AIPolitics 18: The Political Cleavage Rotated and AI Has Bisected It

What asymmetric fragmentation tells us about how the next round of tech politics gets fought

Something interesting caught my attention this week, and well, dear reader, I am going to inflict it on you.

Democrats’ Techlash Trap, sixty-five pages, just released by Dave Vorland at the Chamber of Progress through its new Blue Horizon Project. Vorland was a Biden-era Deputy Assistant Secretary of Defense for Space and Missile Defense Policy. His credentials run through Obama and Biden, his current work runs through the center-left tech-industry coalition, and his argument is sharp.

Vorland’s case is that the Democratic Party spent four years executing a tech-policy posture that lost both the working-class voters it was trying to win and the digitally-fluent young men it was taking for granted. Vorland calls the posture “faculty lounge populism,” borrowing Adam Jentleson’s line about Yale Law School graduates cosplaying as populists. The data points are sharper still. David Shor’s November 2025 polling puts the Mag 7 average net favorability at +34 while “Big Tech” as a political category sits at -2. A thirty-six-point gap between the companies voters actually use and the political category Democrats spent four years attacking. The Pew validated voter analysis of 2024 has Trump beating Harris among non-college voters fifty-six to forty-two. A thirty-one-point youth gender gap between men and women aged eighteen to twenty-nine. Eighty percent of consumers prefer ad-supported services to paying subscriptions, contradicting the surveillance-capitalism public-demand assumption Khan’s FTC used. Musk spent three hundred million for Trump; the tech-sector counter-mobilization at single-donor-PAC scale was effectively zero.

The piece is worth reading. So are the questions it raises, because they’ve been things we’ve been talking about here at SCBBQ from the jump.

The question for this piece is why Vorland’s faction is alone among Democrats in being pro-AI right now, and why the Republican party’s pro-AI coalition looks like three different fractures wearing one coat. The answer to both questions runs through a structural realignment story I started telling in Part 5 of this series, and the cleanest way to see what’s happening in 2026 is to put the realignment side-by-side with the new AI fracture and look at the geometry.

So before we get to Vorland, it’s time to bring back “the clock.” (click to enlarge, you know you want to.)

The Clock, Briefly

In Part 5 I borrowed a framework from my friend Jennifer Victor, along with my friend the late Scott McClurg, who built on Miller and Schofield’s work on partisan realignment. They mapped it on a simple two-dimensional chart that I keep coming back to. Put the economic dimension on the horizontal axis, left to right, redistributive to free-market. Put the social dimension on the vertical, liberal at the top, conservative at the bottom. Now draw the line that best separates the two party coalitions.

In 1960, that line was nearly vertical. The dominant cleavage was economic. New Deal politics. Labor versus capital. You could be socially conservative and vote Democratic if your economic interests pointed that way, and plenty of people did.

Between then and 1996, the line rotated as culture was entering the equation. Religious right activism, the post-Reagan coalition. (Note, dear reader, this is when I started graduate school and was writing with Abramowitz about polarization and ideological realignment and all that jazz.) And here’s where I have to put a marker down, because the next move in this story is one I’ve spent a quarter-century arguing about.

Carmines and Stimson’s Issue Evolution (1989) makes the elite-driven case: party elites strategically positioned on racial issues after 1964, and voters followed along. Top-down. Race as the precipitating issue, parties as the active movers. That’s the standard account, and it’s not wrong about what issue triggered the rotation. I’ve long thought it’s incomplete about the mechanism.

What Alan Abramowitz and I argued in the 1998 Journal of Politics, and through the 2008 polarization debate and after, is that the realignment runs bottom-up at least as much as it runs top-down. As party elites differentiated ideologically through the 1970s and 80s and 90s, voters who held consistent ideological orientations started sorting into the party that matched their underlying views. The elites didn’t drag the voters into a new alignment. The elites finally got clear enough about where they stood that voters could see the choice and act on it. Voters with consistent left-leaning positions across multiple issue domains found themselves at home in the Democratic Party. Voters with consistent right-leaning positions found themselves at home in the Republican Party.

The liberals-as-Democrats and conservatives-as-Republicans realignment was the visible result of that sort, and that was the political equilibrium for a decade-plus.

This matters for the AI fracture coming in 2026. If the realignment is purely elite-driven (the Carmines-Stimson story), then the AI question becomes whatever the party elites decide to make it, and the voters fall in line. If the realignment runs through voter-level ideological sorting (the Abramowitz-Saunders story), then the AI question is going to sort the way voters’ underlying ideological commitments sort, regardless of which elites are currently running the policy machinery. Different mechanisms make different predictions about what 2028 looks like. I want that argument on the table before we look at the matrix.

OK. With that noted, the visual story.

By 2016, and this is when Jen and Scott were writing the original piece on the clockwork rise of Donald Trump’s populism, the line had rotated further, almost becoming the x-axis, separating the upper from the lower (aligning the liberals with the educated cosmopolitans on one side, and the populists and the conservatives on the other). The working cleavage was the cosmopolitan, socially liberal, economically left-of-center versus the populist, socially conservative, economically skeptical of redistribution folks. Sides, Tesler, and Vavreck’s Identity Crisis named identity politics as the dominant 2016 sorting variable, which is consistent with the rotation framing and adds a specific mechanism (identity-based rather than abstract “culture”).

And by 2024? The cleavage divides almost completely the cosmopolitans from the populists. Not old school left and right. Not liberal to conservative. Cosmopolitan vs. populist. The single strongest predictor of where you fall on that line isn’t income, isn’t race, isn’t religion (though all of those matter). It’s education. College-educated voters sorted Democratic. Non-college voters sorted Republican. Six decades of voters with consistent ideological commitments matching their party affiliation to those commitments, and education turning out to be the cleanest demographic proxy for the underlying ideological cluster.

Now: this line doesn’t perfectly sort within the parties, and that matters. The Democratic coalition still has a left-populist wing, the Sanders constituency, working-class Democrats, the people whose politics are populist but who ended up on the cosmopolitan side of the partisan divide anyway. They’re Democrats, but they were never credentialed-class Democrats. They didn’t sort into the coalition because of the credential. The Republican coalition still has its cosmopolitan-Republican wing, the Liz Cheney and Mitt Romney holdouts. So, the sort is real, but it’s never clean. The cleavage on the left panel of the graphic above is drawn with a slight upward-left tilt because of exactly this. The line is real. The line is porous.

A note on precision. The degrees in this story are visual approximations. Victor and McClurg didn’t put specific angles on the cleavage line; I have, well, because that’s how I visualize this in my own brain and because the rotation visual is doing analytical work in this piece. The framework is a heuristic for thinking about realignment, not a statistical estimate of how the underlying issue space sorts. Don’t take “ninety degrees” too literally.

So, that is the clock. It has been rotating clockwise since 1960. By 2024 it had completed roughly ninety degrees of rotation, and the credentialed-cosmopolitan-liberal class had sorted from a small Republican wing into the dominant Democratic majority while the populist-conservative class made the symmetric journey in the other direction. The class structure didn’t change. The partisan affiliation of each class changed, because voters within those classes held consistently positioned ideological commitments and sorted into the party that matched them as the parties became clearer about where they stood.

That was the realignment that preceded all this AI stuff. And you can’t understand what AI is doing to American politics in 2026 without understanding what that realignment did first, and what mechanism it ran on. The bottom-up sorting account matters for what comes next. The AI question is going to sort voters the same way the post-1964 ideological realignment did. People who hold consistent commitments about technology, displacement, regulation, and innovation will find themselves in the party whose elites match those commitments, regardless of which elites are currently running the policy machinery.

What Vorland Says

The piece is a sustained argument that Biden-era tech policy was a strategic loss on both fronts. Vorland names the architects. Inside the administration: Lina Khan at the FTC, Jonathan Kanter at DOJ Antitrust, Gary Gensler at the SEC, Lael Brainard at the National Economic Council. Outside the administration: Matt Stoller at the American Economic Liberties Project, Jeff Hauser at the Revolving Door Project. The intellectual substrate: Khan’s Yale Law Review article “Amazon’s Antitrust Paradox”, Shoshanna Zuboff’s “surveillance capitalism” framework from Harvard, the Brandeisian antitrust revival that Stoller and Hauser have spent a decade institutionalizing.

The faculty-lounge-populism critique runs like this. The administration’s regulators wanted to be populists. They believed that taking on Big Tech and Big Capital would deliver working-class voter gains. They believed the public wanted aggressive antitrust enforcement and the breakup of Silicon Valley. They believed that the credentialed-cosmopolitan policy class could lead an economic populist revival from inside the federal regulatory apparatus.

The frame fits Khan, Stoller, and Hauser cleanly. They are credentialed Brandeisian antitrust advocates pursuing a structural-monopoly agenda from inside the federal regulatory apparatus, and “faculty lounge populism” captures the disconnect between their elite-academic policy framing and the voter coalition they assumed they were serving. The frame is a worse fit for Bernie Sanders, whose anti-AI position is a displacement frame rooted in a half-century of working-class organizing rather than in elite-academic populism. Lumping Sanders with Khan blurs an analytically important distinction inside the anti-AI Dem coalition. Both wings critique Big Tech. They get there from different starting points and they would govern through different remedies.

The voters did not agree with the regulatory program either way. Slingshot Strategies polled voters in August 2024 about what they wanted the FTC to focus on. The top-ranked actions were data security and breach notifications (eighty-seven and eighty-four percent support), banning hidden fees (seventy-nine percent), and “click to cancel” rules for subscriptions (seventy-seven percent). The bottom-ranked actions were the headline-grabbing tech merger lawsuits Khan’s FTC built its reputation around: suing Meta over Instagram and WhatsApp (twenty-nine percent), suing Meta to block its acquisition of a VR fitness company (twenty-four percent), suing Microsoft to block Activision Blizzard (nineteen percent). Voters wanted protection. The FTC delivered antitrust theater. (The standard caveat: Slingshot is a Democratic firm and the poll is commissioned partisan research. Take the framing accordingly. The directional pattern matches other public polling on FTC priorities.)

The polling on advertising tells the same story. Eighty percent of consumers prefer ad-supported services to paying subscriptions. Roughly ninety percent favor ads tailored to their interests. Khan’s FTC proposed a sweeping “commercial surveillance” rulemaking explicitly built on the surveillance-capitalism framework, which assumes voters loathe targeted advertising. The rulemaking stalled. The framework was wrong about the public it was supposedly protecting, at least at the stated-preference level.

The crypto crackdown alienated voters Democrats need. Paradigm’s October 2024 polling shows that crypto owners are sixty-eight percent Gen Z or Millennial, forty-eight percent non-white, seventy percent earning under one hundred thousand dollars. Forty-two percent of men aged eighteen to twenty-nine have used crypto, compared with seventeen percent of women the same age. Senator Warren’s self-described “anti-crypto army” is, in coalitional terms, a tool for alienating the under-forty multiethnic working-class base. Sanders’s broader anti-AI displacement framing reaches the same constituency through a different door, and arrives at the same coalition cost. The two anti-AI positions converge in their voter consequence even though they differ in their theory of harm.

The closed-door policy did the rest. Vorland documents that the Biden Administration never filled the Chief Technology Officer position, kept the chair empty for all four years, after Obama’s three CTOs in succession (Aneesh Chopra, Todd Park, Megan Smith). Secretary of Defense Lloyd Austin did not visit the Pentagon’s Silicon Valley hub at the Defense Innovation Unit until December 2024, after Trump had been re-elected. Marc Andreessen has publicly described his Biden-era engagement attempts as “absolutely horrifying” and unproductive. A Semafor analysis from January 2025 found that “when corporate America called, the Biden Administration often didn’t pick up the phone.”

Vorland’s diagnosis is partisan and funder-aligned. The Chamber of Progress is a center-left tech-industry coalition with corporate partners. The Blue Horizon Project’s explicit mission is to “restore tech optimism in the Democratic Party.” Take the framing with the corresponding skepticism. The data points are sourceable separately, with the standard caveats. Shor is a Democratic-aligned pollster. Slingshot is a Democratic firm. Paradigm’s crypto polling is published. Pew’s validated voter is its own analysis. The “three hundred million Musk for Trump, effectively zero tech-Dem counter-mobilization” framing flattens that Harris raised significant tech-industry money (Hoffman among others); the absence was specifically at single-donor-PAC scale, not at tech-industry political engagement broadly. The structural pattern Vorland is naming, that the Biden-era tech posture lost working-class voters and tech-industry allies simultaneously, is real even if you don’t buy the “faculty lounge populism” frame as a complete account.

Vorland is the only diagnostician in his coalition making this argument at this scale, in this institutional voice, with this much documentation. The Democratic Party’s anti-AI wing, by contrast, has multiple institutional voices, decades of intellectual scaffolding, and a clear faction-internal consensus. The Vorland faction is the minority position inside the Dem coalition. Why?

The Four-Quadrant Map

To see why Vorland’s faction is in the minority position, put the AI fracture next to the partisan cleavage. The graphic above does that.

The left panel is the clock as it has rotated from 1960 to 2024. The right panel is the same two-dimensional space, drawn for 2026. The partisan cleavage is the same line, drawn cleanly horizontal on the right for matrix clarity, drawn with a slight tilt on the left to keep the historical work honest. The right-panel red line is the same partisan cleavage that has been rotating for sixty years. By 2024 it has settled into the horizontal position, separating Democrats above (liberals and cosmopolitans, the credentialed class) from Republicans below (populists and conservatives, the non-credentialed class).

The new line in 2026 is the perpendicular blue dashed vertical on the right panel. That’s the AI fracture. On the left side: anti-AI. On the right side: pro-AI. The line cuts through both parties at the elite-policy level.

A caveat before the walk-through. Whether the AI fracture is a cleavage at the mass-electoral level is a different question from whether it sorts elites. Pew’s recent issue-priority polling shows AI ranks well below inflation, immigration, healthcare, abortion, and crime as voter priority. What the four-quadrant matrix maps is the policy-class fracture among coalition elites and donor networks. The mass-electoral implications are slower and less clear. The matrix still does analytical work; it describes elite factional sorting, not voter behavior at scale.

Now look at where the four factions of the two-party tech debate sort.

The top-left quadrant (anti-AI Democrats, occupying the “liberals” position from the clock) holds the credentialed-cosmopolitan policy class. Lina Khan at the FTC, Elizabeth Warren in the Senate, Bernie Sanders with his displacement-focused critique, Matt Stoller at AELP, Jeff Hauser at the Revolving Door Project, Jonathan Kanter at DOJ Antitrust, and the labor unions (UAW, SEIU) whose policy work fed the regulatory apparatus throughout the Biden term. Within the quadrant, two sub-positions: the credentialed-Brandeisian (Khan, Stoller, Hauser, Kanter) and the displacement-populist (Sanders, the unions). They share the anti-AI position. They get there differently.

The top-right quadrant (pro-AI Democrats, occupying the “cosmopolitans” position from the clock) holds Vorland and the Chamber of Progress, Reid Hoffman, Alex Bores with his AI Dividend bill, Ritchie Torres, Jim Messina, and Gavin Newsom at the state level (vetoed SB-1047, signed the retain-not-replace executive order, signed SB 53). This is the Vorland faction. Real, named, and the minority position among the Democratic policy class in advocating for an institutional pro-AI posture from within the coalition.

The bottom-left quadrant (anti-AI Republicans, occupying the “populists” position from the clock) holds Josh Hawley, Steve Bannon, Tucker Carlson (more anti-Silicon-Valley-elite than anti-AI-technology specifically), Marsha Blackburn with her child-safety legislation, the economic-nationalist wing, and parts of the MAGA base whose techlash is structurally similar to the AELP critique but arrived at from the opposite ideological direction. On AI specifically, Hawley and Khan rhyme rhetorically. Their policy specifics diverge significantly. Khan’s antitrust enforcement targeted mergers; Hawley’s tech legislation focuses on Section 230 reform and TikTok bans. The rhetorical overlap is real. The policy-substance overlap is much weaker.

The bottom-right quadrant (pro-AI Republicans, occupying the “conservatives” position from the clock) holds Marc Andreessen, Peter Thiel, David Sacks (Trump’s AI czar), JD Vance (whose February 2025 Paris AI Action Summit speech was an explicit accelerationist statement at the VP level), Elon Musk, a16z and Founders Fund, and Garry Tan at Y Combinator. The Trump executive branch operates here. Most of the institutional AI policy direction from Washington in 2026 comes from this quadrant. Within the quadrant, two sub-coordinations: accelerationist VCs (Andreessen, Thiel, Sacks, Musk) coordinated through the All-In podcast, donor pipelines, and direct executive access; libertarian classical-liberals (Adam Thierer at R Street, Tyler Cowen at GMU, Mercatus and Cato) operating from think-tank substrate. I have placed the libertarians in the non-partisan zone below the matrix because they are not Trump Republicans.

The four-quadrant matrix surfaces three structural facts that get lost in conventional coverage.

First, each party’s anti-AI wing rhymes with the other party’s anti-AI wing at the level of rhetoric and theory of harm, less so at the level of policy specifics. Khan-Warren-AELP and Hawley-Bannon make structurally similar arguments about Big Tech concentration, corporate power, surveillance, and worker displacement. They arrive at those arguments from different cultural and ideological starting points, and they propose substantially different remedies. Khan’s antitrust enforcement targeted mergers; Hawley’s tech legislation focuses on Section 230 reform and TikTok bans. The rhetorical rhyme is real. The policy-substance rhyme is much weaker, because the two wings disagree about what to do even when they agree about who the villain is.

Second, each party’s pro-AI wing is fragmented internally, though the Pro-AI Republican fragmentation operates with a coordination mechanism (the Trump donor pipeline, the All-In podcast network, Sacks-as-czar centralization, direct executive-branch access) that the Pro-AI Dem fragmentation lacks. Fragmentation in the presence of effective coordination produces a different political result than fragmentation in the absence of coordination. The Pro-AI Republican coalition has been more policy-coherent in 2026 than the fragmentation argument might predict, partly because the donor pipeline provides a substrate the Pro-AI Dems don’t have.

Third, the minority-faction problem. The Pro-AI Democratic quadrant exists, has documented institutional substance, and has a sixty-five-page reset document. The faction is real, named, and institutionally anchored. It is also the minority position inside the Dem coalition. The Pro-AI Republican quadrant has more numerical weight in 2026 (VPs, AI czars, executive-branch direction, three hundred million in campaign spending) and no equivalent institutional reset document in the Vorland format. The Pro-AI Republicans got their tech policy by funding Trump directly and now run it transactionally through executive-branch CEO relationships and the Sacks centralization. The Pro-AI Democrats are doing the institutional reset work because they have to. They lost.

The structural difference between the two coalitions is in their building methods, not in their pro-AI energy. One side is rebuilding through coalition-internal intellectual scaffolding. The other side is running policy through donor-CEO transactions plus think-tank substrate. That asymmetry is the story.

---

(Quick aside for new readers: this is the eighteenth piece in the AI Politics series. The full archive is here. If this stuff is doing analytical work for you, subscribe and the next one shows up in your inbox.)

---

The Asymmetry Compounds

A sixty-five-page institutional reset document does specific work in a coalition. It establishes a vocabulary. It names architects to oppose. It puts data points on the record. It gives a faction a memorandum to circulate, a piece of paper to point at when arguing inside the coalition for staffing decisions or platform positions. The Vorland piece does all of that work for the Pro-AI Democratic faction. The faction now has something to point at.

The Pro-AI Republican faction has not produced an equivalent document in this format. Republican intellectual production happens; it happens in different formats. The Andreessen Horowitz blog, the All-In podcast, R Street and Mercatus policy papers, Heritage’s Project 2025 chapter on tech policy, Sacks’s executive-branch direction, Vance’s Paris speech transcripts, Andreessen’s Techno-Optimist Manifesto. These are real intellectual scaffolding. They are not sixty-five-page Chamber-of-Progress-style reset documents because the Pro-AI Republican faction is in power and doesn’t need to write reset documents. They execute policy directly.

So the asymmetric institutional production is about format and political position more than about quantity. Vorland writes a coalition-internal reset document because that’s what an out-of-power faction in the institutional-policy track does. Andreessen runs Trump’s AI policy through David Sacks because that’s what an in-power faction with executive-branch access does. Both are intellectual work. Both are institutional production. They look different because the political positions are different and the formats serve different coalition needs.

The asymmetric structure has consequences for governance. A tech policy that runs through CEO-by-CEO transactions and Sacks-style centralization is coherent in the short term and brittle over time. It works as long as the executive branch is responsive to the relevant donors, and as long as those donors agree about what policy to deliver. The moment Andreessen and Vance disagree (and they will, as their coalitional commitments diverge), the transactional framework starts breaking. The Pentagon-Anthropic blacklist that this series has been tracking since Part 14 is one early example: a faction of the executive branch made a policy call that contradicted the Pro-AI consensus the donor pipeline expects to deliver, and the result is a DC Circuit case in active litigation. The coordination held for most decisions and broke on this one. Watch for more of those breaks.

A tech policy that runs through Vorland-style institutional reset documents is slow to implement and durable in the framework it establishes. Vorland’s faction does not run the executive branch. They cannot deliver policy in 2026. What they can do is win the internal Democratic fight for staffing decisions, platform positions, and intellectual influence between 2026 and 2028, with the goal of being institutionally positioned to deliver coherent policy when the political opportunity opens. That work is slower and less visible than what Sacks and Andreessen are doing in 2026. The work is also more durable. If Vorland’s faction wins the intra-party fight, the next Democratic administration’s tech policy will be more coherent than the current Republican one, for a few cycles at least, until the next coalition shift.

The asymmetry shapes 2026 American tech politics. Slow durable institutional rebuilding on one side. Coordinated transactional execution on the other. Both sides know the AI question is consequential. Both sides have factions that want to take it seriously. The faction that has done the documented coalition-internal reset work is the one out of power, and the faction in power is running policy through transactions plus think-tank substrate that don’t add up to a comparable coalition-internal reset.

Where Anthropic Fits

This series spent Part 17 arguing that Anthropic’s appearance at the Vatican alongside Pope Leo XIV’s first encyclical was a frontier-lab choice to align with the Catholic Church over the White House. The four-quadrant map gives that move a second reading. Three readings, actually, since the simpler explanations need to be on the record first.

The simplest explanation for the Vatican move is that Pope Leo XIV is a mathematician by training, his encyclical’s interpretability framing required a frontier-lab interlocutor with serious mechanistic-interpretability research, and Olah’s work matched the encyclical’s intellectual needs. Part 17.5 walked the Teilhard de Chardin connection that explains why this specific theological framework matched this specific research program. None of that requires the four-quadrant analysis. The Vatican move was also a brand-and-positioning play. The Pope generates international media reach no American political alliance could match. The simpler explanations hold.

The four-quadrant map adds a compatible structural reading. Look at the four American political quadrants from Anthropic’s commercial position. The Pro-AI Democratic quadrant does not run the executive branch. The faction cannot offer Anthropic the institutional substrate it needs. They are rebuilding. The Anti-AI Democratic quadrant explicitly antagonized Anthropic-style frontier labs for four years. The Pro-AI Republican quadrant runs the executive branch transactionally, but the dominant accelerationist intellectual framework within that quadrant rejects the safety-encoded posture that Anthropic’s identity is built on. The Pentagon-Anthropic blacklist case is the live evidence of that rejection. The Anti-AI Republican quadrant treats Anthropic as another Big Tech firm to discipline; Hawley would happily regulate Anthropic alongside Meta and Google.

None of the four American political quadrants was offering Anthropic the kind of institutional substrate Anthropic’s identity wanted, at this moment in the political cycle. The Catholic Church was. That’s why the Vatican move was rational even after you account for the simpler explanations. The Vatican was the available substrate that fit; the American political substrate that would have fit was either not in power, not interested, or actively hostile.

The Vatican move also gets an even sharper reading at the coalition level. Vorland’s faction is the closest available American institutional voice for Anthropic-style AI engagement. The Chamber of Progress framing (pro-innovation, pro-engagement, regulated-and-responsible) overlaps substantially with Anthropic’s public posture. If the Pro-AI Democratic faction were in power, Anthropic might still have walked to Rome for the brand-and-positioning reasons, but the institutional walk-away would be less pronounced. The faction is not in power. Rome was the next institutional substrate.

The cross-quadrant compatibility framing in Part 17 needs a qualifier here. Compatibility with deployment across factions is a generic feature of frontier-lab products at scale. Pro-AI Republicans use OpenAI’s ChatGPT enterprise contracts. Anti-AI Democrats use OpenAI’s workplace tools. The same is true for Anthropic. What’s distinctive about Anthropic-Vatican specifically is the intellectual content of the match: interpretability research, the Teilhardian theological substrate, the safety-encoded posture mapping onto the encyclical’s framing. Other frontier labs are commercially compatible across quadrants. Anthropic was intellectually compatible with the encyclical’s specific theological program in a way the other labs were not. The Vatican picked Anthropic because of the intellectual content, not because of generic compatibility.

So the structural reading the four-quadrant map adds is narrower than Part 17’s frame. The Vatican move was a brand play plus an intellectual match plus a coalition-substrate walk-away. The four-quadrant analysis explains the walk-away. The other two explanations explain why Rome specifically.

What This Predicts

The four-quadrant map is a snapshot of where the elite factions are in May 2026. It is not stable. The next two years of American politics will move pieces around the matrix, and the question is which faction gains and which loses in the rearrangement.

The 2026 midterm tests one specific thing: whether the Anti-AI Democratic frame (Khan-Warren-AELP credentialed populism plus Sanders displacement-populism) produces electoral gains. Vorland’s argument is that it will not. His argument depends on the polling cluster he’s documented holding through the cycle. If Shor’s Mag 7 favorability gap holds, if the youth gender gap holds, if non-college voters continue to break Republican in the mid-fifties, then Vorland’s faction has the empirical evidence to win the intra-party fight after 2026 about whether to continue the Biden-era posture or reset. If those polling clusters move (if Big Tech favorability drops, if young men shift back to Democrats on tech-policy grounds, if non-college voters return to the Democratic coalition), then Vorland’s faction has lost the empirical ground for its argument.

The 2028 presidential cycle is the more consequential test. Vorland’s piece ends with “lessons for politics” and “lessons for governance” sections that read like a primary-platform document for a hypothetical Democratic candidate willing to break with the Khan-AELP wing. Newsom is one possible carrier of that platform, given his vetoes and his retain-not-replace position. Whitmer is another. Buttigieg has the policy background and the institutional standing. The Bores AI Dividend bill is the legislative test case for whether redistributive-pro-AI framing can carry a Democratic primary.

For the Republicans, the 2028 question is whether the accelerationist VC bloc can keep the populist-right techlash in the same coalition. Vance is the structural inheritor of the Trump tech-policy track, and his Paris speech is publicly accelerationist, but his Catholic identity and his Thiel-funded coalition donor base pull in directions the accelerationist position cannot fully satisfy. The Pope’s encyclical names AI as “an instrument of domination, exclusion and death” when deployed in the accelerationist mode Vance has championed. Vance has not publicly resolved the tension. The 2028 primary may be when he’s forced to.

For the frontier labs, the question is whether the Anthropic-Vatican move is the start of a pattern. Anthropic chose Rome because the intellectual match was specific and because the American political substrate at this scale wasn’t offering what the company needed. Other frontier labs face the same problem in different ways. OpenAI’s transactional posture toward the Trump administration is structurally similar to what Anthropic chose to walk away from. DeepMind’s UK-based Google parent operates in a different institutional substrate. xAI runs through Musk and is currently inside the Trump donor pipeline. Mistral and other European labs run through European institutions. Watch for additional frontier-lab moves into institutional substrate outside US partisan politics over the next twenty-four months. The IMF, the BIS, the WHO, and the OECD are candidate venues. The Vatican won’t be the only one.

The predictions above are worth tracking, probably not worth betting on. Predicting primary-cycle outcomes three years out has a poor empirical track record, and Cohen et al.’s “Party Decides” framework has been complicated by Trump 2016 and the subsequent revisions. This series has been better at structural-institutional analysis than at electoral prediction. The four-quadrant map is a tool for thinking about what is happening. The map is not a tool for predicting which party wins in 2028.

Where We End Up

The clock has been rotating since 1960. By 2024 it had completed roughly ninety degrees of rotation, and the credentialed-cosmopolitan-liberal class had sorted from a small Republican wing into the dominant Democratic majority while the populist-conservative class made the symmetric journey in the other direction. The rotation tracked race, culture, education, and income in shifting weights through the period. The rotation is the structural fact preceding everything this series has been writing about for two years.

The AI fracture in 2026 is the perpendicular line at the elite-policy level. The mass-electoral implications are slower and less clear. Both party coalitions now have anti-AI populist-aligned wings whose arguments rhyme rhetorically even when their policy specifics differ. Both party coalitions have pro-AI elite wings whose intellectual frameworks fragment internally. Only one of the two pro-AI wings has produced an institutional reset document in the Vorland format. The other coalition produces intellectual work in different formats (podcasts, executive orders, think-tank papers, VC blogs) and doesn’t need to produce a Vorland-format reset document because it’s running policy directly.

The asymmetric institutional production is the analytical move worth holding onto. The Pro-AI Democrats are writing sixty-five-page reset documents because they lost in 2024 and have to rebuild through coalition-internal scaffolding. The Pro-AI Republicans are running tech policy through CEO-by-CEO transactions and Sacks-style centralization because they won in 2024 and execute policy directly. Both factions are pro-AI. Their institutional behavior in response to being pro-AI looks different because they are in different political positions and the format conventions of their respective coalitions are different.

The frontier labs are observing the asymmetry and routing around it. Anthropic walked to the Vatican because the Pope-Anthropic intellectual match was specific and because none of the four American political quadrants at this institutional scale was offering a Vorland-coalition-style substrate that was politically available. If that’s the leading indicator, the next twenty-four months will see additional frontier-lab moves into institutional substrate outside American partisan politics. The frame moves while the institutions sleep.

The question really is whether either party builds a coherent post-2024 tech-policy framework before the 2028 cycle, or whether the AI question continues to be adjudicated by frontier labs in foreign and trans-national institutional venues. Vorland’s piece is one of the few attempts on either side of the partisan aisle to rebuild through coalition-internal documents rather than route around. The faction that has produced the document is the only American institutional voice currently building a Vorland-format substrate that frontier labs could engage with. That faction is also not in power. Whether they win the intra-party fight, and whether the next Democratic administration carries their framework, determines whether the AI question gets brought back into American democratic politics or stays where Anthropic just took it.

The clock has rotated, and well, AI politics has bisected it. The four-quadrant map is, at least to my eye, the geometry of where we are now. What happens to the geometry between now and 2028 is probably all we’re going to talk about as AI as an issue becomes even more salient than it already is.

Yay?

---

The 2x2 and who is where is open for argument—just an attempt at mapping! Where do you see yourself on it? Where do you see me getting the placements wrong? The bottom-up sorting prediction is testable, and I’ll be tracking how voters actually move on AI through the cycle. If you’ve got a faction I missed, a placement that doesn’t sit right, or a piece of polling that complicates the picture, drop it in the comments. The Sanders-Hawley populist-rhyme is the one I keep going back and forth on, in particular.

And if this piece moved your thinking on 2026 or 2028, share it, re-skeet it, re-tweet it, restack it, whatever. Series like this one compound when readers share, and the next installment lands…well, when I get it done.