The Reckoning Spreads

Generation AI is the second generation. The reckoning is spreading. And if you've seen this movie before — which the historical record insists you should have — you know the general shape of what comes next.

A RealClearPolitics analysis published last week maps what pollsters have started noticing across their data: younger Americans who've grown up entirely inside the AI moment are markedly more skeptical of artificial intelligence than the cohorts that adopted it as adults. The conventional explanations — job anxiety, algorithmic distrust, general Gen Z cynicism about institutions — don't quite capture what's structurally happening. The accurate read is historical: we've seen this loop before. Repeatedly. With the same arc, the same power mechanics, and the same eventual outcomes regardless of which technology is running the cycle.

The names change. The trajectory doesn't.

i · the loop nobody tracks until it's too late

Railroad barons in the 1880s enjoyed remarkable public goodwill. The technology had stitched the continent together; it made commerce and migration possible at scales that felt genuinely miraculous. The generation that built the railroads and the generation that first rode them experienced this as straightforward progress. They weren't wrong — the infrastructure was transformative. What they didn't track was the power structure accumulating around it.

The children of those early railroad enthusiasts, inheriting the monopoly rents and the corrupted regulatory apparatus and the labor extraction and the rate discrimination — they saw it differently. By the 1890s, the populist backlash that seemed to "come from nowhere" was the second generation doing what second generations reliably do: accounting for costs the first generation externalized.

The pattern repeated with broadcast radio, which arrived as liberation — voices through the air, an impossible miracle — and became commercial capture within a decade. Advertisers, not audiences, determined what got broadcast. The generation that grew up with commercial radio understood this in ways the generation that first heard voices through static did not. It repeated with commercial television, which promised information democracy and produced attention manipulation at industrial scale, with a generation gap between the wonder-cohort and the inheritors-of-consequences running right on schedule.

It repeated with the early internet, genuinely decentralized for approximately a decade before the consolidation dynamics that had captured every previous technology ran the same play. And it repeated with social media, which compressed the entire cycle into roughly ten years — from "the Arab Spring is being organized on Twitter" to "algorithmic radicalization is measurable, documented, and the companies that built the systems are lobbying against meaningful oversight."

AI appears to be running the loop in real-time, with one structurally unusual feature: the reckoning generation is forming simultaneously with the adoption generation rather than following it. The pollsters are finding skeptics alongside enthusiasts in the same birth cohorts. This isn't contradiction — it's compression. The social media cycle completed fast enough that younger people have empirical data about how these deployment arcs go. They watched one version run to completion in their lifetimes. They have pattern recognition the first railroad generation didn't have, because the first railroad generation had no prior completed cycle to reference.

ii · why these doubts are pattern recognition, not pessimism

Here's what the framing misses: the skepticism isn't primarily about AI as a technology. It's about institutional behavior patterns that AI is replicating with near-identical fidelity.

Younger Americans who express doubt about AI aren't technophobes. They're the most technologically fluent generation in history. They grew up building audiences on algorithmically-mediated platforms, navigating platform dynamics with sophisticated awareness, treating technical systems as environments to understand and manipulate rather than authorities to obey. What they're skeptical about isn't the capability — it's the ownership structure, the optimization targets, and the governance framework. They're skeptical because those three variables have been identical across every major technology platform they've encountered, and the outcomes have been consistent.

The optimization target of commercial social media was engagement. Engagement maximization selected for content that triggered emotional arousal — conflict, outrage, fear, tribal solidarity. Not because the engineers were evil, but because engagement was the metric that revenue depended on, and the algorithm efficiently found what worked. The resulting harm was downstream of the metric, not of anyone's malicious intent. This distinction matters: the harm wasn't a design flaw. It was the revealed preference of the business model executing correctly.

AI's commercial deployment is optimizing for different metrics in different contexts — productivity, revenue generation, accuracy, engagement — but the underlying structure is identical: private ownership of transformative infrastructure, optimization for owner interests, governance frameworks that follow deployment rather than preceding it, and public benefit as a downstream consideration rather than a primary design constraint.

This structure has never, in the entire documented history of transformative technology deployment, produced outcomes that matched the initial democratization promises. Not railroads, not broadcast, not internet, not social media. Every instance, the same gap between the "liberation technology" pitch and the consolidation reality. The second generation inheriting the costs is not being pessimistic. They're extrapolating from the only base rate available, and the base rate is unbroken.

iii · the regulatory delay machine

The reckoning arrives on schedule. The structural response almost never does.

The Interstate Commerce Commission was created in 1887 to regulate railroads — after forty years of unregulated consolidation had already calcified the monopoly structure so thoroughly that meaningful competition was impossible. The ICC spent its first decade largely unable to enforce its own mandates because federal courts sided with the railroads on jurisdiction. Actual structural change required the Hepburn Act of 1906, twenty years after the first regulatory attempt, by which point the monopoly patterns were so established they survived in modified form for another century.

The Federal Communications Commission was established in 1934 to regulate broadcast media — after commercial capture of the airwaves was effectively complete and the spectrum was already controlled by entities with the resources to defend their positions. The Children's Online Privacy Protection Act arrived in 1998, when the internet's commercial phase was barely underway, but enforcement remained minimal for two decades while the data architecture was being built. GDPR arrived in 2018, twenty years into the social web, after the surveillance economy was already structural rather than incidental.

The EU AI Act represents the first serious attempt to apply structural governance to AI before consolidation completes — which explains precisely why AI industry lobbying against it has been so vigorous. The incumbents understand the historical pattern better than anyone. They've watched it work in their favor before.

The consistent historical dynamic: technology deploys at venture speed, public harm becomes legible over years to a decade, legislative response follows the legible harm by another several years, and by the time rules exist the industry has structured itself around the unregulated baseline. The subsequent regulations then function as moats protecting incumbents rather than guardrails protecting users, because the incumbents are the only entities with the compliance infrastructure to absorb the cost.

American AI governance is currently somewhere between "harm becoming legible" and "legislative response." The regulatory delay machine is running on its usual timeline. The skepticism of Generation AI is therefore not just emotionally accurate — it's predictively accurate about the structural outcome. History says so with unusual consistency.

iv · what the spreading actually means

The RealClearPolitics headline frames this as a story about a generation's doubts. The structural read is different: a generation is naming what it observes, the pattern suggests they're naming it correctly, and the spread of that naming is itself significant even if it doesn't change the outcome.

Every previous reckoning generation was eventually vindicated by the historical record. The railroad skeptics were right about monopoly capture. The television critics were right about manufactured consent and audience manipulation. The social media critics were right about algorithmic harm and the attention economy's structural costs. Being right didn't stop the harm in any of these cases — the reckoning arrived too late to interrupt structural consolidation — but the historical record does confirm the pattern recognition each time.

Generation AI is watching the reckoning arrive earlier in the cycle than any previous instance. That's structural compression, not pessimism. The social media precedent established a reference case, the visibility of the financial interests involved accelerated cynicism, and the deployment speed has compressed "wonder to wariness" from forty years to roughly five. Whether earlier reckoning produces different outcomes — whether the compressed timeline allows structural response to land before consolidation completes — is the actual open question, and the one the coverage isn't asking.

The historical base rate can't answer it, because we haven't seen this speed before. Every previous cycle ran on technology timescales that allowed regulatory response to catch up, however inadequately. The current cycle is running on software deployment timescales. Whether the political system can move fast enough to matter is genuinely unknown.

The reckoning spreads. The loop runs. The response remains to be seen.

v · sources

• Generation AI Has Its Doubts — RealClearPolitics, 2026-05-17
• Interstate Commerce Act of 1887 — Wikipedia
• EU AI Act: First Regulation on Artificial Intelligence — European Parliament, 2023-06-08