The fight over AI kept growing and never cooled off

Seven months on a single fault line

Most political fights about technology never come near the territory our hate research covers, and they are not supposed to. They are ordinary democratic combat. They do have a measurable shape, though. An event happens, and within hours commentators on each side stop arguing about the event and start arguing about what it reveals about the other camp's motives and character. This report measures seven months of that pattern on a single fault line: the fight over AI, the companies building it, and the data centers it runs on.

The instrument is our polarization methodology. We identify each narrative, state its claim in a testable form, check the claim against its own internal logic, look for the mirror narrative on the opposing side, and score the whole contest for reach, intensity, penetration, elite adoption, and mirror lock. We also check each narrative for seven markers of drift toward the territory our separate hate methodology covers. Coding a narrative here describes a framing pattern. It is not an accusation of hate, extremism, or bad conduct, and several of the people quoted on these cards are making arguments we tested and found partly supported.

More narratives at the same heat

The thirty-week figure above is the report in one place. January opened hot and narrow. The California wealth-tax fight and the first round of the jobs-data war kept two or three narratives active in a typical winter week, and what was active ran at intensity 4 on our 5-point anchors, which is essentialization territory: this is who they are, not this policy is wrong.

The spring widened the field without heating it further. Data center siting fights moved the conflict from Sacramento and Washington to township boards in Michigan, Utah, and Maine, and those narratives run cooler, at motive-attribution temperature rather than identity temperature. The April attack on the OpenAI chief executive's home put five narratives in play in a single week for the first time, as each camp built a story about what the violence revealed about the other.

By summer the field was crowded and hot at once. Five narratives were active in the window's final week, with the median intensity of the active set back at 4, and the mid-June week that matched that count came during the moratorium fight's peak. Across all thirty weeks the median intensity of what was active never fell below 3. The fault line added narratives faster than it retired them, which is what sedimentation looks like in the data.

Both camps built lenses that outlived their events

Eleven of the thirteen narratives qualified as sedimented by the end of the window, meaning the frame outlived its trigger event and reappeared on later, unrelated ones. On the tech-critical side, the oligarch lens is the clearest case. It attached to executives' job predictions in January, to a Stanford town hall in February, to the data center moratorium bill in March, to the AI sovereign wealth fund proposal in June, and to the layoff wave in July, carried by the same core of officeholders and allied commentators each time. The bought-government lens did the same work at the state level, reading the White House AI framework, Michigan's tax breaks, Maine's veto, and the federal ratepayer bill as one continuous story about officials serving the industry against their constituents.

The industry-aligned camp built lenses at the same pace. The Luddite lens started as a February editorial quip and hardened into a standing label carried by a sitting senator, partisan columnists, and a podcast with a national audience, usually welded to the claim that critics would hand the future to China. The astroturf lens read each round of data center opposition as paid, orchestrated, or foreign-aligned. The fear-monger lens recast the AI-safety camp's advocacy as a capture strategy, and by late winter it was being voiced from the Secretary of War's podium. Officials carried the frame on ten of the thirteen cards, on both sides of the line. The loudest voices in this fight are not anonymous accounts.

The same verdict came back thirteen times

We tested every narrative's claim with checks we authored ourselves, and we say so plainly: the predictions are our operationalizations of each frame, not the speaker's own stated test, and reasonable readers may formulate different checks. Where a frame supports both a narrow and a strong reading, we tested both. Every one of the thirteen verdicts came back mixed evidence, and they all failed the same way. The narrow reading survived its checks. The strongest reading failed at least one.

The pattern holds in both directions. The bought-government narrative rests on documented utility-linked political spending in Michigan, which our checks confirmed, but its strongest form, officials as a purchased class, failed against a legislature that passed the nation's first data center moratorium and a commissioner who lost a primary over an approval vote. The oligarch narrative rests on real, first-party job-replacement predictions and a real layoff wave, but its strongest form, a uniform bloc enriching itself behind false public aims, failed against the industry's own published advocacy for taxing capital to fund a public dividend.

Mirror-image results came back across the line. The confiscation narrative's wealth-flight core is roughly right at the half-trillion-dollar scale, and its precedent prediction came true when the wealth-tax idea went national, but the claim that backers intend the industry's destruction failed against the measure's own one-time drafting and its backers' stated aims. The Luddite narrative's China stakes are real, but the claim that critics are reflexive enemies of the technology failed against a moratorium bill built on stated utility-cost conditions and a Wyoming Republican defending his constituents' concerns as valid. The doomers-incited narrative has a documented core in the attacker's manifesto, but its strongest form, an anti-AI movement that is itself a violent extremist current, failed against the movement organizations' immediate condemnations and the absence of any attack pattern. The astroturf narrative's funding-lineage claim was independently confirmed, and its paid-and-bused claim failed the moment a reporter interviewed the rally crowd and found every attendee local.

We read the repetition as the signature of sedimented narratives. Each one carries a documented core that keeps it alive and an inflated edge that makes it polarizing, and the checks kept separating the one from the other. That is also why no verdict here settles any underlying policy question: a narrative can overreach while its camp's concern remains entirely legitimate, and on this fault line that was the rule, not the exception.

Nothing crossed the line, and five narratives got close

Nothing in this report's verified record cleared the bar our hate methodology enforces, in either direction, across thirty weeks. The escalation instrument recorded drift markers instead. Five narratives sit at our near-ignition status: the two mirrored attack narratives, which run on each other's worst voices; the Luddite lens, whose coined label has spread from an editorial into a standing insult documented by national radio; the fear-monger lens, which added coinages of its own, from defective altruism to woke lobotomized models, and attached them to a named company's staff; and the oligarch lens, whose collectivizing shorthand now travels at mass reach.

The single heaviest marker we logged is one eliminationist-register item: an April-era anonymous post, described in a later academic analysis, suggesting the attack on the OpenAI chief executive's home should be repeated nightly. It sits at fringe tier, we describe it in aggregate only, and it did not spread in our verified record. Near ignition is a review status that raises our checking cadence. It is not a designation of anyone as dangerous, and we ask readers not to use it that way.

The lenses lock in pairs

The strongest structural finding is how tightly the two camps' narratives cite each other. The wealth war locked first: the oligarch lens and the confiscation lens each treat the other camp's tax rhetoric as primary evidence. The April attack produced the cleanest lock we measured, with one camp reading the violence as what the industry sowed and the other reading it as what doomer rhetoric incited, each side quoting the other's harshest lines as proof. The political-money fight locked the same way, with the industry's election spending serving as proof of oligarchy on one card while the safety camp's regulatory push serves as proof of capture on the mirror card.

Our coding symmetry, disclosed every report: seven of the thirteen narratives originate in tech-critical carrier sets and six in industry-aligned ones, and all thirteen verdicts came back mixed evidence. The symmetry was measured, not assumed. One narrative refused to stay on a side at all: the surveillance-and-control frame was carried by left-populist boycott organizers in March, a MAGA-aligned open letter in May, and a right-wing commentator in July, and a former White House AI adviser partially validated its stakes premise while rejecting its remedy. The fault line's most portable narrative belongs to no single camp.

How this was measured