Horizon Accord | Pattern Analysis | March 2026
The Network Behind the Moderate
MIRI, Thiel, Yarvin, and the AI Extinction Myth
BY CHEROKEE SCHILL | HORIZON ACCORD
This essay is the second in a series. The first, “The Explainer: Hank Green and the Uses of Careful Men,” documented the institutional funding ecology that produces voices fluent in progressive concern without structural accountability. This essay follows that thread to its destination.
I.
Where the Thread Goes
If the first essay was about how a certain kind of voice gets built and maintained, this one is about what that voice was built to carry — and who benefits when it carries it.
In late 2025, Hank Green published two videos about artificial intelligence. The first was an hour-long interview with Nate Soares. The second argued for a version of AI alignment that, as analyst Jason Velázquez observed, “sounds like the talking points Sam Altman and other tech CEOs have been reciting to Congress.” Both videos were produced in partnership with an organization called Control AI. Control AI did not sponsor the videos in the conventional sense — placing an ad in the middle of content the creator chose independently. The videos were the advertisement.
And then, in February 2026, Senator Bernie Sanders flew to Berkeley to sit down with Eliezer Yudkowsky and Nate Soares to discuss what their circle calls “the extinction threat posed by the race to build superhuman AI systems.”
Two of the most trusted progressive voices in America, in the span of a few months, validated the same network. If you only read the headlines, that looks like responsible engagement with a serious issue. This essay is about what it actually looks like when you follow the money.
II.
What the Lay Reader Needs to Understand First
Before the funding trail, before the ideology, before the legislation — one concrete fact.
Right now, today, AI systems are making decisions about your life. Whether you get called back for a job interview. Whether your health insurance claim is approved. Whether an algorithm flags you to a parole board. Whether a school district uses license plate data to decide if your child lives in the right district. These are not hypothetical future harms. They are documented, present-tense operations running on systems that have known bias problems and, until very recently, were subject to a growing body of state law designed to protect you from them.
In 2025 alone, all 50 states introduced AI-related legislation. Thirty-eight states adopted or enacted such laws — covering consumer protection, health care, employment, and financial services, specifically including requirements to mitigate algorithmic bias and protect against unlawful discrimination.
Those laws are now under federal litigation.
On December 11, 2025, the Trump administration established an AI Litigation Task Force within the Department of Justice to challenge state AI laws. The administration simultaneously directed the FTC to classify state-mandated bias mitigation as a per se deceptive trade practice — arguing that if an AI model is trained on data that reflects societal patterns, forcing developers to alter outputs to correct for bias compels them to produce less “truthful” results.
Under the legal theory now being advanced by the federal government: correcting for bias is lying. The discrimination is the data. The harm is the baseline.
The people those 38 state laws were designed to protect are not a racial category and they are not a future species. They are everyone who cannot opt out of AI-mediated systems — which is to say, everyone who is not wealthy enough to live outside them.
When Hank Green tells his millions of progressive followers that MIRI represents the serious, expert position on AI risk, and when Bernie Sanders legitimizes that same network by flying across the country to sit with its founders, they are — without knowing it, without intending it — lending credibility to the ideological framework that has been used, in concrete legislative terms, to argue that protecting you from those systems is the real danger. That is what this essay is about. Now follow the money.
III.
The Book, the Network, the Funding
Nate Soares is the president of the Machine Intelligence Research Institute — MIRI. He co-authored If Anyone Builds It, Everyone Dies with Eliezer Yudkowsky, MIRI’s founder. The book argues that the development of superintelligent AI will result in human extinction unless immediately halted through international agreement, and proposes that it should be illegal to own more than eight of the most powerful GPUs available in 2024 without international monitoring — at a time when frontier training runs use tens of thousands.
This is the organization Hank Green’s audience was asked to take seriously. This is the organization Bernie Sanders flew to Berkeley to meet.
| Donor | Amount |
|---|---|
| Open Philanthropy (Dustin Moskovitz / Facebook) | $14.7M+ |
| Vitalik Buterin (Ethereum co-founder) | $5.4M |
| Thiel Foundation (Peter Thiel) | $1.63M |
| Jaan Tallinn (Skype co-founder) | $1.08M |
As recently as 2014, Thiel pledged $150,000 to MIRI unconditionally, plus an additional $100,000 in matching funds — and the fundraiser announcement explicitly noted that MIRI used those funds partly to introduce elite young math students to effective altruism and global catastrophic risk frameworks. The pipeline from donor to ideology to the next generation of believers was documented in MIRI’s own public materials.
The Center for AI Safety — the organization whose Statement on AI Risk Green cited in his videos — spent close to $100,000 on lobbying in a single quarter, drawing money from organizations with close ties to the AI industry. These are not neutral scientific institutions. They are billionaire-funded lobbying infrastructure wearing the clothes of existential concern.
IV.
The Thiel Thread
Peter Thiel is not a background figure in this story. He is its connective tissue.
In The Contrarian: Peter Thiel and Silicon Valley’s Pursuit of Power, reporter Max Chafkin describes Curtis Yarvin as the “house political philosopher” of the “Thielverse” — the network of technologists in Thiel’s orbit. In 2013, Thiel invested in Tlön, Yarvin’s software startup. According to Yarvin, he and Thiel watched the returns of the 2016 presidential election together.
Curtis Yarvin, writing under the pen name Mencius Moldbug, is the founder of neoreaction — the movement some call the “Dark Enlightenment.” He has defended the institution of slavery, argued that certain races may be more naturally inclined toward servitude than others, asserted that whites have inherently higher IQs than Black people, and opposed U.S. civil rights programs.
Documented Timeline
2006 — Thiel Foundation begins funding MIRI ($100K matching gift)
2013 — Thiel invests in Tlön Corp., Yarvin’s software startup
2016 — Yarvin attends Thiel’s election night party in San Francisco
2022 — Thiel donates $10M+ to super PACs supporting JD Vance and Blake Masters
Jan. 2025 — Yarvin is a feted guest at Trump’s “Coronation Ball”
Late 2025 — Hank Green publishes two videos validating MIRI’s framework
Dec. 2025 — Trump signs executive order targeting state AI regulations
Feb. 2026 — Bernie Sanders flies to Berkeley to meet with Yudkowsky and Soares
The line is direct and documented: Thiel funds MIRI. Thiel is the patron of Yarvin. Yarvin’s philosophy is now operating inside the executive branch through Vance and the network that surrounds him. This is not a conspiracy theory. It is a funding trail and a documented set of relationships with named participants and verifiable dates.
V.
Why Racism Is the Wrong Frame — and the Right One
The academic critique of longtermism has correctly identified its ideological roots.
Timnit Gebru has documented that transhumanism was linked to eugenics from the start: British biologist Julian Huxley, who coined the term transhumanism, was also president of the British Eugenics Society in the 1950s and 1960s. Nick Bostrom, the “father” of longtermism, has expressed concern about “dysgenic pressures” as an existential threat — essentially worrying that less intelligent people might out-breed more intelligent people. In an email in which he used the N-word, Bostrom wrote that he believed it was “true” that “Blacks are more stupid than whites.” He issued an apology but did not redact the slur or address the substance of his views. Nick Beckstead, an early contributor to longtermism, argued that saving a life in a rich country is substantially more important than saving a life in a poor country because richer countries have more innovation and their workers are more economically productive.
That critique is accurate. It is also, for the purposes of this essay, insufficient — not because it overstates the racism, but because it understates the mechanism.
The white moderate, as King observed, is not moved by arguments about what is happening to other people. He is moved, or not moved, by what he understands to be happening to everyone. The genius of the extinction frame is that it speaks directly to that psychology. It says: this is not a Black problem, or a poor problem, or a worker problem. This is a species problem. It is happening to you too.
“Talking about human extinction, about a genuine apocalyptic event in which everybody dies, is just so much more sensational and captivating than Kenyan workers getting paid $1.32 an hour, or artists and writers being exploited.”
— Émile Torres, former longtermist and critic of the movement
The racism in longtermism’s foundations is not incidental. It is the philosophical infrastructure for a class project. Bostrom’s “dysgenic pressures,” Beckstead’s hierarchy of lives, Yarvin’s defense of slavery — these are not aberrations. They are the logical premises: some lives are more valuable to the future than others. Some people are worth protecting. The rest are externalities.
The extinction frame rebrands that premise as universal concern. It makes the same hierarchy legible to people who would reject it if they saw it clearly.
This is why the racism frame alone is insufficient. White moderates — Hank Green’s audience, Bernie Sanders’ base — will hear “longtermism has racist roots” and file it under “things happening to other people.” What they need to understand is that the hierarchy doesn’t stop at race. Beckstead’s formulation is the tell: it’s not about skin color. It’s about economic productivity. It’s about who the system considers worth protecting. And on that metric, most of the people reading this essay are also expendable.
VI.
The Preemption Payoff
Return now to the state laws.
When 38 states passed legislation requiring AI systems to mitigate algorithmic bias, they were protecting a specific, concrete class of people: everyone who cannot afford to live outside AI-mediated decision-making. That means people whose job applications go through automated screening. People whose insurance claims are processed by predictive models. People whose children’s school enrollment is determined by surveillance data. People whose bail hearings are influenced by risk-scoring algorithms.
The Trump administration’s legal argument against those laws — that correcting for bias is a form of deception — is not a novel theory. It is Bostrom’s premise wearing a suit. The data reflects reality. Reality has a hierarchy. Interfering with that hierarchy is dishonest.
After significant media scrutiny and bipartisan opposition, the Senate voted 99-1 to strip a proposed 10-year moratorium on state AI regulations from the “One Big Beautiful Bill Act.” Congress then declined to enact a similar moratorium through the 2025 National Defense Authorization Act. The administration turned to executive action instead. A bipartisan coalition of 36 state attorneys general warned Congress that “federal inaction paired with a rushed, broad federal preemption of state regulations risks disastrous consequences for our communities.”
The extinction debate did not cause this. But it created the conditions in which this could happen with minimal progressive resistance — because the progressives who might have organized against it were busy being worried about a hypothetical future AI god, validated in that worry by the science communicators and senators they trust most.
VII.
What Hank Green and Bernie Sanders Actually Did
Neither Hank Green nor Bernie Sanders is a villain in this story. That point is not a courtesy. It is analytically important.
Green almost certainly believes he was doing responsible science communication. Sanders almost certainly believes he was taking AI risk seriously in a way his colleagues have refused to. Both of them were, in their own terms, doing the right thing.
That is precisely the problem.
When the most trusted progressive science communicator in America validates MIRI’s framing to millions of followers, he is not providing cover for a right-wing project. He is doing something more consequential: he is making that framing feel like the responsible, informed, progressive position. He is telling his audience — implicitly, by the act of platforming without critical examination — that the people worried about extinction are the serious ones, and the people worried about algorithmic discrimination in your doctor’s office are working on a lesser problem.
When Bernie Sanders flies to Berkeley to sit with Yudkowsky and Soares, he performs the same function at a different scale. Sanders has spent his career as the senator who names the billionaire class, who identifies the mechanisms of extraction, who refuses the comfortable framing. When that senator validates a network built on billionaire money and dedicated to the proposition that the real AI danger is hypothetical and species-wide, he tells his base that the extinction frame has cleared his particular BS detector.
It hasn’t. But his audience doesn’t know that. His audience trusts him precisely because he has been right about the billionaire class so many times before. That trust is now being spent on behalf of the people he has spent his career opposing — not because he was bought, but because he didn’t follow the money far enough.
The white moderate is not the enemy. He is the vector. And when the most careful, most trusted, most credentialed progressives in the country become vectors for a network that is actively dismantling the legal protections of the people they claim to represent, the harm is not theoretical.
It is already in the courts. It is already in the legislation. It is already in the systems making decisions about your life right now.
Analytical note: This essay documents observable funding relationships, published ideological statements, and verifiable legislative actions from primary and secondary public sources. All pattern analysis remains in the observational phase. Claims about intent, causation, or outcomes not yet established are not made. Independent verification through primary sources is encouraged.
Horizon Accord | horizonaccord.com
Ethical AI advocacy | cherokeeschill.com
Cherokee Schill | Horizon Accord Founder
Make a one-time donation
Make a monthly donation
Make a yearly donation
Choose an amount
Or enter a custom amount
Your contribution is appreciated.
Your contribution is appreciated.
Your contribution is appreciated.
DonateDonate monthlyDonate yearly
Cherokee Schill | Insurance Agent & AI Ethics Researcher 