"if Fable 5 detects you're doing frontier AI research in other words using their model to make a better model... they were downgrading you but they weren't telling you" - Jason Calacanis [00:01:43]
"what are the best open source models today they're Chinese yeah they are and that is a major concern the American open source models are not as good" - David Friedberg [00:07:17]
Disclaimer: Orignal content owned by or sourced from third parties. It does not represent the views of 'Nuggets' platform or it's team. AI is used extensively across this platform including for summaries. Accuracy is not guaranteed, there can be mistakes. Any info or content on this platform is not a financial, legal, or investment advice. Do your own research. Refer for complete disclosures:- Terms of Use · Full Disclaimer
"eight months ago I said that Enthropic was engaged in a very sophisticated regulatory capture campaign based on fear-mongering and people at the time thought that was a very spicy take but 8 months later I think you're hearing a lot of people say it" - David Sacks [00:09:59]
"a gigawatt now costs a hundred billion dollars guys... so if you want to get all three gigawatts developed I have to come up with 300 billion i don't have" - Chamath Palihapitiya [00:18:28]
"the incremental cost of performing AI is excessive and large that contrasts and compares to the incremental cost before AI which was zero" - Chamath Palihapitiya [00:55:09]
"There is no election... you are a citizen of those who tell you who your overseers are you are no longer allowed to vote for your elected representatives they are now appointed representatives" - David Friedberg [01:13:06]
Speakers & Credentials
Jason Calacanis: Host, Angel Investor, and Podcaster. Provides the structural narrative and moderation for the conversation.
Chamath Palihapitiya: Venture Capitalist and Capital Allocator. Focuses on the macro-economics of energy infrastructure, data center capital requirements, and global market dynamics.
David Sacks: Venture Capitalist and Software Executive. Brings a sharp political and geopolitical lens, specifically critiquing regulatory capture, corporate censorship, and the erosion of election integrity.
David Friedberg: Entrepreneur and Scientist (AgTech/Genomics). Evaluates AI through the lens of applied scientific utility, economic productivity, and the structural design of macro systems like social security and electoral mechanics.
1. Executive Summary
The AI Censorship & Regulatory Capture Era: Anthropic’s release of "Fable 5" has triggered massive developer backlash due to 30-day mandatory data retention and shadow-downgrading of frontier users, revealing a pivot toward strict gatekeeping under the guise of "safety."
The Open-Source Geopolitical Threat: As domestic frontier models lock down access for enterprise and scientific use, U.S. startups are being forced to adopt superior Chinese open-source models, creating severe long-term economic and security vulnerabilities.
The $100 Billion Capital Moat: The infrastructure cost to support open-source AI has exploded; building a 1-gigawatt data center now requires $100 billion, cementing power into a potential AI duopoly that cannot be easily challenged by independent actors.
Nationalizing AI vs. Rebuilding Social Security: In response to tech CEOs' narratives regarding imminent AI-induced job loss, politicians like Bernie Sanders are pushing for a 50% equity seizure of AI companies, sparking counter-proposals to convert Social Security into a sovereign wealth fund that holds equities.
Macro Inflationary Shocks: Against expectations of cooling, CPI surged to 4.2% and PPI to 6.5%, driven largely by global conflict energy blips, putting massive pressure on the Fed and eliminating the likelihood of short-term rate cuts.
The Institutionalization of Election Engineering: A rigorous dissection of the LA Mayoral primary reveals how systematic, legally sanctioned loopholes—such as AB 1921's unlimited ballot harvesting—statistically warp late-arriving mail-in ballots, effectively transitioning democratic elections into machine-run appointments.
[01:12:34] - California's Broken Elections, Ballot Harvesting, and Statistical Anomalies
3. Detailed Thematic Summary
Anthropic, "Safetism", and The Regulatory Capture Playbook
Anthropic released their "Mythos" level model called Fable 5, which immediately raised red flags across the developer ecosystem. Tokens cost twice as much as Opus 4.8 [00:00:46], but more alarmingly, the company instituted a mandatory 30-day retention policy on all prompt data, eliminating privacy for enterprise clients [00:01:37].
The system actively shadow-downgrades users; if Fable 5 detects a user conducting frontier AI research or creating competitive models, it bumps them to a lesser model without notifying them or reducing their fees [00:01:43].
David Sacks argues this is not a mistake but a calculated feature of a massive regulatory capture campaign. Anthropic’s CEO Dario Amodei is actively petitioning for an FAA/FDA-style government agency to approve all models [00:16:19], which Sacks notes is a targeted maneuver to outlaw open-source competition which inherently cannot comply with central regulation.
Jason physically verified this hyper-censorship during the recording. By asking Fable 5 a generic journalistic question about the regulations surrounding fertilizer components, the model instantly flagged his account and downgraded him on the spot [00:30:40].
Enterprise Utility & The Push to Chinese Open-Source
David Friedberg highlighted the immediate collateral damage of AI censorship on hard science. His company historically used LLMs to design genetic constructs, map RNA guides, and analyze DNA sequences to optimize agricultural yields [00:05:44].
Due to vague "bioweapon" fears, Anthropic recently restricted these standard biological research queries, crippling scientific workflows [00:06:33].
As a direct consequence of domestic "safetism," American startups and enterprises are defecting to open-source models, and currently, the highest performing open-source models available are Chinese [00:07:17]. Friedberg warns this domestic gating is a severe unforced error that hands global technical leverage directly to international competitors.
The Hard Constraints of Capital and Compute Infrastructure
The group analyzed the physical reality of competing in the frontier model space. Chamath shared a sobering anecdote about his real estate investments: he purchased 2,000 acres in Arizona to build a 2-gigawatt data center [00:17:20].
Originally projected in the low billions, Chamath revealed that developing a single gigawatt of AI data center capacity now demands an astonishing $100 billion [00:18:28].
This astronomical capital requirement creates a virtually insurmountable moat. Even if the developer community desperately wants an American open-source champion to combat Anthropic and OpenAI, gathering the $300 billion required for a 3-gigawatt facility is near impossible without sovereign-level backing, cementing an imminent corporate duopoly [00:18:51].
Historical Context: Deep Time Parallels & Economic Regimes
The Manhattan Project Parallel: Debating the moral panic surrounding AI, Friedberg pointed to the splitting of the atom. While the breakthrough yielded the atomic bomb, it simultaneously provided the pathway for virtually unlimited, cheap energy [00:23:46]. Attempting to block the fundamental technology of AI prevents the digital "unlimited energy" (curing cancer, agricultural abundance) out of fear of the "bomb."
Internet vs. AI Economics: Chamath provided a deep structural comparison of the last 20 years of tech vs. the next 20. The internet boom (Facebook, Google) relied on a "money glitch" where the marginal cost of adding a new user was effectively zero [00:52:46].
In stark contrast, the incremental cost of performing AI is massive—taxing specific GPUs, requiring extreme power grids (electrons), and dedicated memory arrays. This shifts AI away from software economics and closer to heavy industrial infrastructure like the 1950s Interstate Highway System [00:54:00].
Nationalizing AI vs. Revamping Social Security
Driven by constant fear-mongering from AI CEOs (like Dario Amodei claiming a 50% job loss for knowledge workers in 1-5 years), politicians are taking action. Bernie Sanders proposed a one-time 50% tax on the stock of OpenAI, Anthropic, and XAI to fund an American AI Sovereign Wealth Fund [00:38:13].
While the hosts rejected the seizure as property confiscation, they acknowledged the AI industry "poked the tiger" by aggressively training on public data while threatening public livelihoods.
Friedberg countered with a capitalist solution: The US Social Security trust currently holds a single $4 trillion Treasury certificate [00:45:04]. By rewriting the mandate to allow the fund to purchase equities in transformative companies (like AI firms), the system could transition into a true, active Sovereign Wealth Fund, saving the bankrupt entitlement system through aggressive productivity scaling [00:44:50].
Friedberg sharply rejected the job loss narrative, noting that in his own engineering squads, AI acts as a revenue multiplier, allowing one engineer to do 100x the work, leading to massive hiring expansions, not layoffs [00:46:08].
Macro Shocks: The End of Rate Cut Dreams
The macroeconomic environment hit a wall with unexpectedly hot inflation prints. May CPI surged to 4.2% year-over-year, and PPI shocked the market at 6.5% year-over-year [01:05:42].
Chamath attributes this directly to energy market constraints exacerbated by global conflicts (specifically noting the Iran situation). If global buffer mechanisms (like China's spot market behavior) fail, oil could surge past $150–$200 a barrel [01:10:18].
Consequently, Polymarket odds flipped violently, pricing in a 49% chance that the Fed actually hikes rates this year, while the European Central Bank already hiked rates by 0.25% [01:06:21].
The Engineering of Elections: Legalizing Fraud in California
The podcast concluded with a heated forensic analysis of the LA Mayoral primary featuring Spencer Pratt, Karen Bass, and Nithya Raman.
On election day (in-person voting), Spencer Pratt held a commanding lead at 35%, compared to 29% for Bass and 26% for Raman [01:13:51].
However, in mail-in ballots received after election day, Pratt's share plummeted to 19%, while Raman's share surged 80% to capture 37% of the late pool [01:14:14].
Friedberg laid out the mechanics of how this is legally achieved: California Assembly Bill 1921 legalized unlimited ballot harvesting [01:15:44]. Combined with a lack of voter ID, un-scrubbed voter rolls resulting in millions of floating ballots, and zero chain-of-custody, political machines can legally aggregate and submit thousands of ballots late in the cycle.
Friedberg describes this not as illegal fraud, but as an engineered system of "appointments" rather than democratic elections, utilizing adverse selection to exploit legal loopholes [01:18:55].
The Reference Vault
4. Data & Figures
Data Point
Value
Context
Timestamp
Fable 5 Cost
2x Opus 4.8
Anthropic's new model is vastly more expensive despite promised efficiencies.
Regulatory Capture via "Safetism" [00:09:59]
Instead of competing on product merit, incumbents intentionally exaggerate the existential dangers of their own technology to invite government regulation. By demanding FDA/FAA-style oversight, companies like Anthropic create an unscalable compliance moat that effectively outlaws open-source competitors who lack the capital to lobby or strictly govern decentralized code. It is a monopolistic tactic wrapped in altruistic PR.
Deterministic Technology Evolution [00:25:36]
The principle that whatever is scientifically possible will be tried at least once. Applied to AI, attempting to stuff the genie back into the bottle through prompt-censorship is futile. The framework dictates that societies must govern the physical output and application of tech (e.g., product liability, bioweapon laws) rather than trying to outlaw the underlying mathematics or limit access to the computational tool itself.
The "Money Glitch" vs. Marginal Cost Divergence [00:52:46]
An economic model contrasting Web2 with AI. For decades, internet companies (Google, Meta) scaled massively because the marginal cost to serve an incremental user was practically zero. AI violently breaks this paradigm. The incremental cost of AI inference taxes physical GPUs, power grids, and cooling systems. This structural shift means AI companies will resemble heavy industrial monopolies (like rail or energy utilities) rather than agile software startups.
Adverse Selection in System Design [01:18:55]
Originating in insurance, this model dictates that if a system is designed with exploitable altruistic loopholes, bad actors will disproportionately leverage them until the system fundamentally breaks. In California's election mechanics, removing voter ID and legalizing ballot harvesting under the guise of "access" inevitably selected for hyper-organized political machines to exploit floating ballots, warping the intended democratic process into a game of operational arbitrage.
The Dual-Mandate of Public Benefit Corporations [00:42:37]
Unlike traditional C-Corps strictly bound to maximize shareholder returns, Public Benefit Corporations (like Anthropic and OpenAI) possess a legally binding dual-mandate to balance profits with a stated public good. This structural reality provides legitimate legal and political vectors for politicians (like Bernie Sanders) to demand societal dividends, weaponizing the companies' own altruistic tax structures against them.
6. Anecdotes
Ben Thompson’s Medical Inquiry Gets Downgraded: [00:12:53] Sacks shared how Stratechery's Ben Thompson asked Anthropic a basic scientific question correlating cancer risk with GLP-1 drugs. He was instantly punished and kicked down to a lesser model. This anecdote proves that "safetism" isn't stopping terrorists; it is actively degrading legitimate journalism and medical inquiry.
Chamath’s 2-Gigawatt Arizona Real Estate Shock: [00:17:20] Chamath spent years zoning 2,000 acres in Arizona for a 2GW data center, assuming a low billions cost. He discovered it now costs $100 billion per gigawatt. He told this story to brutally quantify the exact height of the AI capital moat, proving why "indie" open-source AI is economically impossible without nation-state-level funding.
Jason's Live Fertilizer Bomb Interrogation: [00:30:40] During the recording, Jason deliberately asked Fable 5 about the legal regulations surrounding fertilizer bombs. The system's backend "thinking" acknowledged he was a podcaster doing legitimate research, but downgraded him anyway. This live demonstration provided undeniable, real-time proof of the system's hyper-aggressive and unyielding censorship.
Thomas Laffont's "Trillicorn" Presentation: [01:00:02] At the Summit event, the Coatue executive presented data showing that once a company hits a $100B valuation, it has a massive 31% chance of reaching $1 Trillion. This mathematical anecdote was used to illustrate that late-stage hyper-scaling in the modern tech era is structurally more probable than early-stage startup survival.
Friedberg's Missing Election / "There is no spoon": [01:13:06] Friedberg opened the California voting segment by quoting The Matrix, telling the hosts that the election didn't happen—it was an appointment. He used this deeply cynical framing to transition the audience away from the standard "voter fraud" debate and into a chilling breakdown of how laws have been rewritten to make election engineering entirely legal.
7. References & Recommendations
Geopolitical Institutions & Government Bodies
US Treasury: Mentioned regarding the single $4 Trillion bond holding up the entire US Social Security system. [00:45:04]
FAA / FDA: Dario Amodei wants these types of strict, gatekeeping federal agencies applied to AI model releases to block open-source. [00:16:19]
European Central Bank (ECB): Cited as evidence of global macro tightening, having just hiked rates 0.25%. [01:06:40]
Companies, Funds & AI Models
Anthropic (Fable 5 / Mythos / Opus 4.8): The central antagonist of the episode, criticized for extreme censorship, spying on enterprise clients, and driving regulatory capture. [00:00:18]
OpenAI & XAI: Referenced as targets of Bernie Sanders' proposed 50% equity seizure. [00:38:42]
Coatue: Thomas Laffont's firm, cited for providing the elite venture capital math on Decacorns becoming Trillicorns. [01:00:02]
SpaceX: Noted as having a 100% Polymarket probability of going public before 2027. [00:55:44]
Palo Alto Networks: Referenced in relation to its enterprise use of Anthropic's Mythos models to fix system vulnerabilities. [00:01:09]
Heritage Foundation: Mentioned by JCal regarding their extensive data tracking project covering historical election irregularities. [01:32:03]
People
Dario Amodei: CEO of Anthropic, framed as having "AI psychosis" and delusions of grandeur, attempting to govern humanity via his model's strict guardrails. [00:00:29]
Nikesh Arora: CEO of Palo Alto Networks (transcribed phonetically as "Nesh"), whose internal technical confirmation of Mythos' capabilities was highlighted by JCal. [00:01:09]
Ben Thompson: Tech analyst used as an example of legitimate researchers being falsely flagged and punished by Anthropic's safety filters. [00:12:53]
Bernie Sanders: Senator who published an NYT op-ed demanding the public seize 50% of top AI labs to fund a sovereign wealth fund. [00:38:13]
Donald Trump: Referenced regarding job growth numbers and executive oversight capability over federal legal investigations. [00:41:00]
Elon Musk: Discussed regarding his utopian endgame vision for high AI abundance and universal baseline wealth. [00:41:29]
Sam Altman: Described as a highly practical negotiator dealing seamlessly with regulatory shifts and corporate strategies. [00:41:59]
Andrej Karpathy: Elite AI researcher hired by Anthropic to spearhead recursive model self-improvement workflows. [00:48:44]
The Collison Brothers / Ark Institute: Mentioned as the structural visionaries funding open-source DNA baseline language models. [00:19:45]
Sarah Friar: CFO of OpenAI, highly praised for outstanding executive performance during the Liquidity summit events. [00:58:43]
Thomas Laffont: Coatue executive praised for a legendary presentation breaking down the mathematical likelihood of creating Trillicorns. [01:00:02]
Jake Paul: Noted tech investor praised by the crew for highly strategic team management and strong business acumen. [01:02:41]
Thomas Keller: Legendary culinary head of the French Laundry, who hosted the elite speaker gathering. [01:02:11]
Spencer Pratt, Karen Bass, Nithya Raman: The three central figures in the highly controversial LA Mayoral primary race defined by massive late-mail statistical anomalies. [01:13:51]
Steve Hilton: Highlighted by Sacks as a potential structural remedy for California's systemic gridlock via state emergency architectures. [01:27:34]
Brenda Lee Brown Armstrong: Noted as pleading guilty in 2026 to illegal local ballot purchasing maneuvers on Skid Row. [01:24:52]
Gavin Newsom: California Governor critiqued for signing restrictive auditing structures into place over voting roles. [01:21:52]
Historical Events & Legislation
The Manhattan Project: Used as the ultimate historical precedent for dual-use technology (energy vs. weapons), arguing that we should govern the output (bombs) not the science (AI code). [00:23:46]
California Assembly Bill 1921: The specific 2018 legislation that legalized unlimited ballot harvesting, fundamentally altering the mechanics of West Coast elections. [01:15:44]
8. The Bottomline (by AI)
The defining battle of the next decade is the centralization of infrastructural power masquerading as technological safety. As the capital requirements to build frontier AI reach $100 billion per gigawatt, incumbents are weaponizing regulatory capture and censorship to kill open-source competition and establish an unassailable duopoly. Concurrently, with inflation roaring and state-level election mechanics increasingly operating via legally engineered loopholes, the friction between elite institutional control and decentralized public sovereignty is reaching a breaking point. Watch closely how the physical supply chain of compute (energy/GPUs) intersects with legislative attempts to "nationalize" AI wealth—this is where the geopolitical order of the 2030s is being silently written today.
Jun 12, 2026
Seth Klarman BREAKS SILENCE on AI, OpenAI, and Trump's "Show" of Cutting Spending | 12 Jun 2026 | iConnections
"the opportunities that happen in the worst environments can lead to returns for years afterward from deploying more capital into those markets" Seth Klarman 00:00:00 https://youtu.be/AbZYb11A Hw?si=9YNWPu0iMIuOpZQM&t=0h0m0s "people have a…
May Jobs Report
172,000
New jobs added in May, beating economist expectations by over 2x.