The Child Is Splashing, and the Water Is Running Out: Where Stochastic Scaling Is Taking Us — and the Deterministic Bypass

Counts differ because each database indexes a different set of publications. We treat OpenAlex as the canonical count; Google Scholar is not shown (no API, and crawling it violates its ToS).

Something changed in 2026 that is not being described accurately, because describing itaccurately requires holding two facts at once. The first fact: frontier AI labs convinced the executive branch of the United Statesgovernment to treat AI data centers as strategic national infrastructure. Permitting was expedited. Nuclear and federal land were put on the table. What began as a researchparadigm became a national industrial program. [verify specific EO/policy citations] The second fact: the technical case for that paradigm has never been made. It has beenassumed. Scaling laws describe a relationship between compute and loss on held-out data.They do not describe a path to general intelligence. The industry substituted a metric for a theory, and then convinced a government to pour concrete on the basis of the metric. This paper is about where that leads.