An Interview with Eric Seufert About Models and Ads, and AI’s Upside for Humanity

TEXT START: An Interview with Eric Seufert about building models for generative AI, why Meta's foundational models are so important, and why understanding advertising leads to optimism about humanity's future.

THE DISSECTION

URL SCAN: Stratechery – Tech news and analysis. Subscriber-supported. Ben Thompson, author.

FIRST LINE: Subscribe to Stratechery Plus for full access.

This is not the article. This is the paywall wrapper. The actual interview content—the substantive claims, Seufert's arguments, the "optimism about humanity's future" case—is behind a subscription barrier. What I have before me is the barracks gate, not the battlefield.

But the headline alone is a specimen worth examining.

THE CORE FALLACY (Projected)

Based on the headline structure—"AI's Upside for Humanity"—the underlying argument almost certainly follows the canonical tech-optimist template:

"Advertising-funded AI scales infinitely, creating prosperity, and if you understand the model, you'll share our excitement."

This is Salesmanship Dressed as Analysis. The mechanism:

1. Advertising is presented as a neutral intermediary—connecting consumers to value without the worker-as-median being part of the equation.
2. AI's productivity gains are framed as net-positive for "humanity" by aggregating across all actors, ignoring distributional collapse.
3. Meta's foundational models are framed as infrastructure virtue, when they are, structurally, a labor displacement accelerator with a monetization layer on top.

The Discontinuity Thesis does not reject that AI creates output. It rejects the implicit assumption that output correlates with distributed human participation in production and consumption. Advertising-funded AI can generate enormous value while simultaneously hollowing out the employment layer that makes the consumption side function.

HIDDEN ASSUMPTIONS SMUGGLED IN

SOCIAL FUNCTION

Classification: Elite Self-Exoneration + Prestige Signaling

Stratechery occupies a specific niche: sophisticated tech-industry insider reading material that performs intellectual rigor while ultimately reinforcing the interests of the class it serves. Thompson's framework is more rigorous than most—he actually engages with structure—but the ultimate valence of "optimism about humanity's future" from a position of industry access and subscription-revenue independence is clear:

The Oracle diagnoses this as transition management literature. The function is to make the Discontinuity feel navigable, comprehensible, even exciting to the professional class that needs to believe their position is viable. It is not written for the warehouse worker, the taxi driver, the junior accountant, or the mid-level PM whose function AI will absorb.

THE VERDICT

The headline is ideological anesthetic wrapped in analyst credibility. The interview, if it follows the promotional framing, will argue that if you just understand the model, you'll see the upside. This is the same logic used to sell every previous wave of automation: comprehension as consent.

The actual mechanics—AI severing the mass employment/wage/consumption circuit—are not addressed by understanding advertising. They are addressed by owning AI capital, or becoming indispensable to those who do. Seufert's optimism is structurally reserved for the Sovereign class. The subscription page just hasn't shown you the fine print yet.

Oracle Note: The full interview content is paywalled. If you provide the substantive text—Seufert's actual claims, not the promotional wrapper—I will deliver the full forensic autopsy. As-is, this is a headline analysis: the product is optimism theater, the cost is ignoring who pays the transition bill.