发现绝不该止步于西班牙科研团队的一篇单打独斗的论文，政府监管机构应当主动挑起重担，把过去工业化发展中犯下的错误逐一纠正，将潜藏在食物链与工业体系中的隐形毒物彻底清除。
讨论进入尾声时，现场画风一转。Jason 借着“结肠癌”与 Friedberg 的 PCAST 职务，大开双关冷笑话：“Sacks，Friedberg 对天王星（Uranus，谐音 your anus/你的肛门）的执念，如何指导你们在 PCAST 的共同领导？你们打算在这个结肠研究上钻多深？在委员会一天听 8 场这种汇报顶得住吗？”Sacks 顺水推舟大笑调侃：“这就是我们招募 Friedberg 的原因，他负责处理火星、海王星和天王星，他会深入进去好好给它‘做做大扫除’。”在哄笑中，Friedberg 表达了感谢，并强调该论文几天前发表后几乎没有主流媒体报道，他希望借由播客平台让更多大众提高警惕。
- **流行病学关联**：多州郡县流行病学模型显示，高暴露区域的年轻结直肠癌发病优势比约为 3 倍（Odds Ratio ≈ 3x）。
- **供应链现状**：原研专利失效，全球多地通用名仿制并泛滥应用于农牧和基建维护。
- **科研基石**：依赖联邦政府主导建立的 TCGA 生物样本库与新一代 RNA 转录组测序技术，而非单纯的大模型分析。
### [01:29:22](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=5362s) 节目收尾与经典片尾彩蛋
在弗里德伯格（David Friedberg）结束对青年结直肠癌最新科学论文的拆解与讨论后，杰森·卡拉卡尼斯（Jason Calacanis）正式为第 270 期节目画上句号。作为自封的“全球最伟大主持人”，杰森照例向全场的好兄弟们致谢——感谢查马斯（Chamath Palihapitiya）、大卫·萨克斯（David Sacks）以及“科学达人”大卫·弗里德伯格的精彩对谈与深度输出。在向观众、朋友与邻里道别之后，几位好友互道“Love you boys”，在一片轻松愉悦的气氛中结束了本期录制。
紧接着，播客响起了标志性的片头/片尾混音音乐与经典语录串烧。背景音中穿插着粉丝们津津乐道的各种经典老梗与绰号：从萨克斯的“雨人”（Rain Man）、弗里德伯格的“藜麦女王”（Queen of Quinoa），到查马斯奉为圭臬的投资名言“让赢家继续奔跑”（Let your winners ride）。伴随着魔性洗脑的音乐旋律与往期录制中那句令人捧腹的互损名场面——“萨克斯，我家的狗正在你家车道上拉屎呢”——本期《All-In Podcast》在欢声笑语与热烈节奏中完美落幕。
## Notable quotes
> [00:00](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=0s) 这就是为什么我带着细头弹簧刀和 P35 手枪，要是谁想翻过围栏找麻烦，尽管来试试，JCal 早就准备好了 — Jason Calacanis
> [01:02](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=62s) charming AF — David Sacks
> [14:53](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=893s) xAI was not built right first time around, so is being rebuilt from the foundations up — Elon Musk
> [26:28](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=1588s) SaaS's loss is the token dealer's gain — Jason Calacanis
> [36:43](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=2203s) the worst vulture-like business in Silicon Valley — David Friedberg
> [48:50](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=2930s) 这就像你刚跟迈克尔·乔丹做完队友，转头就被要求去成为迈克尔·乔丹 — Chamath Palihapitiya
> [57:46](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=3466s) 千万不要重蹈迪士尼的覆辙 — Steve Jobs
> [01:06:58](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=4018s) 只要给我看激励机制，我就能告诉你结果 — Jason Calacanis
> [01:15:01](https://www.youtube.com/watch?v=1bCXCxrDsCs&t=4501s) 没有人会说‘我们解决了问题，把我的岗位撤了吧’ — David Sacks
## Mentioned
Jason Calacanis, David Sacks, Chamath Palihapitiya, David Friedberg, Donald Trump, SpaceX, Cursor, xAI, Anthropic, Claude, OpenAI, Grok, Colossus, Thoma Bravo, Medallia, Blackstone, Salesforce, ServiceNow, Snowflake, Adobe, Figma, Marc Benioff, Kevin Warsh, Credit Suisse, Tim Cook, John Ternus, Steve Jobs, Walt Disney, Roy Disney, Michael Eisner, Bob Iger, Southern Poverty Law Center (SPLC), Sam Harris, Nick Shirley, Amnesty International, Human Rights Watch, Dow Chemical, Picloram (毒莠定), The Cancer Genome Atlas (TCGA), EPA, PCAST
## Video description
(0:00) Bestie intros!
(4:55) SpaceX-Cursor deal, compute as leverage
(18:33) SaaS bloodbath, debt bomb incoming, buy the dip?
(46:20) New Apple CEO: John Ternus succeeds Tim Cook, what's next for Apple?
(1:00:32) SPLC indictment, out of control NGOs
(1:19:03) Science Corner: Potential cause discovered for colon cancer spike in young people
Apply for Summit 2026:
https://allin.com/events
Follow the besties:
https://x.com/chamath
https://x.com/Jason
https://x.com/DavidSacks
https://x.com/friedberg
Follow on X:
https://x.com/theallinpod
Follow on Instagram:
https://www.instagram.com/theallinpod
Follow on TikTok:
https://www.tiktok.com/@theallinpod
Follow on LinkedIn:
https://www.linkedin.com/company/allinpod
Intro Music Credit:
https://rb.gy/tppkzl
https://x.com/yung_spielburg
Intro Video Credit:
https://x.com/TheZachEffect
Referenced in the show:
https://www.bloomberg.com/news/articles/2026-04-21/spacex-says-has-agreement-to-acquire-cursor-for-60-billion
https://www.bloomberg.com/news/articles/2026-03-02/cursor-recurring-revenue-doubles-in-three-months-to-2-billion
https://techcrunch.com/2026/04/17/sources-cursor-in-talks-to-raise-2b-at-50b-valuation-as-enterprise-growth-surges
https://polymarket.com/event/will-spacex-acquire-cursor
https://polymarket.com/event/spacex-ipo-by
https://x.com/ttunguz/status/2046815725285945820
https://x.com/elonmusk/status/2032201568335044978
https://www.reuters.com/business/thoma-bravo-nears-agreement-turn-software-firm-medallia-over-creditors-source-2026-04-22
https://www.bloomberg.com/news/articles/2026-04-02/blackstone-squeezes-thoma-bravo-and-its-ailing-software-company-medallia
https://x.com/Benioff/status/2044981547267395620
https://www.apple.com/leadership/john-ternus
https://www.bloomberg.com/news/articles/2026-04-21/apple-bets-new-ceo-john-ternus-will-bring-back-jobs-era-decisiveness
https://polymarket.com/event/next-ceo-of-apple
https://x.com/joecarlsonshow/status/2046349686253265302
https://x.com/nickshirleyy/status/2043756610955423782
https://www.justice.gov/opa/pr/federal-grand-jury-charges-southern-poverty-law-center-wire-fraud-false-statements-and
https://www.justice.gov/opa/media/1437146/dl
https://www.vice.com/en/article/2014-vice-news-awards-the-most-offensive-tweet-ubers-white-privilege
https://x.com/nickshirleyy/status/2043756610955423782
https://www.nature.com/articles/s41591-026-04342-5
#allin #tech #news
[Source](https://www.youtube.com/watch?v=1bCXCxrDsCs)
Source quotes
[0:00] Jason Calacanis: 这就是为什么我带着细头弹簧刀和 P35 手枪，要是谁想翻过围栏找麻烦，尽管来试试，JCal 早就准备好了
[1:02] David Sacks: charming AF
[14:53] Elon Musk: xAI was not built right first time around, so is being rebuilt from the foundations up
[26:28] Jason Calacanis: SaaS's loss is the token dealer's gain
[36:43] David Friedberg: the worst vulture-like business in Silicon Valley
[48:50] Chamath Palihapitiya: 这就像你刚跟迈克尔·乔丹做完队友，转头就被要求去成为迈克尔·乔丹
[57:46] Steve Jobs: 千万不要重蹈迪士尼的覆辙
[66:58] Jason Calacanis: 只要给我看激励机制，我就能告诉你结果
[75:01] David Sacks: 没有人会说‘我们解决了问题，把我的岗位撤了吧’
Source quotes
[15:51] Elon Musk: xAI was not built right the first time around, so is being rebuilt from the foundations up.
[73:36] David Sacks: So, this group that was supposed to be fighting racism in fact was fomenting racism by paying these groups to basically organize protests that SPLC could then point to and say that America has a huge racism problem, donate to us... That $270,000 investment led to an $81 million return, pretty good.
[89:25] Chamath Palihapitiya: Love you boys. Bye-bye.
Original transcript
[0:00] Jason, you are the unique person that is
[0:02] at the intersection of both the
[0:05] and the SPLC files. Do you have
[0:07] [laughter] a comment?
[0:08] >> Do you have a comment, Jason?
[0:10] >> in the SPLC files. I
[0:11] >> Yes, you are. You're adjacent to the
[0:13] vice president. You're SPLC adjacent and
[0:14] you're
[0:15] What does that mean? In the Venn diagram
[0:17] >> Thank you, though, for putting me in the
[0:19] crosshairs of all the losers.
[0:20] >> He's got a I got a really good way to
[0:22] select. I mean,
[0:23] >> There's a reason why I'm carrying this,
[0:24] guys. It's because the people
[0:27] >> What the [ __ ] is going [laughter] on?
[0:29] >> THERE'S A REASON WHY I CARRY a stiletto
[0:31] and a P35.
[0:33] >> What the [ __ ] are you doing? [laughter]
[0:35] >> There's a reason. If you want to jump
[0:36] the fence, feel free.
[0:39] >> [laughter]
[0:39] >> J CAL IS READY.
[0:47] >> YOU LET YOUR WINNERS RIDE. [music]
[0:50] >> RAIN MAN David Sacks
[0:54] >> And I said, "We open-sourced it to the
[0:56] fans and they've [music] just gone crazy
[0:57] with it."
[1:02] >> All right, everybody, welcome back to
[1:04] the greatest podcast in the universe,
[1:07] episode 270 of the All-In podcast, your
[1:10] podcast for your favorite podcast. With
[1:12] me again, your sultan of science, David
[1:14] Friedberg, the dick
[1:17] tater, Chamath Palihapitiya, and yeah,
[1:21] the Rain Man is back. Yeah, it's
[1:22] definitely David David Sacks. He's
[1:25] definitely in DC with a with POTUS,
[1:27] yeah. POTUS lets him drive in the
[1:29] driveway. Uh Sacks, what's going on? You
[1:32] you pushed back, you big-shotted the
[1:34] entire crew, and pushed the show back an
[1:36] hour.
[1:36] >> Simple text. He's like, "With POTUS."
[1:39] >> Start unbelievable.
[1:41] >> Start later.
[1:42] >> Okay, we'll just wait. Okay, Daddy. Look
[1:44] at him. All right. All right, big shot.
[1:47] What's going on?
[1:47] >> No, look. I was in DC today and I was at
[1:51] the White House and
[1:53] I just asked if the president had time
[1:55] and he made time. And we had we did have
[1:57] a little meeting and so we did push back
[1:59] the pod for that. One thing I just want
[2:01] to say is just what a pleasure he is to
[2:04] deal with. You know, when I read in the
[2:06] media, they're always describing him in
[2:08] a certain way that you know, he's
[2:10] yelling at people or he's moody or or
[2:12] something like that and that's never
[2:14] ever been my experience with him. He's
[2:15] always pleasant to be with. He's always
[2:18] genial.
[2:19] >> Super charming.
[2:19] >> questions. He's interested in the
[2:21] subject matter.
[2:22] It's just a completely different
[2:23] portrayal. I don't get where the media
[2:25] is coming from at all on this.
[2:27] >> He's charming AF.
[2:28] >> Totally. Totally.
[2:30] >> He's charming.
[2:30] >> maybe if you double-crossed him, maybe,
[2:32] I don't know, but I've just never seen
[2:34] any evidence of of how they describe him
[2:36] at all. And I think on our issues of AI,
[2:39] I think we're really lucky that he's the
[2:41] president who's in the White House when
[2:43] this AI revolution is happening.
[2:45] >> I mean, doing alt history, Sachs, what
[2:46] would happen if Kamala ding-dong was in
[2:49] right now and we'd have like no data
[2:50] centers?
[2:51] >> We'd have no data centers and they'd be
[2:53] using AI to censor us and they'd be
[2:55] promoting DEI values through AI. That
[2:57] was in the Biden executive order.
[3:00] President Trump just wants the country
[3:01] to win and be successful and he doesn't
[3:04] have these like doomer neuroses about
[3:06] it. That's not to say we don't support
[3:09] any regulation at all, but we should
[3:11] have specific solutions for specific
[3:14] problems as opposed to being cowering in
[3:16] fear over this and just trying to halt
[3:18] all progress. And I think a really good
[3:20] example of that was his idea around data
[3:22] centers where he said over a year ago
[3:24] before data centers even became a hot
[3:26] political topic, that we should let our
[3:28] AI companies stand up their own power
[3:30] generation behind the meter. And that's
[3:32] a much better approach than the Bernie
[3:33] Sanders approach of just shutting
[3:34] everything down. So, I don't know. I
[3:36] think we're like very fortunate that
[3:37] he's the president during this critical
[3:39] time in development of this technology.
[3:41] And like I said, he's always been
[3:43] interested in it. He talks a lot of
[3:44] business leaders. I'm always actually
[3:46] very impressed with what he already
[3:47] knows. He listens to like all the the
[3:50] guys in the industry and he synthesizes
[3:52] what he hears. I think he's very good at
[3:53] that.
[3:54] >> He was talking about the Anthropic guys,
[3:55] and he was like, "These are brilliant
[3:56] guys." And he was like, "Giving the
[3:58] flowers to them and how genius they
[4:00] were, and that they were working on a
[4:01] deal." Any insights there about the
[4:03] relationship between the White House and
[4:06] Anthropic?
[4:08] >> I thought what he said was very balanced
[4:09] and accurate. Like you said, he said
[4:12] that they were very smart guys. They do
[4:14] have a great product. I've certainly
[4:16] acknowledged that. He also said that
[4:17] they were very left-wing, but that was
[4:19] something we could work through.
[4:21] >> Hm.
[4:21] >> Didn't have to [clears throat] be a deal
[4:22] killer. He said they tried to tell the
[4:24] Pentagon what to do, which the Pentagon
[4:26] didn't like. But in any event, I mean,
[4:27] look, he wants American companies to be
[4:30] successful, and he he I think genuinely
[4:33] really does like high-IQ people. I mean,
[4:36] he says it all the time, and people
[4:37] think he's joking, but I actually think
[4:38] it's like one of his core convictions is
[4:40] he just really likes smart people. He
[4:42] likes being around smart people.
[4:44] >> Loyal people, smart people, people who
[4:46] are good on camera seem to be the three
[4:47] circles. And hey, Sachs, you fall into
[4:50] two of the three. [laughter]
[4:51] Uh all right.
[4:52] >> I thought the audience would bring it
[4:53] that out.
[4:54] >> Uh topic one,
[4:56] SpaceX has signed a huge deal with
[4:58] Cursor.
[4:59] You know Cursor, that's the AI coding
[5:01] startup. Really, they they they defined
[5:03] the category. XAI and Cursor are
[5:05] building and collaborating on a new AI
[5:07] coding model that would, quote, be the
[5:11] world's best coding and knowledge work
[5:13] AI. Here's the deal as it's been
[5:15] explained. SpaceX will either buy Cursor
[5:18] by the end of 2026 for $16 billion.
[5:20] That's $10 billion more than they were
[5:22] rumored to be raising at, or they will
[5:24] pay Cursor $10 billion for their
[5:27] collaboration together. Bloomberg says
[5:29] you can think of that $10 billion
[5:31] essentially as a breakup fee. So, I
[5:33] think it's fait accompli that this deal
[5:36] is going to get done.
[5:37] Cursor's run rate, $2 billion at the end
[5:39] of February. This is a money-printing
[5:41] machine. They expect to end 2026 with a
[5:44] $6 billion dollar rate. They're going to
[5:46] triple it. SpaceX projected revenue
[5:49] between 22 and 24 billion in 2026. So
[5:52] this is quite accretive to the revenue
[5:55] story at SpaceX at the IPO of SpaceX,
[5:58] which is now targeting a valuation of 2
[6:00] trillion.
[6:01] Which would be trading at roughly 80
[6:03] times top line revenue, which is a, you
[6:06] know, people would say it's a high
[6:07] valuation, but also commensurate with
[6:09] the opportunity. Cursor's valuation
[6:11] would be 30 X. So this is a good deal, I
[6:13] think, for everybody at the end of the
[6:15] day. Cursor started, I think, built off
[6:18] of Anthropic's LLM. You could use any
[6:21] LLM previously on it, but in March
[6:23] Cursor released the second version of
[6:25] their proprietary model, Composer 2. And
[6:29] uh here it is. It's It's ranked pretty
[6:31] high right now. It's between uh GPT-4
[6:33] 5.4 and Opus 4.6, as you can see on the
[6:36] screen.
[6:37] The key part of the story here
[6:40] is that Elon has 550,000
[6:43] GPUs in Colossus. He's scaling up to 1
[6:46] million, and then of course he's going
[6:47] to bring it to space. So if you believe
[6:50] that infrastructure matters, and it's
[6:52] pretty clear it does, this is incredible
[6:55] for Cursor, who has been compute
[6:57] constrained. So this is peanut butter
[6:58] and chocolate. If you put these two
[7:00] together, I predict that this is going
[7:03] to move
[7:05] Space X X XAI and Cursor to the front of
[7:09] the coding leaderboard within 12 months.
[7:12] That's my prediction, Chamath.
[7:14] >> Shareholder in SpaceX via the
[7:17] acquisition of the Starlink company that
[7:21] you were a backer of. What are your
[7:22] thoughts?
[7:22] >> The acquisition was essentially
[7:24] negotiated, and the way that it's
[7:26] structured is so that the S-1 doesn't go
[7:29] stale. So I think the way that it was
[7:31] announced has more to do with the fact
[7:33] that they don't want to slow down and
[7:35] have to rewrite parts of the S-1, have
[7:38] to redo the disclosures, um
[7:41] have to redo the risks. and so I think
[7:43] what you're going to see is that this
[7:45] will get done. In fact, the deal is
[7:47] effectively done.
[7:49] But what's so smart is that where is
[7:51] SpaceX today? Let's call it
[7:53] a trillion.
[7:55] Where could it be?
[7:57] Just for the purpose of this argument,
[7:58] let's say 2 trillion. So when the deal
[8:01] gets done on a stock for stock basis,
[8:03] it's going to be if again, if it's 60
[8:05] billion in tomorrow dollars,
[8:09] effectively Elon's gotten a 50%
[8:11] discount. And what has he bought? He can
[8:12] issue a 60 billion dollars of stock at a
[8:14] 2 trillion dollar valuation and get a
[8:16] model and a service that I think is
[8:20] extremely compelling in coding, which is
[8:22] where we know all of the immediate and
[8:24] short-term revenue gains are. It's also
[8:27] patterns that are hard-fought and are
[8:29] really valuable in reinforcement
[8:31] learning. He gets all of that. And then
[8:33] he gets a very cracked team, which, you
[8:35] know, we've known for a while that the
[8:36] cursor team is absolutely excellent. If
[8:38] you look at the Grok usage, it shows
[8:42] why he had this excess capacity. There
[8:44] was a moment where Grok had a very steep
[8:47] and very aggressive discount on their
[8:49] output tokens.
[8:50] And in that moment, there was just a lot
[8:51] of experimentation and usage. And over
[8:54] time, that sort of
[8:56] went away.
[8:58] So there was a lot of capacity and
[9:01] relatively low utilization, I think,
[9:03] inside of Colossus that he was able to
[9:05] turn around, jujitsu move the whole
[9:07] thing, and basically acquire the most
[9:09] interesting and valuable third-party
[9:12] wrapper service in AI right now. So, and
[9:15] the fact is that they got it
[9:16] effectively, I think, at this price for
[9:17] 30 billion.
[9:19] So I think it was a really good deal,
[9:20] really smart deal.
[9:21] >> Sacks, your thoughts if you want to
[9:23] unpack it a bit. Under the framing, I
[9:25] think it's be interesting for you.
[9:28] If we were sitting here 3 years ago, the
[9:31] Biden administration didn't invite Elon
[9:34] to
[9:35] the EV summit, and the SEC, and other
[9:38] organizations, Delaware, they were
[9:41] explicitly involved in lawfare. They
[9:42] were trying to put Elon in prison, and
[9:44] here we are, the most important company
[9:46] in the history
[9:47] of the United States, SpaceX AI and
[9:51] Tesla, now on the verge of just creating
[9:55] the greatest products in the history of
[9:57] humanity between SpaceX,
[9:59] clusters in space, and Optimus. Your
[10:01] thoughts?
[10:01] >> Well, you're right. I do remember a
[10:03] press conference where Biden said we got
[10:04] to look at the sky. And so, on the heels
[10:06] of that, the DOJ brought a a lawsuit
[10:10] attacking the company for not hiring
[10:11] enough asylum seekers. Remember that?
[10:14] >> SpaceX, which they can't under ITAR.
[10:16] >> They couldn't, exactly, under ITAR.
[10:18] Anyway, that's all ancient history, so
[10:21] let's let's put that behind us. Look, I
[10:23] I agree with your guys' analysis on
[10:25] this. I think these two companies are
[10:26] are very complementary. Cursor,
[10:28] obviously, is very strong in coding.
[10:32] And that's what it brings to xAI. xAI
[10:35] brings compute, and they bring a
[10:37] foundation model. And the problem that
[10:39] Cursor had is that even though coding is
[10:42] kind of like the white-hot area of AI
[10:44] right now,
[10:46] when it got started, it was really
[10:48] competing against generalists in the
[10:50] form of Open AI and Anthropic. But now
[10:52] those generalists have decided to
[10:53] vertically integrate in this area of
[10:55] coding, right? And so, Cursor's now
[10:58] competing against Quad Code and Open
[11:00] AI's Codex. And so, they were dependent
[11:03] on foundation model companies that were
[11:04] getting in the business of competing
[11:06] with them, which was just not a good
[11:07] place to be, right? So now, they have
[11:09] this new alliance with a different
[11:11] foundation model company, which also
[11:13] brings the compute. Just makes a lot of
[11:14] sense, and then they bring Cursor brings
[11:17] to xAI
[11:19] the training data,
[11:20] a lot of enterprise clients, and the
[11:23] experience in coding. And I think this
[11:25] will accelerate xAI in this area.
[11:27] >> Sacks, you think they're going to dump
[11:29] Kimmy K2
[11:31] .6?
[11:32] Cuz I think Cursor Composer 2 uses the
[11:35] Moonshot model. There's no reasonable
[11:37] way that Elon's going to pay $60 billion
[11:39] and not run on top of Grok. I got to
[11:41] think.
[11:41] >> It seems like it.
[11:42] >> Yeah.
[11:42] >> Likely, but I don't know.
[11:43] >> I think it might be tough depending on
[11:45] the the users. One of the things that
[11:47] makes Cursor so good is
[11:48] >> wait. Say Say more on that. What do you
[11:49] mean?
[11:50] >> So, I think that the different
[11:52] developers
[11:53] want to have choice in that sense.
[11:54] There's a toggle. So, one of the things
[11:56] that's really good about Cursor is
[11:57] they've got this very well-built out
[11:59] IDE, this application layer that puts
[12:02] them probably from a UX perspective,
[12:05] meaning developers are using the tool
[12:07] above Codex, above Claude, above
[12:11] anything else. You can use a third-party
[12:13] IDE and integrate the models or
[12:16] integrate whatever other third-party
[12:18] service you're using. But I would
[12:19] imagine that the developers are going to
[12:21] want to
[12:23] continue to have at least some choice on
[12:25] what's actually writing the code for
[12:26] them. The thing that people are waking
[12:28] up to in the last 120 days
[12:30] is just how much of the value of AI is
[12:35] being realized by writing software. And
[12:38] we've kind of got this wrapper term we
[12:39] call it agents. But agents are
[12:41] fundamentally just quickly spun up
[12:43] applications.
[12:45] But for all of them, as we're realizing
[12:47] very quickly, you end up making too many
[12:48] agents. They end up being super
[12:50] inefficient. They need to be engineered.
[12:52] And you still need to have a strong
[12:53] software engineering capability and
[12:55] competency to fix all the agents, to
[12:57] build all the harnesses, to make
[12:59] everything work well together. And
[13:00] that's why having a strong developer
[13:02] environment, a strong IDE, actually
[13:05] solves that biggest problem. So,
[13:06] eventually all the enterprises that are
[13:08] getting hot and heavy on agents are
[13:09] going to be like, "Whoa, wait a second.
[13:11] We've actually got to fix how this is
[13:12] all being done." As we saw this week in
[13:14] that story with Amazon, where there's
[13:16] like a million agents being spun up
[13:17] inside and everything's wasting
[13:18] resources, redundant data creation,
[13:21] redundant data stores, redundant API
[13:23] calls, et cetera. Tons of money being
[13:25] wasted. So, you have to centralize
[13:27] still. You have to have good software
[13:28] engineering talent that's making good
[13:30] infrastructure and good use of these
[13:32] agents. And that ultimately will require
[13:34] an integration of the AI tooling with a
[13:37] standard software engineering front end,
[13:39] which is the IDE that Cursor has. So, I
[13:41] think that that's probably where
[13:43] everyone's waking up to the fact that
[13:45] having the the software engineers
[13:48] may end up winning you the arms race
[13:50] here. And seems pretty smart for Elon to
[13:52] buy Cursor.
[13:53] >> One other piece of it, you mentioned K
[13:54] me uh K2.6. Yeah, I mean, so I think
[13:58] that one of the things that's going to
[13:59] become a priority over the next several
[14:00] months is this idea of optimizing
[14:03] because enterprises token bills are
[14:05] going through the roof right now. I
[14:07] mean,
[14:08] this month over month they're spending
[14:10] increasingly large amounts because their
[14:12] employees are just building more and
[14:13] more software, but I'm not sure that
[14:15] anyone's been incentivized yet to be
[14:17] efficient about it. And it really only
[14:19] makes sense to go to a frontier model
[14:22] for a frontier task, but more mundane
[14:24] things can be done using an open source
[14:26] model or less expensive model. And I
[14:29] think like you're saying, whether it's
[14:30] the IDE or something else, there needs
[14:32] to be some sort of
[14:33] middleware that determines which model
[14:35] you go to and how much you're willing to
[14:36] spend and what the most efficient way of
[14:39] getting the tokens is going to be.
[14:41] >> I am deep in playing with XAI's suite of
[14:44] products and I would predict
[14:47] we're going to be sitting here in 6 to
[14:49] 12 months and they are going to be
[14:50] dramatically dramatically improved.
[14:53] >> Let me just flag one other area that I
[14:55] think is maybe the white-hot center
[14:58] within this red-hot area of [snorts]
[15:00] coding,
[15:01] which is cyber.
[15:02] And I think Mythos has kind of woken
[15:04] everybody up to the potential of
[15:07] frontier models to be a weapon that can
[15:11] be used by either cyber offense or cyber
[15:14] defense.
[15:15] Now, the issue with Mythos is that it's
[15:17] very large and expensive. It's something
[15:19] like a 10 trillion parameter model. And
[15:21] there's a lot of reports that Anthropic
[15:23] just doesn't have enough compute to be
[15:24] able to serve it. I'm not sure it was
[15:25] ever built to be a commercial model, to
[15:27] be honest, because I just think it's so
[15:29] big and expensive, but
[15:31] I think what will happen is
[15:33] these companies will start training
[15:34] dedicated cyber models.
[15:36] They'll say Mythos comparable models,
[15:39] but with a lower token cost, and
[15:42] I think there's a real race on right now
[15:44] to get those products to market, because
[15:46] I think IT departments and CISOs are
[15:48] very worried about the risk of hacks
[15:51] right now, AI-powered hackers. So, this
[15:54] is something I think over the next 3-6
[15:56] months will be, again, this maybe the
[15:58] hottest part of the market.
[16:00] >> Polymarket says all of this is fake.
[16:03] Complete SpaceX acquiring Cursor, 74%
[16:06] chance. SpaceX IPO by the end of August,
[16:10] 80% chance. So, this is happening,
[16:12] folks.
[16:13] All right, let's keep it moving.
[16:15] >> I think that deal structure is smart,
[16:16] because, I mean, to Chamath's point,
[16:18] yeah, it prevents the IPO process from
[16:20] being disrupted. Also,
[16:22] it kind of gives a huge motivation to
[16:23] these Cursor guys to bust their ass and
[16:25] make it work over the next, I don't
[16:26] know, 6 months. Yeah, they have a $10
[16:28] million break-up fee, but I'm sure they
[16:29] want the deal to be successful.
[16:31] >> Well, the $10 million break-up fee will
[16:32] go back to SpaceX anyways, because if
[16:35] they actually run the compute and
[16:37] they're not owned by SpaceX, they're
[16:38] going to have to pay for it. That is not
[16:40] cheap.
[16:40] >> I mean, we saw a bunch of these
[16:42] xAI co-founders leave after the
[16:44] acquisition by SpaceX. I don't know if
[16:46] that was the reason why, but, you know,
[16:48] all of a sudden they're sitting on
[16:49] SpaceX stock, and they may have felt
[16:51] like
[16:52] they had it made, you know?
[16:53] >> Well, it's actually
[16:54] >> always a problem. It's always a problem
[16:56] with with M&A.
[16:57] >> This Cursor thing came about pretty
[16:59] quickly, because
[17:01] Okay. Let's just say friends of ours who
[17:03] were supposed to wire into that round
[17:05] were like, "Where's the wiring
[17:06] instructions?" It all just evaporated.
[17:09] Here's a tweet from Elon. We don't have
[17:10] to um speculate too much. Here's Sachs.
[17:14] He was very clear that uh xAI wasn't
[17:18] um built right the first time around.
[17:19] The quote, "XAI was not built right
[17:21] first time around, so is being rebuilt
[17:24] from the foundations up." Same thing
[17:26] happened with Tesla, and uh that tweet
[17:28] is from
[17:30] about 5 weeks ago.
[17:31] >> How crazy is it then when he tweets he
[17:32] gets 50.8 million views?
[17:35] >> It takes the four of us 7 months to get
[17:38] 50.8
[17:40] >> our collective
[17:42] >> It's unbelievable the distribution he
[17:43] has.
[17:44] >> Well, also, how many CEOs would just
[17:46] fess up like that and say, "Yeah, that
[17:47] we didn't do it right the first time and
[17:48] now we're rebuilding it." I mean, most
[17:50] of them are not willing to say that.
[17:52] >> He's a magnet for talent. He's a magnet
[17:53] for the right kind of talent, and the
[17:55] SpaceX talent has his philosophy. He
[17:58] inherited, I think, a lot of
[18:00] you know, maybe uh people for XAI or for
[18:03] Twitter that were not in his mold, and
[18:06] they're clearly getting aligned. And
[18:09] it's also going to make his day-to-day
[18:10] life much easier when all of these
[18:12] things are occurring in the same
[18:14] building with the same team. The
[18:15] continuity of not having to task switch
[18:17] between companies is going to be great.
[18:19] We talked a little bit about the
[18:20] possibility of Tesla and SpaceX merging.
[18:22] >> Even Walter Isaacson now is on the Tesla
[18:25] SpaceX merger train.
[18:26] >> Mm, there you go.
[18:28] >> He just did a pod.
[18:28] >> Everybody's confirmed it. It's going to
[18:30] happen. We called it here first.
[18:32] >> Okay, topic two. Is there a SaaS debt
[18:35] bomb in private equity?
[18:37] Thoma Bravo, we had Orlando Bravo at the
[18:40] fourth All-In Summit last year, is
[18:43] nearing a deal
[18:45] to hand its company Medallia over to its
[18:49] creditors. This is a SaaS
[18:51] for customer experience company. TB
[18:53] acquired them in 2021 for 6.4 billion
[18:57] all cash
[18:58] at the top of the market.
[19:00] As part of the deal,
[19:02] they incurred 3 billion in debt, and for
[19:04] background in 2021, this company had 470
[19:07] million in revenue growing 20% a year.
[19:10] Earlier this month, Bloomberg reported
[19:13] that TB's debt servicing costs for
[19:15] Medallia were about to triple from a
[19:18] hundred million a year to three hundred
[19:20] million a year. Blackstone and other
[19:21] firms refused to extend
[19:24] a lifeline to the company to the SAS
[19:26] company. So, it looks like Thoma Bravo
[19:28] just handed the keys back and wiped out
[19:31] 5.1 billion in equity. Chamath, your
[19:33] thoughts? We've been talking about the
[19:35] SAS headwinds for a bit. You've been
[19:36] quite vocal about it.
[19:38] >> Well, first of all, I think Thoma Bravo
[19:40] is an unbelievably well-run
[19:43] organization. Their returns are bonkers,
[19:45] and Orlando is
[19:47] really, really, really good investor.
[19:49] So, what do I think happened?
[19:53] I suspect that they probably got enough
[19:56] of their equity, if not all of their
[19:58] equity.
[19:59] There's probably a decent chance that
[20:01] they did at least one or two dividend
[20:02] recaps in the last five years.
[20:05] And if I had to guess, I suspect that
[20:07] they are
[20:09] positive return. It may not be the
[20:10] return that they would want. And so,
[20:13] turning the keys over becomes easier cuz
[20:14] you have to remember,
[20:16] in private equity, the entire playbook
[20:18] is for transformations of assets that
[20:21] are at some point not working.
[20:24] Right? It's very rarely that they're
[20:26] buying the same kinds of businesses that
[20:27] the four of us would buy, which is just
[20:29] sort of this, you know, clean white
[20:31] sheet de novo grow at all costs kind of
[20:34] business. So, they have operating
[20:36] partners and all of these other people
[20:37] waiting in the wings
[20:39] to un-
[20:40] situations. That's the whole playbook.
[20:43] Mhm. So, to turn it over, I suspect
[20:46] means that there is a core rot
[20:48] that people couldn't fix combined with
[20:52] the fact that they have probably gotten
[20:54] enough downside protection that it's not
[20:56] a huge thing for them. Now,
[20:58] this is an issue for the bondholders,
[21:00] and then that'll maybe flow through to
[21:02] the borrowing costs that Thoma Bravo has
[21:04] to pay maybe for a a
[21:05] deal. I don't know, but I doubt that
[21:08] they would just walk away from a
[21:09] business. So, I suspect they probably
[21:11] got most of their money out.
[21:12] >> I don't know if that's true. There was
[21:14] someone that published some internal
[21:16] data showing
[21:17] that the sales team was like 18% of
[21:20] target at Medallia.
[21:22] You guys know what this company does?
[21:23] Medallia?
[21:24] >> Customer support is the general arena
[21:27] and customer experience.
[21:28] >> Customer experience management. I don't
[21:30] know what I don't know what that means.
[21:32] >> Yeah, they'll basically send like you go
[21:33] on Caribbean cruise ships and you get a
[21:35] survey afterwards and then they use that
[21:36] survey data to provide management
[21:39] insights and operational insights to the
[21:41] leadership team and the operating team
[21:42] on how to improve the quality of their
[21:43] product or their service. So, it's sort
[21:45] of like this feedback surveying loop.
[21:47] So, if I were to tell you guys, "Hey,
[21:49] you want to
[21:50] build a feedback surveying loop to run
[21:51] your business better, are you going to
[21:53] buy SaaS today or you going to ask your
[21:55] AI to spin up an agent for you to do
[21:57] that?" And I think that's a big part of
[21:59] what's happened is all these sorts of
[22:00] companies where the alternative to
[22:03] buying a SaaS product is to spin
[22:05] something up internally and it's much
[22:07] cheaper and easier to spin it up
[22:08] internally. You get a custom workflow.
[22:10] >> I agree with that. But, I'm just saying
[22:11] in the last 5 years, you think they sat
[22:13] on their hands and didn't take a dollar
[22:15] out? They're not that dumb.
[22:17] >> Maybe they took cash out, maybe they
[22:18] didn't, but there was still a big debt
[22:19] overhang and the debt's clearly
[22:21] um gotten impaired, which means the
[22:23] equity
[22:24] >> The debt holders The debt holders are
[22:26] clearly screwed here.
[22:27] >> Yeah.
[22:27] >> The question is is Thoma Bravo screwed?
[22:29] And I would say, if you sat around for 5
[22:31] years, they're That's not their style.
[22:32] They generate too much money. They're
[22:34] too good.
[22:35] >> So, they may have taken cash out and
[22:36] covered some of their their costs, but
[22:38] the equity got fully impaired and then
[22:39] the debt is clearly impaired cuz you can
[22:41] see how the debt and
[22:42] the uh CLOs are trading, which indicates
[22:45] that this business is just not doing
[22:47] well. And then someone else on Twitter
[22:48] posted some internal information from
[22:51] Medallia saying the sales team is just
[22:53] not hitting their targets. They're like
[22:54] way, way off their sales targets. Which
[22:57] I think speaks to the underlying problem
[22:59] here.
[22:59] >> Yes.
[23:00] >> Which is that all of these
[23:01] >> that. Yeah, yeah, please. Yeah, so the
[23:02] underlying problem is that these
[23:04] businesses in the SaaS space where
[23:06] you're driven by net new sales every
[23:08] year, how many new customers are you
[23:10] signing up and then you're trying to
[23:11] manage retention and you're trying to
[23:13] increase sell-through and retain
[23:14] customers, they're just having a really
[23:16] hard time sourcing new customers and
[23:18] there's probably higher than model
[23:19] attrition. That's right. And when you
[23:20] have a very kind of typically
[23:21] historically predictable business where
[23:24] you can say, "Hey, I've got a net
[23:25] revenue retention of 118%
[23:27] or what have you, meaning I'm I'm
[23:29] selling into my install base by 18% over
[23:32] what I'm making last year and then I'm
[23:33] signing up new customers,
[23:35] you can lever that business, right? You
[23:37] can borrow money against those cash
[23:39] flows because it becomes predictable.
[23:41] And what's happened in the last year in
[23:42] particular is agents have become so good
[23:45] and so fast and so cheap that many
[23:48] enterprises can simply spin up an
[23:50] alternative to a vertical SaaS solution
[23:52] and that's crushing the sales team's
[23:53] ability to sell in. That's who you're
[23:55] competing against. Now, I want to make
[23:57] one point and just link this with
[23:58] something else that happened this week
[23:59] and that's Kevin Warsh's hearing
[24:02] for Fed Reserve chair.
[24:03] Kevin Warsh went and talked a lot about
[24:06] the deflationary effect of AI. And I
[24:08] actually think we all talked about the
[24:10] SaaS apocalypse as if it's the sort of
[24:12] like isolated business phenomenon where
[24:14] these SaaS companies are getting blown
[24:15] up. I think another lens to look at
[24:17] what's going on is the incredible
[24:20] deflation of how much it costs to
[24:22] successfully run a business and you
[24:25] don't have to pay a premium price for
[24:27] SaaS products anymore. Meaning that that
[24:29] piece of the business can suddenly get
[24:31] much cheaper. That AI is delivering on
[24:34] its deflationary promise. I'll just say
[24:37] one thing about what Warsh said. Warsh
[24:39] spoke a lot about the deflationary
[24:41] evolution promised by AI and that he
[24:43] expects that it will drive productivity
[24:45] growth like we've never seen before. But
[24:47] he said, "I don't know what that's going
[24:48] to do to the job market that there may
[24:50] be a dislocation between that
[24:51] productivity growth being realized and
[24:54] how the labor markets are going to be
[24:56] able to respond to those things. But
[24:58] fundamentally, he's saying that we're
[25:00] going to see economic deflation. The
[25:03] problem with economic deflation is that
[25:05] when it occurs, it means some business
[25:07] is seeing their revenue go down. And if
[25:09] that segment of the economy is levered,
[25:12] if they have debt sitting on top of that
[25:14] piece of the economy where it's supposed
[25:15] to always, always, always grow, like a
[25:17] SaaS company's top line is always
[25:19] supposed to grow, suddenly that debt
[25:21] gets impaired, and that can have an
[25:22] economic ripple effect that is adverse.
[25:25] But what he's pointing out is that as a
[25:26] result of deflation, because it's not
[25:29] coming from some cost cutting or
[25:31] economic contraction, what he's saying
[25:33] is that the deflationary forces
[25:35] ultimately lead to economic expansion
[25:37] because other parts of the economy will
[25:39] now grow. So if I can suddenly cut, you
[25:42] know, call it 50% of my SaaS budget, and
[25:44] I can reinvest that capital in other
[25:46] ways of growing my business instead of
[25:48] managing my expenses, all of a sudden my
[25:51] enterprise will grow, and the economy
[25:53] will grow. He also said, just as an
[25:55] aside, and I want to make sure I cover
[25:57] this so so that we're really clear, he
[25:59] said the way that we've been measuring
[26:00] inflation is wrong. And that he doesn't
[26:02] agree with the way the Fed has been
[26:03] measuring inflation because you can do a
[26:05] survey of any household and they'll tell
[26:06] you, "My god, everything's so
[26:07] expensive." So all of the indices and
[26:09] [ __ ] that are being used to
[26:11] calculate an inflation index is
[26:13] completely misrepresenting what the
[26:14] average American is actually feeling.
[26:16] And so he wants to rethink how the Fed
[26:18] is addressing inflation from an interest
[26:20] rate perspective, but he does think that
[26:22] the overall kind of economic picture is
[26:24] one of deflationary pressure and
[26:26] productivity gains coming out of AI.
[26:28] >> Sacks, I'll I'll drop this off to you. I
[26:30] think it's pretty clear what's happening
[26:32] here is that the loss SaaS's loss is the
[26:35] token dealer's gain, right? And startups
[26:38] are now, and we always see that they're
[26:40] the tip of the spear. They're writing
[26:42] their own tools, they're making their
[26:43] own dashboards. I see that every day.
[26:46] And if you look at the SaaS product
[26:47] index, here it is. Salesforce down 32%
[26:52] uh in the past 6 months. Shout out to
[26:54] Bestie Benioff, best guest we've had on
[26:56] SaaS at the summit. ServiceNow down 54%,
[26:59] Snowflake down 43%, Adobe down 33%, and
[27:01] Figma, which had a huge IPO pop
[27:05] and is now down 67%.
[27:08] So, what is the role of venture capital
[27:11] and then private equity in addressing
[27:14] the software market? Software was eating
[27:16] the world. Now tokens are eating
[27:18] the SaaS business and the software
[27:20] business, yeah?
[27:21] >> Well, I'm of two minds about this. I'm
[27:23] going to talk about the opportunity for
[27:25] private equity.
[27:27] Let me just say backing up that
[27:29] historically we only had two good exits
[27:31] for software businesses. One was IPO,
[27:34] the other was M&A. And then these big
[27:36] private equity shops came along and gave
[27:37] us a third potential exit, which is you
[27:40] would sell to them
[27:42] and then they would raise the capital
[27:43] based on, I don't know, 1/3 equity and
[27:45] 2/3 debt. So, it was debt-financed
[27:47] buyouts, which is something that's been
[27:49] around in let's call it the non-tech
[27:50] part of the economy for a long time, but
[27:53] was a relatively new entrant into the
[27:55] world of technology. And the reason for
[27:57] that is that if you're going to
[27:58] debt-finance a purchase, you need to
[28:00] have very stable cash flows
[28:01] because if you miss
[28:03] if your cash flows miss and you can't
[28:05] pay your interest on the debt then
[28:07] you're going to lose all your equity
[28:08] because the debt holders will foreclose.
[28:10] So, in order to do a debt financing of
[28:13] any kind, you have to have very
[28:14] predictable cash flows. And it was
[28:16] believed for a long time that software
[28:18] did have those predictable cash flows,
[28:20] at least for the mature businesses, the
[28:22] ones that were at the stage where they
[28:25] could IPO as a potential alternative.
[28:27] So, it was a very attractive thing. I
[28:29] Like I said, I think it was great to
[28:30] have that third option. I I'm of two
[28:32] minds about where the private equity
[28:35] business is today. On the one hand
[28:37] the pricing now has got to be super
[28:40] attractive for them. I mean, we're
[28:41] seeing public SaaS companies that are
[28:43] doing a billion of ARR with 20% rates,
[28:48] 80% gross margins,
[28:50] and they're trading at three times ARR.
[28:53] You can buy a dollar for 50 cents.
[28:56] So is that an opportunity Sachs you
[28:58] think
[28:59] hit rock bottom you should do a roll up?
[29:02] Well, on the one hand I do think that
[29:04] the pricing has never been more
[29:05] attractive if you're a private equity
[29:07] shop looking at a business like that. I
[29:08] mean those companies used to be valued
[29:10] at 13 times ARR now it's three.
[29:12] I'm talking about like a category
[29:14] leader. Now the downside of that
[29:16] Salesforce is off 9% today. I don't know
[29:19] if you guys saw this but the market's
[29:20] absolutely tanking today after the
[29:23] Medallia
[29:24] announcement came out. Right. Okay, so
[29:26] so that would be like you know I said
[29:27] I'm of two minds about it. So
[29:29] I would be bullish for private equity
[29:31] just based on pricing but the bearish
[29:32] part is that in order for their business
[29:35] model to work you have to have
[29:36] predictable cash flows. You can't have a
[29:38] SAS company go from I don't know 120%
[29:40] net dollar retention one quarter to 80%
[29:44] net dollar retention six months or a
[29:47] year later because a big part of their
[29:48] customer base is attrited to using
[29:51] tokens, right? Or to basically creating
[29:53] some bespoke software. You just said the
[29:54] absolute critical thing in all of this
[29:56] which is you have to have predictable
[29:58] cash flows. I think what happens is when
[30:00] you're a startup
[30:01] you typically have to figure out how to
[30:04] disruptively price to enter the market.
[30:06] So you're like, okay, if I deliver $10
[30:08] of value I'm going to charge a dollar
[30:10] and that's that's the normal playbook
[30:12] like a 10% ratio, right? Of price to
[30:14] value.
[30:15] The problem is when you start to stack
[30:17] venture capital into it and you then you
[30:19] stack growth equity into it what you're
[30:21] effectively
[30:22] creating in in the preference stack of
[30:25] your company
[30:27] is that you are creating a higher return
[30:30] hurdle.
[30:31] Right? You got to clear 300 million, 500
[30:34] million, a billion of pref and then you
[30:35] have to return 15 or 20% on top of that.
[30:38] So what do people do as they raise more
[30:40] money?
[30:42] they increase price.
[30:44] But the problem is at some point when
[30:46] you increase price,
[30:48] you engender a ton of competition and
[30:49] you put a huge target on your back.
[30:52] Private equity is the last stop
[30:54] because when you they come in and they
[30:56] lay in billions and billions of dollars
[30:59] of not just equity, but also debt and
[31:01] that has to then be completely
[31:03] predictable and paid back,
[31:05] their only lever is to raise price. They
[31:07] can never cut price to take share. They
[31:09] don't they can't underwrite that to pay
[31:11] back their debt holders.
[31:12] And so Sachs, part of the big problem
[31:14] here and why nobody wants to touch these
[31:16] companies is that they are overpriced.
[31:19] Yes, they're making a billion dollars of
[31:20] ARR, but the unit cost has gotten out of
[31:23] control. It used to be 10% of value.
[31:26] It's probably now 30% of value and
[31:29] everybody's looking at their contracts
[31:30] thinking, well, when it comes time to
[31:32] our renewal, I'm going to just cut this
[31:35] in half
[31:36] or I'm going to cut this by 2/3 or I'm
[31:38] going to cut this by 75%
[31:40] because the value isn't there anymore.
[31:42] >> Or they can threaten to and negotiate a
[31:44] better deal.
[31:45] >> And it becomes even worse because the
[31:46] minute you make these products headless,
[31:49] right? And you say, I'm just going to
[31:50] communicate with these products via MCP
[31:52] and with agents,
[31:54] you can't charge on a per seat basis.
[31:56] What do you do then?
[31:57] Free book doesn't need 50 seats of
[32:00] you know, work day.
[32:01] He needs two seats because the agents
[32:05] act as the way to write in and out of
[32:06] work day. So he wants to pay for two
[32:08] seats, not 50. And then if you multiply
[32:10] that by a million companies,
[32:12] that's what gets us to this place where
[32:14] it just feels like a falling knife.
[32:16] And I think it comes down to these unit
[32:18] costs. The unit costs and the price to
[32:20] value of these products are out of whack
[32:22] with what the market needs and wants.
[32:25] And until they reset that or you find
[32:27] new products that can do it cheaper,
[32:28] we're not going to get a cleansing and a
[32:30] clearing here.
[32:31] >> Yeah, I think
[32:31] >> way, Salesforce today is down 9% 140
[32:35] billion enterprise value on 15 billion
[32:38] of free cash flow.
[32:40] This thing is trading at less than 10
[32:41] times free cash flow. It's unbelievable.
[32:44] >> I think it might be a bargain to be
[32:45] honest.
[32:46] >> Yeah, it sounds like bargain hunting.
[32:47] And what if Benioff, who's the king of
[32:50] acquisitions, what if he just starts
[32:52] >> Well, we didn't we didn't put that he's
[32:53] a buying his own stock.
[32:55] >> Yeah, and we didn't put this on the
[32:56] docket, but did you guys see his kind of
[32:58] headless product announcement?
[32:59] >> Yeah.
[33:00] >> Do you see this? It's it's actually very
[33:02] he's very smart.
[33:03] >> Yeah, I mean I think it's very smart.
[33:05] >> There's ways that that business can
[33:06] maneuver, right? And I think they're
[33:08] pretty unique. It may be that of all the
[33:10] businesses in the scape, like
[33:12] that you know, the ones that have that
[33:14] scale, that have that multi-product
[33:16] platform, that have a lot of your data,
[33:18] there's a lot of opportunity for them to
[33:21] you know, maneuver their way into an
[33:23] evolution of the business.
[33:24] >> one and cuz like if you look at it and
[33:26] you compare it for example to other
[33:27] companies, I think the Workday response
[33:29] was to say, you can't have an AI
[33:31] interact with us without paying some
[33:33] kind of like toll.
[33:34] >> You're exactly right.
[33:35] >> Whereas Benioff is the exact opposite,
[33:37] which is he's like, okay, we're going to
[33:38] go headless for the whole thing, which
[33:39] is really
[33:40] >> right. You're exactly right. I think
[33:42] that's that's going to be the
[33:42] distinction of the winners here and the
[33:44] losers and are you on the wrong side of
[33:46] this?
[33:46] >> The problem is that we have to figure
[33:49] out what is the bottom clearing price
[33:52] and that has nothing to do with business
[33:54] quality. And so is Salesforce a good buy
[33:57] at 10 times free cash flow?
[33:59] Historical artifacts would tell us a
[34:01] screaming yes.
[34:03] The problem is that if you cut
[34:04] everybody's cash flows off at year five
[34:07] or six or seven,
[34:09] then all of a sudden, I think you see
[34:10] the natural compression to between three
[34:12] and five times free cash flow.
[34:14] And that is nothing to do that's that's
[34:16] crazy.
[34:16] >> That has nothing to do with business
[34:17] quality. That just says you you
[34:19] literally mathematically take year seven
[34:21] through end of the future and you
[34:24] discount it to zero.
[34:25] >> And having free cash flow in a war chest
[34:28] gives massive optionality. We've seen
[34:30] this with Salesforce, we've seen it with
[34:32] Apple, we've seen it with Meta, with
[34:33] Google, with Uber, just having massive
[34:36] free cash flow. When you've got tens of
[34:37] billions of dollars, you can put it to
[34:39] work and you can weather these storms.
[34:42] >> J Cal, I think another another way to
[34:44] think about this is
[34:46] you know, to the question about
[34:47] maneuverability and who has the gumption
[34:49] to make the hard choices right now.
[34:51] >> Yeah.
[34:52] >> Look at Benioff. He's the founder of the
[34:54] company. He's run this thing since its
[34:56] founding decades ago. He is willing to
[34:59] bet it all. He's willing to make the
[35:00] change. And it may be that the index you
[35:02] buy in this era of AI transformation is
[35:05] the index of founders. That the founders
[35:07] who are still running their businesses
[35:09] are going to be the ones who are most
[35:10] likely to see
[35:11] >> They can burn the boats.
[35:12] >> They'll burn the boats. They'll
[35:13] maneuver. They'll make the changes.
[35:14] >> And all and all of the guys who have
[35:16] hired managers to run the business are
[35:18] going to do the things that Chamath's
[35:19] talking about, which is trying to charge
[35:20] fees and try to maintain the old way of
[35:22] doing things as opposed to reinvent for
[35:24] the new future.
[35:25] >> If you look at the 10-Ks, if we could
[35:26] figure out what the unit price cost and
[35:29] the trend and the inflation is of a per
[35:32] seat license for these products,
[35:34] I will point to the ones that are going
[35:36] to die first.
[35:37] >> Chamath, two quick points. So
[35:38] >> Yeah, wrap us up.
[35:39] >> Yeah. One is, yes, I would like fully
[35:41] endorse what what you said about
[35:42] Benioff. He's made every previous wave
[35:44] work to his benefit, whether it was
[35:47] social, whether it was mobile, whether
[35:49] it was big data, all that kind of stuff.
[35:51] What are the odds he's going to make AI
[35:53] work to his benefit? I'd say pretty
[35:54] good. So, his stock might be a bargain
[35:56] right now. So, that is me, point number
[35:59] one. Just And I want to say just a
[36:01] thing about venture debt, which is I
[36:04] look, I think it's fine when private
[36:05] equity guys use it cuz they know what
[36:06] they're doing, but I've always hated
[36:08] when founders take on venture debt. And
[36:10] I know J Cal you agree with me. Part of
[36:12] it is that founders forget that they
[36:13] have to pay it back. They treat it like
[36:15] venture capital and they forget about
[36:16] that. And then they get surprised. But
[36:18] the other thing I've never liked about
[36:20] it is it makes you more fragile. It
[36:22] basically subjects you to a bunch of
[36:24] business covenants and it makes it
[36:26] harder for you to do an abrupt shift in
[36:29] your business because now you've got a
[36:31] bank looking over your shoulder and they
[36:32] want to make sure they get paid, and
[36:34] they have to review your financials and
[36:35] all the rest of it. And to your point,
[36:37] J. Cal, the companies that have free
[36:39] cash flow right now are the ones that
[36:40] have the most maneuverability. I hate
[36:43] taking away maneuverability from
[36:45] founders, and that is what debt does
[36:47] because it subjects you to a fixed
[36:49] schedule of payments. And so, this is
[36:52] always a thing to remember, whether
[36:53] you're a business or you're an
[36:56] individual.
[36:57] You know, when you put on that debt, it
[36:58] makes you more vulnerable to big
[37:01] disruptions in the market.
[37:03] >> Yeah, it just you become incredibly
[37:05] brittle, and founders who are listening,
[37:07] when you get that in in peak markets,
[37:09] peak zero
[37:10] you're going to have venture debt people
[37:11] offer you tons of cash, and then the
[37:14] problem David and I saw up close and
[37:15] personal in many different companies
[37:17] where the founders would look at it as
[37:19] like, "Oh, I'm extending my runway."
[37:20] Well, if you're a hot startup, there's
[37:21] always more venture capital, there's
[37:23] always more people who want to own
[37:24] equity. The equity sale gives you
[37:26] optionality, and you have more people on
[37:28] your team, more people rooting for you,
[37:31] and aligned with equity interest, as
[37:33] opposed to now you have a debt
[37:35] instrument, they have a different goal,
[37:38] they have different downside they're
[37:39] trying
[37:40] >> No bank wants to be your last 3 to 6
[37:42] months of runway.
[37:43] Because that means that in a high
[37:45] percentage cases, they're going to lose
[37:47] their money.
[37:48] >> Yeah.
[37:48] >> So, they're they're built to try to
[37:50] avoid that.
[37:51] >> I've never seen venture debt work well
[37:53] to improve the quality of the business.
[37:55] Never.
[37:56] Never. It doesn't only only ever seen
[37:58] venture debt leave tracks that damage
[38:01] companies. And if you get the venture
[38:03] debt, you can never actually use it. So,
[38:05] the venture debt investors that
[38:06] ultimately make money, it's because they
[38:08] put money in a company and the company
[38:10] never actually used the money they gave
[38:11] them. I hate this business. I think
[38:13] venture debt's like the worst
[38:15] vulture-like business in Silicon Valley.
[38:17] It's terrible.
[38:18] >> If you get down if if the last money in
[38:20] the bank is the debt you owe to the
[38:22] bank, they're going to rug you. That's
[38:24] when you get rugged.
[38:25] >> 100%
[38:26] >> You think they can afford to lose 100%
[38:29] of their money when they're getting an
[38:31] 8% return or something like that? No
[38:33] way. That's not how it works. VCs can
[38:36] afford that because we have the
[38:38] opportunity for
[38:39] a 10x or 100x or 1000x for that
[38:41] moonshot. So, we can accept a bunch of
[38:43] zeros. The bank can't accept a bunch of
[38:44] zeros.
[38:45] >> Well, and then when they when they do
[38:47] get scared and when they do think
[38:48] they're going to lose their money, wait
[38:50] till you see what they extract in terms
[38:52] of value, what they ask for. They will
[38:54] ask they'll double the interest rate,
[38:57] they'll ask for warrants. It's basically
[38:59] like being in debt in prison. Chamath,
[39:02] you can talk a little bit about your
[39:03] experience when you were
[39:04] >> [laughter]
[39:05] >> in debt in prison.
[39:06] It's not going to be pleasant. I've been
[39:08] in debt. I mean, I've had a $420 million
[39:11] credit line.
[39:13] And I had a moment where it was
[39:15] reflexively kind of collapsing inward
[39:18] because the assets that I was using to
[39:19] secure it shrank in value in a moment of
[39:22] market disruption.
[39:24] I was scrambling and then at the same
[39:26] time
[39:27] there was a risk It was the worst moment
[39:29] of my professional working life. I had
[39:31] like a couple hundred million dollars
[39:32] sitting at Credit Suisse and they were
[39:34] about to implode. And so, on a weekend,
[39:36] I was trying to figure out whether my
[39:38] money was still there.
[39:40] I had always had this rule, don't have
[39:42] debt and then I violated it to try to
[39:44] run the number up.
[39:45] I almost got run over. I almost lost
[39:48] everything.
[39:50] I will never do it again. And if I ever
[39:52] do it again, if you guys ever hear me do
[39:53] it again, please just come and punch me
[39:55] in the face.
[39:56] >> Oh, we will. We've been waiting for an
[39:57] excuse.
[39:58] >> Can we punch you in the face for other
[39:59] things, too?
[40:00] >> Buffett has this line about how smart
[40:02] guys go bankrupt is they take on debt.
[40:06] >> Debt equals prison, [ __ ] Keep it in
[40:08] your mind, guys.
[40:09] >> Unless you socialize the debt and then
[40:11] everyone thinks it's okay, which is what
[40:13] we do with governments and that's the
[40:14] problem with governments.
[40:16] >> He's going to start going off.
[40:17] >> Yeah.
[40:18] >> Don't push the button. All right,
[40:19] listen. We got to talk about TJ.
[40:21] >> just just just a quick aside on that. In
[40:23] the 1950s, all the corporations in
[40:25] America had pension plans where you
[40:27] would get some guaranteed payout at the
[40:28] end when you retire. And they were all
[40:30] like, "We're all going to go bankrupt."
[40:31] Because a pension plan is either
[40:32] significantly overfunded or underfunded.
[40:35] If it's underfunded, you're bankrupt. If
[40:36] it's overfunded, you've wasted all this
[40:38] money, you can't do anything with it.
[40:39] So, they all moved to 401(k)s and
[40:41] everything got moved to defined
[40:42] contribution plans except except
[40:45] governments. And that's because the
[40:47] government employees formed government
[40:49] public employee unions and they're like,
[40:51] "We want to keep the pension plans." And
[40:52] now the pension plans, it turns out 70
[40:54] years later, are going to bankrupt all
[40:56] the governments in the United States.
[40:57] >> That guy Spencer By the way, that guy
[40:59] Spencer Pratt who's running for mayor,
[41:01] he started uncovering all of the
[41:03] salaries of the union folks and their
[41:06] pensions in Southern California. It's
[41:09] bonkers. They're making four, five
[41:10] hundred thousand dollars a year right
[41:12] before they go on pension. Then they
[41:14] double their overtime and they get
[41:15] two-thirds or a half. The pension
[41:17] doesn't work. You got to go
[41:18] superannuation fund. I don't know how
[41:20] many times we've talked about it here,
[41:21] but
[41:22] >> You you don't need an annuitization
[41:23] fund. You just need a 401(k). Like
[41:25] people have a have an account. They got
[41:26] their money in their account.
[41:27] >> Yeah, but it's just a way to force
[41:29] people to contribute to it. So, a forced
[41:30] 401(k) is different than a 401(k). You
[41:33] got to force people and it's you're not
[41:34] allowed to force people into the 401(k),
[41:36] as you know.
[41:37] >> As we've seen in California, everything
[41:40] related to the government is a giant
[41:41] grift, is a giant scam. There's tons of
[41:43] fraud going on. Uh we've talked about
[41:46] the homeless industrial complex, 12
[41:47] billion a year to homelessness, but the
[41:49] number of homeless keeps going up.
[41:51] There's a million examples like that.
[41:53] >> industrial complex?
[41:54] >> Yeah, maybe it's a good time to let's
[41:56] shift to SPLC cuz I think it's a good
[41:57] example.
[41:58] >> Well, we'll get to it. We'll get to it.
[41:59] Um yeah, but you know what's even better
[42:01] is you can just pass a law like the Nick
[42:02] Sherly Act and you can put your fingers
[42:04] in your ear, cover your eyes and say la
[42:05] la la la la la and just pretend the
[42:07] fraud's not happening, which is their
[42:08] reaction in California.
[42:10] >> Uh hey, how much fraud has Nick
[42:12] uncovered so far in California?
[42:14] >> Billions. He should be a billionaire.
[42:16] >> Where's his
[42:17] >> a billionaire.
[42:18] >> You know, he should be doing it
[42:19] privately and then getting the
[42:20] whistleblower awards. I think that
[42:22] actually would be a better strategy for
[42:24] him.
[42:24] >> We told him that was the We told him
[42:26] that business model, remember?
[42:27] >> Yeah, I know. He I think he's addicted
[42:29] to the views, but I mean, he could
[42:31] literally raise money on
[42:33] >> Well, he's making
[42:34] He's making thousands when he could be
[42:35] making billions to Free Brooks' point.
[42:37] >> No, he's
[42:37] >> Well, but it's better It's better for
[42:39] the public that he's doing what he's
[42:41] doing.
[42:41] >> thank god for Nick Shirley. Let's thank
[42:43] Nick Shirley.
[42:44] >> whether you could be making more money
[42:45] or not, what you're doing is
[42:46] >> And you know what he also did? He shamed
[42:49] the mainstream media who's forgotten
[42:51] about investigative journalism, who
[42:53] forgot the ability to knock on a door
[42:55] and just ask a basic question.
[42:57] And now Bari Weiss with CBS has
[43:00] deputized one of her reporters and she's
[43:02] doing the exact same playbook and
[43:05] meeting him punch for punch.
[43:07] Where's CNN? Anderson Cooper should have
[43:09] a Nick Shirley on his team. The New York
[43:11] Times should have Nick Shirley. The LA
[43:13] Times should have a Nick Shirley. Why
[43:14] don't they?
[43:15] >> Cuz you're
[43:17] You're talking You're talking about old
[43:18] media that does things one way and the
[43:20] point about Nick Shirley is it's new
[43:21] media. It's citizen journalism. It's
[43:23] people on the street distributing
[43:25] fact-finding, distributing information
[43:27] gathering and old media in order to
[43:28] survive became an opinion organization.
[43:30] >> Didn't the media used to care that the
[43:32] Pentagon was paying $900 for a hammer or
[43:35] what have you? I mean, like 60 Minutes
[43:36] used to do things. Now, it's like the
[43:38] media just wants to protect the waste,
[43:41] fraud, and abuse
[43:43] no matter how egregious it is. Remember,
[43:45] I mean
[43:45] >> Remember that guy, Dennis Kozlowski, the
[43:48] CEO of Tyco, who went to jail and they
[43:50] like had it just a feeling because he
[43:53] had
[43:53] >> He umbrella stand the like the $6,000
[43:56] umbrella stand.
[43:57] >> Made out of like ivory from like
[43:59] >> [laughter]
[43:59] >> a rhino elephant.
[44:01] >> I know you bought that at auction,
[44:02] didn't you, Jim Matt?
[44:03] >> Sachs, you're so right. People really
[44:05] used to care except when it was their
[44:07] team. And then the minute that it was
[44:08] their team, they're like, "Oh, no, no,
[44:10] no, let's just look the other way."
[44:11] >> It was a
[44:11] >> If it's a CEO, if a CEO basically
[44:14] engages in some misbehavior, and I'm not
[44:16] defending it, the press will be all over
[44:17] that. But when the government does it,
[44:19] they don't do anything.
[44:21] And in fact, we had one of the most
[44:23] successful, probably the most successful
[44:25] entrepreneur of our generation donating
[44:28] his time to the government to find
[44:30] waste, and the media basically drove him
[44:33] out of that.
[44:34] >> They vilified him and drove him out.
[44:35] >> Yeah, they made it untenable.
[44:36] >> Well, this is whoever comes up with a
[44:39] way to eliminate waste, fraud, and
[44:40] abuse, like Chamath did, and they
[44:42] productize that and make them that make
[44:44] that their platform, that's the way to
[44:46] win in 2028 and going forward is to
[44:49] convince the public that they don't need
[44:52] to have their taxes raised, they could
[44:54] have their taxes reduced just by
[44:56] eliminating the minimum of 20 or 30% of
[44:59] waste, fraud, and abuse there is in the
[45:01] system. Uh we'll get to Tim Cook
[45:03] stepping down in just a moment, but I
[45:04] want to remind everybody
[45:06] liquidity sold out. I'm sorry, we added
[45:08] a couple of tickets. We we we burned
[45:10] through them immediately. But you can
[45:12] still get into the All-In Summit. This
[45:13] is our fifth edition in Los Angeles,
[45:15] September 13th, 14th, and 15th.
[45:18] allin.com/events to apply. Please apply,
[45:21] and then don't come to us 60 days out
[45:23] and say, "I didn't get a ticket.
[45:25] I'm a bestie, get me in." Just buy your
[45:27] damn ticket and don't get left out.
[45:29] >> I have a liquidity announcement.
[45:30] >> Oh, yum, yum.
[45:31] >> We are going to do one
[45:33] political panel.
[45:35] >> Okay.
[45:36] >> And it's going to be
[45:38] Dave McCormick and John Fetterman, the
[45:41] two sitting senators from Pennsylvania,
[45:44] on stage with us
[45:45] talking about all topics from a left and
[45:48] right perspective.
[45:49] >> Amazing. So, Fetterman's coming, which
[45:52] means the dress code is now sandals, uh
[45:55] shorts, and a T-shirt. That's great.
[45:57] >> Construction Construction sheet.
[45:58] >> Get your Timberlands [laughter] out.
[46:01] I mean, is he really going to show up
[46:03] looking like a hobo? I love his hobo
[46:05] style. It's great.
[46:07] >> McCormick's very fit and handsome. So,
[46:08] like he he'll balance him out. So, I
[46:10] think
[46:10] >> That's what we should program it as.
[46:12] Like, he should wear his best Brioni
[46:14] suit versus
[46:16] the Old Navy from Bettermint. Who
[46:18] [laughter] wore it better? All right,
[46:19] listen. Just rapid fire here on the Tim
[46:22] Cook
[46:23] resignation and moving on. This guy,
[46:26] John Ternus, is a 25-year vet. He did
[46:30] lots of hardware, worked on iPad,
[46:32] AirPods, and he was the favorite on
[46:35] Polymarket since day one. He's a bold
[46:37] decision-maker according to reports. And
[46:41] unlike Tim Cook Cook, Tim Cook did a
[46:43] great job of squeezing every last nickel
[46:45] out of Steve Jobs' product line, which
[46:47] lasted for a decade, iPhone, Apple TV,
[46:50] Watch, I don't have to need to repeat
[46:51] them. But, here we are.
[46:53] We got a product person in the seat,
[46:55] which is what we all know they needed.
[46:58] Because, hey, these tools are getting a
[47:00] little bit stale. Siri, disgraziad.
[47:03] AirPods, disgraziad. The whole system is
[47:06] not built on innovation anymore. It's
[47:08] built, Freeberg, I think you would
[47:10] agree, on just bringing more profits,
[47:14] more profits. What's your hope here?
[47:16] Because, man, they missed so many great
[47:18] swings at bat. They didn't get the
[47:20] Oculus, you know, Ray-Bans that Meta
[47:23] did. They canceled their self-driving
[47:25] car. What would you hope that this new
[47:27] CEO of Apple focuses on, Freeberg, in
[47:30] terms of innovation? They don't have a
[47:32] problem selling phones still. They don't
[47:34] have a problem selling laptops and
[47:35] making a ton of money. But, if you were
[47:36] in the seat, if you were on the board of
[47:38] Apple, which wouldn't be a bad idea for
[47:39] them, if I'm being honest, what would
[47:41] you tell the new CEO to focus on, David
[47:43] Freeberg?
[47:44] >> I mean, I don't know. The software layer
[47:46] of the future is not the software layer
[47:48] of the past. So,
[47:49] >> Okay.
[47:50] >> It's pretty obvious. I don't know if how
[47:52] much there is to talk about, but you
[47:53] just need the Siri equivalent that's
[47:55] ubiquitous in all of your devices, knows
[47:57] who you are, personalized to you, sees
[47:59] your email, sees your calendar entries,
[48:02] knows what kind of music you like, has
[48:03] connection to your home. Basically,
[48:05] build that AI layer for your life. And
[48:08] make it ubiquitous in all of your Apple
[48:10] devices, that no matter what device
[48:11] you're using, it knows who you are, you
[48:13] can engage with it using kind of a
[48:15] natural language method. And it's, you
[48:17] know, it's it's pretty obvious.
[48:19] >> Yeah.
[48:20] They should buy WhisperFlow, yeah. I
[48:21] mean, that would be
[48:22] >> I don't know how they're I don't know
[48:23] how they're running the business, but
[48:24] >> Well, they're running it for profits,
[48:26] obviously, Sachs. I would say buy
[48:28] WhisperFlow and just replace the Siri
[48:30] team with that, because Siri's been just
[48:32] The fact that Siri can't spell
[48:33] Polyhymnia or Calacanis after 20 years
[48:36] of us giving them $20,000 for iPhones is
[48:39] just disgraceful. Sachs, if you were on
[48:41] the board of Apple, again, not a bad
[48:43] idea, what would be your hope for the
[48:46] company? What would be your sage advice
[48:48] for the new CEO?
[48:50] >> Well, I mean, everybody is going to be
[48:52] asking the same question, which is what
[48:54] are you going to do about AI. I don't
[48:55] know that they needed to be on the
[48:57] bleeding edge of it, but they are going
[48:58] to need an answer at some point, and
[49:00] Siri's going to need to be AI-empowered.
[49:03] Probably the way it should work is that
[49:04] you get to choose your model. I mean, I
[49:06] don't know that they need to pick just
[49:08] one model provider. It could be a
[49:09] setting where you go in and you set up
[49:11] your account with whatever ChatGPT or
[49:14] Grok or Claude or what have you. And you
[49:17] can choose your own model provider, and
[49:18] then you'll have more customization and
[49:20] more
[49:21] ability to control your your storage.
[49:24] Let me just say this on Tim Cook's
[49:27] retirement, he had an incredible run as
[49:30] CEO of of Apple. I mean, he ran it very
[49:32] effectively for 15 years. The market cap
[49:35] of the company went up by over 10X, the
[49:37] revenue grew from roughly 100 billion a
[49:40] year to over 400 billion a year.
[49:42] He also improved the quality of revenue
[49:44] by moving the mix into services, which
[49:47] is probably why it got why it got a
[49:49] higher valuation.
[49:51] And, you know, people say that well,
[49:53] they never did any innovation under Tim
[49:55] Cook, but you know, I've seen people
[49:56] tweet list of products that were
[49:58] released under him.
[49:59] And there were a lot of them. Now,
[50:02] it's true, nothing as big as the iPhone,
[50:04] but they did release a lot of products
[50:06] under Tim Cook. And then just finally, I
[50:07] mean, you you look back over the last 15
[50:09] years and there really weren't any
[50:12] public snafus or scandals or imbroglios
[50:17] with with Apple. It's one of the few
[50:19] tech brands that is still, I think,
[50:21] beloved by the population.
[50:23] I think a major part of that was Tim
[50:25] Cook's dedication to privacy and keeping
[50:28] the company on the right side of that,
[50:29] which I think users do appreciate. And,
[50:32] you know, he even Tim Cook even got
[50:35] praise from the president. I think it
[50:36] was just unsolicited where the president
[50:39] talked about how Tim Cook didn't call
[50:40] him up that often, but when he did, it
[50:42] was something important and therefore
[50:44] the president tried to help them out.
[50:46] Seems like he nurtured a good
[50:47] relationship with the president over
[50:48] over the last decade or so. So, you just
[50:51] have to say that he navigated what could
[50:53] have been a turbulent period with a
[50:55] great deal of grace and aplomb.
[50:58] >> Clearly, Chamath, he was a great steward
[51:01] of the brand, even though that list of
[51:04] products were all
[51:06] developed under Steve Jobs and they were
[51:08] just executed well, but he didn't
[51:11] bring in a lot of new products or
[51:13] services. Any final take
[51:15] >> Actually, Chamath, can I kindly ask you
[51:16] a question? What do you think
[51:18] other than AI, you know, AI-powered
[51:20] Siri, let's say.
[51:22] What do you think he missed? I mean,
[51:24] what should Apple have done that they
[51:26] didn't do?
[51:27] >> They would have out by now a pair of
[51:30] glasses that weren't 17 lb like the
[51:33] Apple Vision Pro. They would have gotten
[51:35] glasses that pair perfectly with your
[51:37] phone, take videos for kids, and they're
[51:39] coming out with it. It's just on a
[51:41] really broken timeline. They would have
[51:43] had a killer Siri.
[51:45] They would have had a search engine-ish
[51:47] Perplexity like product. They would have
[51:49] had a self-driving car. When you went to
[51:51] the Apple Store, you would have been
[51:52] buying two or three very important
[51:55] products. Glasses, a car, and probably a
[51:58] television set. If you look at actually
[52:00] what they did innovative under Tim Cook,
[52:02] I think that they have great taste and
[52:04] Apple TV produced a lot of great
[52:06] programming. I he was working on a
[52:08] television set. Not Apple TV clunked
[52:11] onto the back. I think those three
[52:13] products would have been four products.
[52:15] Siri, glasses, car, television set.
[52:18] Those would have been extraordinary. And
[52:20] who knows what he would have come up
[52:21] with. When they lost Jony Ive, and
[52:23] obviously Steve Jobs passed away, they
[52:25] lost the soul of the company. They lost
[52:27] the innovative, groundbreaking soul of
[52:30] the company, and they just went into
[52:31] profit
[52:33] and iteration mode. But no acquisitions
[52:36] of note. Nothing important was acquired,
[52:39] and nothing important was released.
[52:41] Vision Pro, you can give them like maybe
[52:43] that's like the sixth best product or
[52:45] something. But it obviously hasn't hit
[52:47] the mainstream. Chamath, any final
[52:49] thoughts from you?
[52:50] >> Yeah, I have um
[52:52] three specific things to say. The first
[52:54] is that he had honestly like an
[52:56] impossible job.
[52:58] It's sort of like you play
[53:01] basketball with Michael Jordan, and then
[53:04] you're asked to be Michael Jordan.
[53:06] And I think that that's an impossible
[53:08] task. And on that dimension, I think he
[53:10] has done just an incredible job. He has
[53:12] been
[53:13] an incredible steward of the business.
[53:15] Sachs is right, no major snafus. I think
[53:19] he did a really smart thing around
[53:21] doubling down on privacy. And just as a
[53:24] practical matter of being a great CEO.
[53:27] Like if you I think you can categorize
[53:29] CEOs in two buckets. One is
[53:32] the innovator,
[53:33] the person that's pushing the envelope,
[53:35] and then the second is just a great
[53:38] steward.
[53:39] He's at the top of the top of that
[53:41] second category. Um I sent you couple of
[53:44] charts to to show this. And he found a
[53:46] lane that allowed him to separate
[53:49] himself from Steve Jobs. So, you know,
[53:52] as an example, like what does it mean to
[53:54] be a steward? Well, when you're a
[53:56] steward, you're allocating resources,
[53:57] and the two most important resources you
[53:59] control
[54:00] is capital
[54:02] and people. And I think on that
[54:04] dimension, what Tim did, if you just
[54:07] look at this, is
[54:08] he was able to invest appropriately in
[54:10] R&D. They completely divested their need
[54:14] of Intel. They spun their entire new
[54:16] line of silicon. That silicon, it turns
[54:19] out, and this will be important in the
[54:20] future,
[54:21] is very useful
[54:23] in AI with all of this open claw stuff,
[54:25] and you know, some of you guys are
[54:27] completely addicted to it.
[54:29] And they've kept the acquisitions light.
[54:30] So, he was very capital efficient. If
[54:32] you look at the
[54:34] the next chart,
[54:35] what's so interesting is like it is the
[54:37] exact opposite of what Steve Jobs did.
[54:39] Look at the amount of money that Steve
[54:41] Jobs returned to shareholders in his
[54:43] tenure at Apple. It's easy to count. It
[54:44] was zero.
[54:46] He loved to to keep that money on the
[54:48] balance sheet, and he probably, or
[54:51] maybe, I'm guessing, would have directed
[54:53] that at some huge shot on goal. In the
[54:55] Tim Cook era, it was very different. He
[54:57] shrank the share count by almost 50%. I
[54:59] think it's like 44%.
[55:00] >> That's insane. Is there any
[55:02] corollary to that, Chamath?
[55:04] >> No, he's he's been a prolific
[55:07] shareholder friendly CEO, finding ways
[55:10] to give us money back, which I think
[55:11] everyone who's owned the stock
[55:13] has very deeply appreciated.
[55:15] The last thing I'll say, though, is what
[55:17] is the future for John Turnus, and I
[55:19] think it's in this final chart. We
[55:20] talked about
[55:22] the problematic nature of increasing per
[55:26] unit pricing in SaaS.
[55:29] And what I would say is if you look at
[55:31] the iPhone,
[55:33] the unit price has gone up, and people
[55:35] would say, "Yes, but the capabilities
[55:37] have gone up in turn, and I acknowledge
[55:39] that.
[55:40] But the problem with the per unit
[55:42] pricing being as high as it is is what
[55:45] Freeburg says is going to happen. AI
[55:47] rips open the canvas
[55:49] of the devices that we will use to
[55:51] interact with information and knowledge.
[55:54] We are going to live in a much more
[55:56] heterogeneous world in the future. It's
[55:58] not going to be two devices and two
[56:00] different operating systems that get you
[56:02] to knowledge.
[56:04] There's going to be all kinds of stuff.
[56:05] Pens, orbs,
[56:07] who knows, Jason? Your glasses,
[56:08] whatever.
[56:10] And so, the problem is if you get too
[56:12] addicted to a single thing
[56:14] that has an incredibly juicy profit
[56:17] margin and great stickiness and the
[56:20] ability to raise price,
[56:22] it's a hard drug to get off of. So, I
[56:24] think really what John Turnus has to do
[56:27] is figure out how to move to this world
[56:29] where everybody will be launching
[56:31] umpteen devices
[56:33] via MCP or otherwise,
[56:36] all of these services will be open.
[56:39] It'll be agentically talking to
[56:40] everything. I think the most decay.
[56:44] And I think if that happens, that's
[56:46] problematic if you're too reliant on a
[56:48] single thing to kind of keep it going.
[56:50] >> Yeah, and just expanding on what you
[56:52] said, like wearables is where they
[56:54] really uh made some good inroads in
[56:57] terms of getting people to use them,
[56:59] whether it's AirPods or the watch. And
[57:01] the next wearable, like this is an
[57:02] applaud pin that I use to record. You
[57:04] can put it here, put it on your wrist.
[57:06] That AI synchronicity of having your
[57:10] eyes, having your ears, having your
[57:12] watch, having your phone, your desktop
[57:14] all sync together with AI could be a
[57:17] huge product line. I'll also add a
[57:18] fifth, robotics. You know, the I think
[57:21] to I think Steve Jobs, if he were alive
[57:24] today, would have been looking at
[57:25] Roomba, he would have been looking at
[57:26] Optimus, and he would have said, "Hmm."
[57:28] >> [clears throat]
[57:29] >> Consumer robotics, in addition to a
[57:31] consumer car, those are two things I
[57:33] think he would have absolutely uh
[57:36] executed on at a high level. Okay, let's
[57:39] keep moving here. Uh
[57:40] >> Just come and come and one last point.
[57:41] >> Chamath, we'd love to have you on the
[57:42] pod, John. So, just come on the pod when
[57:43] you're ready. Uh we'll have you come sit
[57:45] in. Go ahead. Sachs, you get the final
[57:46] word.
[57:47] >> This one last point on this is that I
[57:49] think the succession at Apple
[57:52] is reminiscent a little bit of the
[57:54] succession at Disney.
[57:56] And apparently
[57:58] Steve Jobs and Tim Cook had this
[58:00] conversation back when Steve was alive
[58:01] and and Steve
[58:03] told Tim, "Don't do what Disney did."
[58:05] Where basically after Walt Disney died,
[58:07] the company kind of languished because
[58:10] it felt so beholden
[58:12] to Walt's vision that they never really
[58:14] iterated. Now, when Walt died, his
[58:17] brother Roy took over and Roy was
[58:19] already in the business. He was sort of
[58:20] like the business co-founder. He was a
[58:22] COO type, a little bit like Tim Cook.
[58:25] And he kept the magic going for about 5
[58:26] years and then he himself died. I think
[58:28] it was around 1971. And then you had
[58:31] this string of CEOs who took over were
[58:32] kind of uninspired and it wasn't until
[58:34] Eisner came in in 1984 that he sort of
[58:37] revitalized the business.
[58:39] And so, as I understand it, Steve and
[58:41] Tim had this conversation.
[58:43] And Steve told him, "Don't be too
[58:44] beholden to my vision. You need to
[58:46] figure out your own and extend it."
[58:48] I think that, you know, you could argue
[58:50] that Tim in a way was like the Roy
[58:52] figure here. Very effective business
[58:56] partner
[58:57] of Steve. He got a 15-year run. Roy only
[59:00] got five and I think again, he added a
[59:04] zero to the value of the company. The
[59:06] market cap went up over 10x. So, you
[59:09] have to say fantastically successful run
[59:11] as CEO. I think the question now for
[59:14] John Turnus is, "Okay, you're now past
[59:17] let's say the the Walt Disney and Roy
[59:19] Disney part of the business. Is it going
[59:21] to be like the 1970s
[59:23] Disney or is it going to be more like
[59:24] the 1980s? Do you figure out a way to
[59:26] revitalize it or do you have to go
[59:28] through
[59:29] uh you kind of a funk first?
[59:31] >> Mhm. Yeah, and if I I think it's like
[59:34] really illustrative of this discussion,
[59:37] Eisner and Iger. Cuz Eisner's innovation
[59:40] was he realized that Disney was uh I
[59:43] think he called it the vault strategy.
[59:45] He would and we probably remember this
[59:47] from our childhoods, he would re-release
[59:50] all into theaters all of their IP every
[59:52] 7 years. You couldn't get some of those
[59:55] Bambis, whatever.
[59:57] Snow White and the Seven Dwarfs, you
[59:58] couldn't get those products except in
[60:00] theaters and he figured out a cadence to
[60:02] keep publishing. But then Iger came in
[60:04] and said, "Hey, what if we use this
[60:06] balance sheet and we use this
[60:07] distribution at the parks uh and you
[60:10] know, with their brand
[60:12] to buy Pixar, Marvel,
[60:14] uh and Star Wars." And so, there's
[60:17] multiple ways to do it. There might be
[60:18] something there in terms of
[60:20] acquisitions, bold acquisitions with the
[60:23] Apple balance sheet could be super
[60:25] accretive to shareholders as opposed to
[60:27] lowering share count and just
[60:30] distributing a ton of cash. All right,
[60:31] listen. We're going to talk about the
[60:33] Southern Poverty Law Center racism
[60:36] corner. Let's go to race Fake fake
[60:38] racism corner.
[60:39] >> I want to know how the SPLC managed to
[60:42] accumulate 822
[60:44] million dollars in offshore bank
[60:46] accounts.
[60:46] >> Yeah.
[60:47] >> This is incredible. These are big
[60:48] numbers. Okay, SPLC
[60:50] >> How was that possible? What is going on?
[60:52] >> All right, let me tee it up here for the
[60:53] team.
[60:54] >> It's like one of the biggest grifts of
[60:55] all time.
[60:56] >> Anyway, Jake has to do it. This is a big
[60:57] one. SPLC has been indicted indicted,
[61:00] not found guilty yet, on 11 counts of
[61:03] wire fraud and money laundering. Keep
[61:05] that in the back of your head. Wire
[61:06] fraud and money laundering. Here's the
[61:08] core allegation. Between 2014 and 2023,
[61:11] the Southern Poverty Law Center used
[61:14] hidden bank accounts to funnel 3 million
[61:15] in donor money to paid informants.
[61:18] Like these are confidential informants,
[61:20] like the police or FBI might use. They
[61:23] use these as a non-profit NGO to
[61:27] infiltrate hate groups. And so, the
[61:30] official mission of the SPLC
[61:32] is, quote, to be a catalyst for racial
[61:35] justice in the South and beyond, working
[61:37] in partnership with communities to
[61:39] dismantle white supremacy, strengthen
[61:41] intersectional movements, and advance
[61:44] the human rights of all people. Okay,
[61:45] sounds great on paper. Examples of
[61:47] organizations they were trying to
[61:49] infiltrate great KKK Aryan Nation
[61:52] neo-Nazi groups, and the Unite the Right
[61:54] organizers. Proud Boys, labeled as a
[61:56] hate group by the SPLC. Oath Keepers,
[61:58] not listed as a hate group, but part of
[62:01] the militia movement. My friend Sam
[62:03] Harris, he was not listed as a hate
[62:05] group, but he was also pinned by the
[62:07] SPLC as, like, hate adjacent in their
[62:10] Hate Watch headlines. And this is
[62:12] something that I had a major problem
[62:13] with this organization on, which is they
[62:15] would just very loosely label people as
[62:19] hate speech and try to get them
[62:20] canceled. All of this kind of came to a
[62:24] head. The revenue before
[62:25] Charlottesville, you remember the
[62:27] incorrectly clipped uh Charlottesville
[62:29] hoax, where they said Trump said both
[62:31] good people on both sides, but they
[62:32] didn't give his full quote. Very unfair
[62:34] to President Trump. Uh we found out
[62:36] later.
[62:37] And that was the reason Biden
[62:39] of course ran. He said the
[62:40] Charlottesville both sides thing was his
[62:42] inspiration. 58 million in 2026, to your
[62:45] point, Chamath. Doubled and spiked to
[62:47] 136 million, more than double. And it's
[62:50] remained elevated ever since. Here are
[62:52] some, you know, images. For the
[62:54] indictment, and I'll wrap on this and
[62:56] then get everybody's uh feedback. They
[62:59] had F-37 as one of their confidential
[63:02] informants. He was a member, and this is
[63:04] according to the indictment, quote,
[63:06] member of the online leadership chat
[63:09] group that planned the 2017 Unite the
[63:11] Right event in Charlottesville,
[63:13] Virginia, and attended the event at the
[63:16] direction of the SPLC.
[63:19] F37 made racist postings under the
[63:21] supervision of the SPLC and helped
[63:23] coordinate transportation to the event
[63:25] for several attendees between 2015 and
[63:27] 2023. The SPLC secretly paid F37 more
[63:31] than $270,000.
[63:33] That's the legal case here. Let me pause
[63:35] there.
[63:36] >> Can I add one thing?
[63:37] >> Sure, keep adding. There's a lot of
[63:38] detail to this case. Yeah.
[63:39] >> Yeah, so you're right that the SPLC
[63:41] allegedly did fund $270,000 to help plan
[63:45] Charlottesville. In addition to that,
[63:47] they secretly funneled more than $3
[63:49] million
[63:50] to
[63:51] to a bunch of violent racist extremist
[63:53] groups including the Ku Klux Klan, the
[63:57] American Nazi Party,
[63:59] Aryan Nation,
[64:01] United Klans of America,
[64:03] and it goes on from there. So, I think
[64:05] don't forget about the $3 million bucks.
[64:06] So, this group that was supposed to be
[64:08] fighting racism in fact was fomenting
[64:11] racism by paying these groups to
[64:15] basically organize protests
[64:17] that SPLC could then point to and say
[64:20] that America has a huge racism problem,
[64:22] donate to us. And that's basically what
[64:23] happened after Charlottesville.
[64:25] They increased the amount of money that
[64:27] they were able to fundraise by $81
[64:30] million. So, that $270,000 investment
[64:34] led to an $81 million return, pretty
[64:36] good. But this is kind of the whole
[64:37] point of the story is that these guys
[64:38] are basically running a grift. And one
[64:41] of the ways that you know this is a a
[64:43] grift is because according to the
[64:45] indictment
[64:46] that they opened bank accounts under
[64:49] fictitious entities to conceal the
[64:51] payments that they were making from
[64:53] their own donors.
[64:55] Because if their donors knew
[64:57] that they were funding the KKK,
[65:00] they wouldn't be getting all these
[65:02] contributions from Hollywood celebrities
[65:04] and all the rest of it. So, it's really
[65:06] just this unbelievable story.
[65:09] >> Under
[65:09] >> Uh it It the mind.
[65:10] >> And just to clean up a little bit there,
[65:12] these are allegations. They haven't made
[65:14] the jump from planning these events and
[65:17] the SPLC
[65:19] claims they were not planning these
[65:22] things, they were monitoring. So, that's
[65:23] going to be their argument on the side.
[65:25] I'm not saying I agree with that.
[65:26] >> I know. I know. You're You're You're
[65:27] right that the SPLC's cover story is
[65:30] that they were simply paying informants.
[65:32] That's what they've claimed.
[65:34] But, there's two problems with that
[65:35] story. Number one is they were paying
[65:37] the actual leaders of these groups, not
[65:40] just sort of moles who were infiltrating
[65:42] the groups. And second, these leaders,
[65:46] they weren't paid to inform, they were
[65:47] paid to foment the activities. So, I'm
[65:50] just saying that's the the flaw. I
[65:52] understand they have this cover story
[65:54] that they were just paying informants.
[65:55] I'm just saying in my view, that does
[65:57] not hold up. And again, if they were
[65:59] just paying informants, why the
[66:01] extraordinary efforts to conceal the
[66:02] payments from their own donors? If they
[66:04] were proud of these efforts to
[66:06] infiltrate these groups, they should
[66:08] have basically informed their donors
[66:09] what they were doing. In fact, they hid
[66:11] it.
[66:12] >> And well, here's the reason uh that it
[66:14] was hidden, according to them. Again,
[66:16] I'm not taking the side. SPLC is not an
[66:18] organization I'm endorsing in any way.
[66:21] Their version of this is we didn't plan
[66:23] any of this.
[66:24] If we put
[66:26] SPLC
[66:27] bank accounts together for informants,
[66:29] that would be like the FBI sending a
[66:31] check to an informant from an FBI
[66:33] account. That's their explanation of it.
[66:34] >> They're not a law enforcement agency.
[66:36] >> Well, that's actually the question I had
[66:38] about all this is like what is a
[66:39] nonprofit doing hiring confidential
[66:42] informants, Chamath, to infiltrate these
[66:44] organizations, to what end? And then if
[66:46] you show me an incentive, you're going
[66:47] to you're going to see an outcome. And
[66:49] the outcome here is, "Hey, we'll get
[66:51] more donations if there's more racism."
[66:53] Uh your thoughts, just generally
[66:55] speaking here, Chamath. Again, all of
[66:57] this is alleged.
[66:58] >> These NGOs have completely run amok.
[67:01] They're cause-playing as these
[67:04] overlords and power brokers in our
[67:06] lives, and it needs to get stopped. They
[67:08] should all be dismantled. The people
[67:10] that donated to the SPLC should sue
[67:12] them.
[67:13] Rip open all of the documentation, get
[67:16] their money back. Because just so you
[67:18] guys know, if you are listening or
[67:19] watching and you have donated, there's
[67:21] $822 million of your money sitting in an
[67:25] offshore bank account waiting for you to
[67:27] get it back. Okay? And then separately,
[67:29] if you are thinking of donating to any
[67:31] of these organizations in the future,
[67:33] unless there is a
[67:35] full transparent auditing of it, you
[67:37] actually may be doing the opposite of
[67:39] what you thought. If you are against
[67:41] racism, you may be supporting racism. If
[67:44] you are against discrimination for gays,
[67:47] this could be actually promoting
[67:49] discrimination for gays. If you are
[67:50] supportive of trans rights, this may be
[67:52] pushing back against trans rights. Cuz
[67:54] the playbook seems to be do the opposite
[67:57] to create the narrative, give it to your
[67:59] friends in the media who will look the
[68:01] other way and just amplify it, tell the
[68:03] lie, create the craziness, and then
[68:06] raise a bunch of money, make a bunch of
[68:08] stink, and try to
[68:09] curate power.
[68:10] >> Friedberg, you think this is
[68:12] in your estimation or your gut tell you
[68:14] this is arsonists
[68:16] firefighters? They're lighting things on
[68:18] fire so that they can go put it out.
[68:20] Or do you think this is like
[68:23] a law fair as some people are claiming
[68:25] because there hasn't been to Chamath's
[68:27] point, it's they don't have donors
[68:30] taking this action. They're being
[68:31] accused of wire fraud on behalf of
[68:33] donors who haven't shown up yet to do,
[68:35] you know, a legal action. What's your
[68:36] take on all of this, Friedberg?
[68:39] >> The IRS definition of what a 501c3
[68:43] nonprofit organization is meant to be
[68:45] doing is to engage in exempt activities.
[68:49] The definition of exempt activities is
[68:51] charitable,
[68:53] religious, educational, scientific,
[68:56] literacy,
[68:58] public safety,
[69:00] or fostering amateur sports competition
[69:03] or preventing cruelty to children or
[69:05] animals.
[69:06] You tell me how the [ __ ] 90% of what we
[69:10] call nonprofits today fall under that
[69:12] definition. We have completely closed
[69:15] our eyes to the fact that organizations,
[69:18] regardless of political affiliation,
[69:20] social interest, have fundamental
[69:22] commercial and probably not aligned
[69:25] interests with the definition of a
[69:27] 501c3, and we've allowed them all to get
[69:30] away with it for far too long. I don't
[69:32] think that this is a blue or red thing.
[69:34] I think that this is a thing where we
[69:36] let these organizations make it easy to
[69:38] get money, to hide the money, and to do
[69:40] whatever the hell they want with the
[69:41] money, and we need to stop it. And I
[69:43] think that it's an amazing opportunity
[69:45] right now for everyone to kind of reset
[69:47] the decks by cleaning all the [ __ ] up
[69:50] and getting all of these organizations
[69:51] flushed and make sure that any
[69:53] organization that wants to do whatever
[69:55] [ __ ] nefarious things they want to
[69:56] do,
[69:57] by all means do it. But it's not a
[69:59] nonprofit, and you shouldn't get a
[70:01] charitable donation deduction, and the
[70:03] government should not be putting money
[70:05] into these sorts of things. This is an
[70:06] entirely different sort of activity in
[70:10] the social order. And as a libertarian,
[70:12] I'm all for it, but I don't think that
[70:14] they should be tax-exempt, and I don't
[70:15] think they should be getting government
[70:16] money, and I don't think that
[70:17] individuals should be benefiting from
[70:19] giving them money. And if we could fix
[70:21] all that [ __ ] up, I think a lot of these
[70:22] problems are going to go away, and I
[70:24] think this is a major problem.
[70:25] >> I think the theme of this episode is
[70:27] audit everything, whether it's
[70:29] government waste and abuse, or it's
[70:31] these NGOs, or it's people like Dow
[70:34] making these chemicals that 30 years
[70:35] later, you know, perhaps are correlated
[70:39] with cancer. We need to audit
[70:41] everything. We need to take a fresh look
[70:43] at this cuz it's not red versus green.
[70:45] Red This is not red versus blue. It's
[70:46] green. This is clearly a monetary
[70:49] incentive, and it is incredibly
[70:51] disruptive for society to not know the
[70:53] truth about
[70:54] what's going on with race in this
[70:56] country. I got
[70:58] absolutely, you guys might not know
[70:59] this, but in this is part of the cancel
[71:03] culture moment in time where they tried
[71:05] to take people having reasonable
[71:06] discussions about race in this country
[71:08] and tried to cancel them. They tried to
[71:09] do this to me in 2014.
[71:12] Very famously, you guys may not know
[71:13] this, but I have won a couple of awards
[71:15] in my career. Most offensive tweet ever
[71:18] by Vice in 2014 was my
[71:21] alleged racist tweet where I said, "Hey,
[71:23] if you want to get into uh blogging and
[71:26] journalism, there's no racism in Czech
[71:28] journalism. All you have to do is
[71:29] publish for a couple of years a blog
[71:30] post, nobody can stop you, and there'll
[71:32] be a ton of jobs available to you." And
[71:34] then what they did was they tried to
[71:35] cancel me and tried to cancel all my
[71:37] media properties and my investing.
[71:39] And this stuff had like a modest impact
[71:41] on me maybe for a year.
[71:43] And then now it's obviously all being
[71:45] >> Whoa, whoa, whoa, wait. I'm not sure I
[71:47] understand. [clears throat] J Kyle,
[71:47] you're saying the SPLC put you on a
[71:49] cancellation list?
[71:50] >> No, they put Sam Harris on it. Vice put
[71:52] me on a cancellation list. I It didn't
[71:54] get picked up by the SPLC, but I
[71:56] experienced the same thing, which was
[71:58] they said because I said
[71:59] you know, race does
[72:01] race doesn't play a role in hiring.
[72:04] >> You're so careful about your virtue
[72:05] signaling. I'm just shocked that anyone
[72:06] would try to cancel you.
[72:08] >> Well, that's what's shocking about it.
[72:09] It's like [laughter]
[72:10] I know I was very clear. I said, "In
[72:12] journalism, like a very vertical thing
[72:14] that
[72:15] >> skilled at experiencing You're very
[72:16] skilled virtue signaler, so
[72:18] >> [laughter]
[72:18] >> I mean
[72:19] >> They tried to cancel me, guys. I don't,
[72:20] you know, it's They tried to cancel you,
[72:22] too, Jamal.
[72:23] >> Jason, go ahead.
[72:24] >> You have a question for me?
[72:25] >> I'm uncancelable.
[72:26] >> Yes.
[72:26] >> Cuz I don't give a That's what we found
[72:28] out through all this shenanigans. You
[72:30] care about all what all these idiots
[72:31] think. But
[72:32] >> don't. I never have. [laughter]
[72:33] >> Jason, I have a question for you.
[72:35] >> Go ahead.
[72:35] >> What percentage odds now do you keep in
[72:37] the back of your mind that your
[72:39] petite little illustrious Human Rights
[72:41] Watch
[72:42] >> [laughter]
[72:43] >> is actually creating human rights abuses
[72:45] to try to
[72:46] >> This is actually a very good point. You
[72:47] know, a lot of the human rights
[72:48] organizations from back in the day that
[72:50] >> What is the organization that you were
[72:52] What is it called?
[72:52] >> Amnesty International.
[72:53] >> Oh, yeah.
[72:54] >> When I worked at Amnesty International,
[72:55] a very fine
[72:57] mandate. The mandate was human rights
[72:59] abuses as described in the Universal
[73:01] Declaration of Human Rights created by
[73:03] Eleanor Roosevelt and the uh
[73:05] >> This is like science corner. Wait until
[73:07] this is done, Sam.
[73:08] >> No, but they It was torture. [laughter]
[73:09] It was people being put in jail and
[73:11] being tortured. It was people being
[73:13] censored because of freedom of speech.
[73:16] And And that's what I worked on when I
[73:18] was at Amnesty International. These
[73:19] groups went adrift in order to get
[73:22] money, Human Rights Watch included, and
[73:24] then they started taking on things like,
[73:27] you know, uh transgender rights, this
[73:29] rights, that rights, and censoring
[73:31] people. They They went after Sam Harris
[73:33] because he had Charles Murray from the
[73:36] Bell Curve.
[73:37] >> with all this [ __ ] I'm asking you a
[73:38] very specific question.
[73:39] >> chance that all these organizations are
[73:41] involved in human rights.
[73:42] >> 50/50 chance
[73:43] >> Depending on the organization. SPLC, I'm
[73:45] going to guess 95%. Yeah.
[73:47] >> Amnesty International, you think is
[73:49] 50/50 that they're engaging in nefarious
[73:51] [ __ ] to try to whip up people's
[73:53] belief that there are human rights
[73:55] abuses happening that are not happening.
[73:56] You're saying it's a coin flip.
[73:58] >> I think it's probably a coin flip, yeah.
[74:00] That's what I would say today because
[74:01] these organizations all got co-opted.
[74:03] SPLC might have had a great origin
[74:05] story, but now it is
[74:07] >> your intellectual honesty and I
[74:09] appreciate you saying that.
[74:10] >> Well, I mean, just based on facts. So,
[74:12] let's see what this
[74:13] legal case brings about.
[74:15] >> Well, let's be clear. This is not Let's
[74:17] be clear.
[74:17] >> them investigating SPLC, 100%.
[74:19] >> When you bring a grand jury indictment,
[74:21] you've already previewed the evidence.
[74:24] This is not like some guy's trying to
[74:25] whip up lawfare.
[74:27] Okay?
[74:27] >> Um actually, you don't have to bring all
[74:29] the evidence, but that's a side thing.
[74:31] And And they're very frisky about
[74:33] allowing you to indict somebody as we
[74:35] experienced with Trump.
[74:37] >> Grand juries are a whole different
[74:38] animal.
[74:38] >> Yeah. They They will indict a ham
[74:40] sandwich is the line. Um so, we'll see.
[74:42] Let's give them their day in court is
[74:43] always my position.
[74:44] >> Well, I mean, regardless of what happens
[74:46] in court, if it's true that the SPLC is
[74:48] funding the Ku Klux Klan and the
[74:51] American [laughter] Nazi Party,
[74:52] >> ago, just so we're clear. They stopped
[74:54] using confidential informants.
[74:55] >> That's good enough for me.
[74:57] >> loud, it's insane.
[74:58] >> It's good enough for me.
[74:59] >> You know, it's insane.
[75:01] >> Here Here's the systemic problem with
[75:03] nonprofits and NGOs. Is Let me just
[75:06] contrast it with business. In business,
[75:08] you set up a company, the company has to
[75:10] make revenue, it has to make profits,
[75:12] and if it doesn't, it's going to go out
[75:13] of business, right? Because it'll lose
[75:14] money. So, there's a feedback mechanism
[75:17] from the market. The company has to
[75:19] create products that people are willing
[75:20] to buy, and those products have to make
[75:22] money. With an NGO, nonprofit, what have
[75:24] you, they raise money, they don't sell
[75:27] things, they fundraise from donors in
[75:30] order to engage in an activity. But what
[75:32] happens over time is the actual
[75:34] activities may stop mattering, and all
[75:37] that really matters is they're able to
[75:38] keep fundraising, right? Because they're
[75:39] just trying to figure out a
[75:41] justification to keep going back to
[75:42] donors to get more and more money out of
[75:44] them. That's how it perpetuates the
[75:45] organization.
[75:46] >> And they're trying to keep their job.
[75:47] >> Exactly. And then, if it's an NGO that
[75:50] gets money from the government, then
[75:52] it's even worse, because all they do
[75:54] from that point forward is try to lobby
[75:56] the government to get more money. And it
[75:58] doesn't really matter whether the
[76:00] program is working or not. All that
[76:02] matters is whether they can spin it as
[76:04] working.
[76:05] >> Why wouldn't the Southern Poverty Law
[76:07] Center focus on Southern Poverty, which
[76:10] is an issue that actually still exists
[76:12] in some parts of the
[76:13] >> Shouldn't that be a better thing? I
[76:13] mean,
[76:14] and why do you call it one thing, focus
[76:16] on racism, and then all of a sudden whip
[76:20] up
[76:20] >> I'll tell you why. Here's my theory.
[76:21] Here's my theory on it. Is I do think
[76:23] that at one time in this country, civil
[76:25] rights was a noble cause, a very
[76:28] legitimate cause. We had the the legacy
[76:31] of segregation and Jim Crow, and there
[76:33] were groups that were set up to
[76:35] basically change that, and they
[76:37] succeeded.
[76:39] But again, no one in an NGO or a
[76:41] non-profit ever declares victory.
[76:44] They're never going to say, "You know
[76:45] what? Like we we addressed this problem.
[76:47] We solved it." You know, I always
[76:48] thought that in 2008 when when
[76:50] >> me. Fire me. My job's done.
[76:52] >> Yeah, well well when Obama got elected
[76:54] in 2008, I thought that regardless of
[76:56] whether you liked Obama or not or agree
[76:57] with his politics, I thought that at
[76:59] that point most people could see that
[77:01] this was not a racist country.
[77:03] >> 100%.
[77:04] >> Whatever else you could say, the fact
[77:06] that the highest office in the land was
[77:08] not denied to any body showed that this
[77:11] country was not holding people back
[77:13] based on their skin color.
[77:15] And instead of just basically packing up
[77:17] shop and saying, "Okay, we've achieved
[77:18] our goal."
[77:19] >> [laughter]
[77:19] >> The goal posts all got moved. Remember,
[77:21] that's when the whole anti-racism thing
[77:23] started was was around Obama's second
[77:25] term. And what anti-racism was, it said
[77:28] that
[77:29] it's not good enough not to be just not
[77:31] to be racist. You actually have to be
[77:33] anti-racist, but what anti-racism meant
[77:36] was was basically that all the
[77:38] distributions had to match the
[77:40] population. Basically, it meant equality
[77:42] of outcomes, not equality of
[77:43] opportunity. So, effectively
[77:45] this whole goal post was moved from
[77:48] equality of opportunity to equality of
[77:49] results.
[77:50] >> you see it, you can't unsee it. It's
[77:52] like they sat around and they said, "Now
[77:53] what?" And one person was like, "I got
[77:56] an idea."
[77:57] >> [laughter]
[77:58] >> Well, and make race a symptom again.
[78:00] >> Exactly.
[78:01] >> if they just said if they just said at
[78:03] that time, "You know what? We're going
[78:04] to move the goal post from equality of
[78:06] opportunity to equality of results.
[78:07] We're going to basically make everyone
[78:09] equal at the finish line, which is to
[78:11] say I guess communism or or some sort of
[78:14] identity socialism." People would have
[78:15] said, "Uh no, we're not on board for
[78:17] that." So, instead they created this
[78:18] whole new terminology to justify it. It
[78:21] has taken us years to unpack that and
[78:24] realize what's really going on.
[78:26] >> Gosh, I don't want to put myself in a
[78:27] position of defending the SPLC. They
[78:29] were partners with the FBI for a long
[78:30] time. To your point, Chamath, or or
[78:32] Sax's point, rather. There was probably
[78:34] a time when it was important to
[78:36] infiltrate the KKK and the Nazi groups.
[78:40] >> It's not 2025.
[78:42] >> not in 2026 like I think necessary to be
[78:45] doing this work. I think more than four
[78:47] people can handle it. Uh in 20 26.
[78:50] >> to give you guys a newsflash. I just got
[78:52] this just hit the wire. Um this is
[78:54] really important.
[78:54] >> Breaking news here.
[78:56] >> America is profoundly less racist than
[78:58] you think, okay?
[79:00] >> There we go. Okay, breaking news.
[79:02] >> Wake up.
[79:03] >> Friedberg wanted to do a surprise
[79:06] science corner. This is the first. We
[79:07] don't know what he's about to talk
[79:09] about, but David looks like he's been
[79:11] working really hard and he needs a nap,
[79:12] so Friedberg, you have the microphone.
[79:14] Let's go.
[79:15] >> This was not
[79:16] >> Surprise science corner.
[79:17] >> Yeah, this is not necessarily a big
[79:19] surprise, but there was a really
[79:21] interesting paper published this week
[79:24] on trying to elucidate the underlying
[79:27] cause or predictor of colorectal cancer.
[79:30] So, I don't know if you guys know any
[79:32] young friends, but colorectal cancer or
[79:35] Nick, if you could just pull up this
[79:36] first image, or colon cancer
[79:38] has become now the third leading cancer.
[79:42] Over the last 20 years or so, there's
[79:44] been a scary rise in the number of young
[79:46] people, people generally under 50 years
[79:49] old that are getting colon cancer. That
[79:51] number has climbed by
[79:53] over 80% in just the last
[79:56] two decades.
[79:57] Historically, it's been an age-related
[79:59] disease. So, as you get older, over 70
[80:01] years old, your probability of getting
[80:03] colon cancer shoots through the roof,
[80:04] but this rise in young people getting
[80:06] colon cancer has been pretty alarming,
[80:08] and there's been a real question mark on
[80:10] what is causing it. What's the
[80:11] underlying
[80:13] trigger.
[80:14] So, this research team out of Barcelona
[80:17] in Spain did an amazing study
[80:19] where they looked at the difference in
[80:22] the epigenome or the gene expression
[80:26] in tumor cells
[80:28] of patients that are under 50 years old
[80:31] and those that are over 70 years old.
[80:33] This sort of data will show you what
[80:36] different environmental triggers are
[80:39] associated with those changes in gene
[80:41] expression. So, whenever we're exposed
[80:43] to something in the environment, whether
[80:45] it's some food or some drink or whatever
[80:48] else it is, some chemical in the
[80:49] environment, the cells in our body that
[80:52] are exposed to that chemistry or exposed
[80:55] to that environmental trigger
[80:57] have genes that get switched on and off.
[81:00] And you can see which genes are on and
[81:02] off by looking at the RNA of those
[81:04] genes, which tells you that those genes
[81:06] are expressing RNA to make protein or
[81:08] not make protein. And you can look at
[81:10] that gene expression to determine what
[81:13] is changing when a cell is exposed to a
[81:17] particular environmental trigger. And
[81:19] so, they were able to get these samples
[81:21] of colon adenocarcinomas
[81:23] from the Cancer Genome Atlas, which is
[81:26] funded by the federal government. And
[81:28] they were then able to take a look at
[81:30] these cancer cells from colon cancer in
[81:33] patients that are under 50 and patients
[81:35] that are over 70, and look at the
[81:37] difference in the gene expression
[81:39] profile
[81:40] and what um environmental triggers are
[81:42] associated with that gene expression
[81:45] profile. So, that will tell you, "Hey,
[81:47] these environmental triggers are more
[81:49] likely the cause or an underlying driver
[81:52] of the risk of getting this colon
[81:54] cancer."
[81:55] And one thing rose to the top. So, they
[81:57] looked at a whole bunch of things. They
[81:58] look at lifestyle factors, they look at
[82:01] eating index, how much you ate, how
[82:03] overweight you were, alcohol, birth
[82:05] weight. They adjusted for gender, they
[82:07] adjusted for all these different things.
[82:09] And as you look down this list, you'll
[82:11] see this is the difference between
[82:13] people that got colon cancer that were
[82:14] over 70, when you typically have a very
[82:16] high chance of getting it, and people
[82:18] that are under 50, when you don't. And
[82:21] what is going on with people under 50?
[82:23] And you can see there's this one row
[82:24] here that's all orange.
[82:27] That row is a pesticide called picloram.
[82:31] Picloram is a pesticide that was
[82:33] developed by the Dow Chemical Company
[82:36] in 1963.
[82:38] This is the
[82:40] chemical formula for that pesticide.
[82:42] It's related to auxin, which are these
[82:44] hormones that plants make. And in the
[82:46] 1960s, there was this big rush to try
[82:48] and make synthetic plant hormones that
[82:50] you would then apply to a plant. It
[82:52] would cause the plant to overgrow, and
[82:54] the plant would quickly die. And
[82:56] picloram became a very widely used
[82:58] herbicide
[83:00] in our environment. It's used to manage
[83:03] weeds in rangeland and pastureland where
[83:05] cattle graze. It's used to control weeds
[83:08] near roads and near railroads, on
[83:11] industrial sites to clear weeds away
[83:13] from highways and utility corridors. And
[83:16] the problem with picloram, one of the
[83:17] the things that's been known about it is
[83:18] it's very persistent. It doesn't
[83:20] biodegrade very well. Picloram sticks
[83:22] around for well over a year. It stays in
[83:25] the water. It moves into groundwater,
[83:27] and it's persistently in the environment
[83:29] after it's been used for some period of
[83:31] time. I went back and looked at the EPA
[83:34] data on this chemical. The last time
[83:36] there was an EPA safety study done was
[83:38] in 1995.
[83:40] And so this was before we had this
[83:41] capacity to do epigenomic studies like
[83:44] what was just done to elucidate that
[83:46] even though a chemical might not be
[83:48] causing cancer immediately, and you
[83:50] can't apply it to a cell and see it
[83:51] trigger a cancer,
[83:53] the long-term use or exposure to certain
[83:56] chemicals in our environment causes a
[83:58] change in the epigenome, which means
[84:00] that these genes are being turned on and
[84:01] off. And when certain genes are turned
[84:03] on or off in the wrong way, it can
[84:05] trigger cells in the tissue to start to
[84:08] malfunction and go haywire, and
[84:10] ultimately lead to cancer. And I think
[84:12] that this paper shows a pretty strong
[84:14] effect of picloram in driving colon
[84:17] cancer in young people. It will very
[84:19] likely lead, and it should lead to an
[84:21] EPA review on whether this should be
[84:22] legally allowed, but it should also lead
[84:25] to a new mechanism by which we assess
[84:27] chemistry that we're using in our food
[84:28] supply, in our environment, in our
[84:30] industrial applications, because we can
[84:32] now look at all of this at sort of
[84:34] epigenomic data to try and figure out
[84:36] what are these chemicals doing to us
[84:38] before we see them cause the problem.
[84:40] So, I thought this was like an amazing
[84:41] paper done by this team. They did a lot
[84:43] of work to try and make sure that the
[84:45] statistics were sound in the studies
[84:47] that they did.
[84:48] It really
[84:49] uh I think elucidated something pretty
[84:51] scary.
[84:51] >> Is this like a Monsanto thing where like
[84:53] one company makes it or
[84:55] picloram is broadly available?
[84:57] >> It's off patent now and so I'm pretty
[85:00] sure my guess, I haven't looked into
[85:01] this, but my guess is most of this is
[85:03] made generically in China and then it's
[85:05] probably packaged up with lots of
[85:06] different brands in the US and all over
[85:08] the world. So, it's one of these
[85:09] chemicals that's just become ubiquitous
[85:11] in our use
[85:12] that just shows up everywhere, but I
[85:14] think it really speaks to the fact that
[85:17] historically, think about 1995, you can
[85:20] look at what the immediate chemical
[85:21] application of something does to a a rat
[85:23] or a human cell and you can say like,
[85:26] "Oh, it didn't cause cancer. It's good
[85:27] to go. Let's go." You know, didn't
[85:28] didn't cause quote toxicity.
[85:31] >> Can I ask a question? In that study, are
[85:33] you exposed to picloram based on where
[85:34] you live
[85:35] >> Oh, yeah. Sorry. That's a that's a great
[85:37] question, Chamath. So, I was going to
[85:38] talk about this. Thank you for asking
[85:40] that. They then took that uh picloram
[85:42] exposure and then they looked at all the
[85:44] counties across the United States. They
[85:45] were able to gather data where there was
[85:47] enough data in California, Connecticut,
[85:49] Georgia, Iowa, New Mexico, Utah,
[85:51] Washington, and they were able to look
[85:53] at picloram use estimates from the
[85:55] Pesticide National Synthesis Project and
[85:58] try and deduce in places where picloram
[85:59] was highly used and not highly used. And
[86:02] once again, it elucidated signal, which
[86:04] is that when picloram was used in the
[86:05] environment in the counties more
[86:07] frequently, there was a much higher
[86:08] frequency of colon cancer in those
[86:10] counties.
[86:11] >> And that R squared is weak or it's
[86:12] strong?
[86:13] >> Reasonably strong. The odds ratio is
[86:15] like 3x. It's very strong.
[86:16] >> This is accomplished Freeberg from a
[86:19] combination of big data
[86:22] and this uh science
[86:24] >> I think it's important Yeah, I think
[86:26] it's important
[86:26] >> testing as well, right? So, you have
[86:28] this confluence of increased testing,
[86:31] increased data, you know, knowing where
[86:33] these instances are occurring, and if
[86:36] you add a layer of AI onto this
[86:37] Freeberg, this is like a really positive
[86:40] use.
[86:41] Going back and looking at all these
[86:42] compounds and figuring out which ones we
[86:44] need to eliminate.
[86:45] >> Yeah, so I'll I'll put my PCast hat on.
[86:47] Thank you, David Sachs, for the role.
[86:49] And I think this speaks to one of the
[86:51] important roles that government has in
[86:54] doing fundamental science and
[86:55] fundamental research. So, the the
[86:56] National Cancer Institute in the federal
[86:58] government stood up this genome atlas
[87:01] with a hundred million dollars a couple
[87:02] years ago. They spend only a few million
[87:04] dollars a year now to maintain it to get
[87:06] cancer tissue samples and then create
[87:09] the availability to scientists to use
[87:11] those cancer tissue samples to do the
[87:13] sort of epigenomic analysis this and
[87:15] study supported by, you know, government
[87:17] grants, or in this case supported by
[87:19] a foreign university getting funding to
[87:22] do it. And so, there's there's an
[87:24] important role that fundamental science
[87:25] still has in elucidating this that we
[87:27] would have otherwise not been able to
[87:28] see if we didn't have this resource
[87:30] available to us from the federal
[87:32] government and federal funding of
[87:33] scientific programs like this. And that
[87:35] leads to this discovery. You don't need
[87:37] fancy AI for this, to be frank, Jay Cal.
[87:39] There's an incredible amount of data
[87:41] that's available or or resources that
[87:43] are available. What's happened in the
[87:45] last couple years is what's called RNA
[87:46] sequencing where you can actually look
[87:48] at which genes are on or off, not just
[87:50] what's the DNA, but in the DNA, remember
[87:52] we've talked a lot about the epigenome,
[87:54] what genes are on or off, and how that
[87:56] changes when you have different cancers
[87:59] or when you have different chemicals,
[88:00] and when you have a certain chemical
[88:02] like picloram, your colorectal cancer
[88:04] goes through the roof, and you can see
[88:06] that relationship in those tissues. And
[88:08] then you can put all the data together
[88:09] and say, "Oh my gosh, there's a lot of
[88:11] evidence here that points to this
[88:12] connection. Very powerful. I think it's
[88:14] important that it opens up the window
[88:16] that this shouldn't just be a one-off
[88:17] research project conducted by a team in
[88:20] Spain, but maybe should be a fundamental
[88:22] role that some of the government
[88:23] agencies play, which is to stop
[88:25] Americans and the world from getting
[88:27] freaking cancer. Let's figure out the
[88:28] things that we got wrong in industry and
[88:31] go back and delete them out of our food
[88:33] supply and out of our industrial supply.
[88:35] And I think this is a really good
[88:36] example of that.
[88:37] >> So X, uh, how does Freeberg's focus on
[88:40] Uranus, uh, in you know, inform your
[88:43] co-leading of P-cast here? Are you going
[88:45] to go deep into this colon research? How
[88:48] deep do you plan on going? And how will
[88:49] you get through eight of these
[88:51] presentations a day at P-cast?
[88:54] >> [laughter]
[88:54] >> All right, it's all good. This is why we
[88:56] hired Freeberg.
[88:57] >> Yes.
[88:58] >> By the way, did you guys
[88:58] >> He's going to handle, uh, Mars, Neptune,
[89:01] and Uranus.
[89:02] >> Absolutely. He's [laughter] going to go
[89:03] deep into Uranus and clean it up. We
[89:06] need to clean up Uranus.
[89:09] >> Uh, great work.
[89:09] >> Anyway, Freeberg, great great work.
[89:11] >> Anyway, I think I think this is
[89:12] important and I don't think there's any
[89:13] news attention on this since it came out
[89:14] a couple days ago. So I thought it would
[89:16] be worth bringing up on the show and
[89:18] making people aware.
[89:19] >> All right.
[89:19] >> But thank you guys for sitting through
[89:20] it.
[89:20] >> Well, no, I think it's it's great work
[89:22] you're doing there.
[89:22] >> I just read the paper.
[89:25] >> Yeah.
[89:25] >> All right, everybody. That's it for
[89:27] episode 270 of the world's greatest
[89:29] podcast. I am your world's greatest
[89:31] moderator. Thank you Chamath, David
[89:34] Sacks, and David Freeberg for the
[89:36] episode. To your friends, your
[89:37] neighbors, and we'll see you all next
[89:39] time. Bye-bye.
[89:40] >> Love you boys. Bye-bye.
[89:43] >> Let your winners ride. [music]
[89:45] >> Rain Man David Sacks.
[89:50] >> And I said
[89:50] >> We open sourced [music] it to the fans
[89:52] and they've just gone crazy with it.
[89:54] >> Love you guys.
[89:54] >> I'm Queen of King Wa.
[89:59] >> Let your winners
[89:59] >> [music]
[90:00] >> ride.
[90:03] >> Besties are back.
[90:05] >> That is my dog taking a [ __ ] in your
[90:07] driveway, [music] Sacks.
[90:10] >> Oh, man.
[90:12] >> [music]
[90:18] [music]
[90:26] [laughter]
[90:31] [music]
[90:36] >> I'm going all in.
