Public discussion of AI in the United States essentially offers only two narratives to choose from: the apocalypse prophecy of Altman and Amodei, and the compute ode of Jensen Huang. But genuine thinking often requires recognizing the tracks themselves before one can see the paths that lie beyond them.
Background and Origin
In June 2026, in a passage of interview commentary, Shen Yi observed that public discussion of AI across American society can essentially be read as a contest between two narratives. The observation looks simple, yet it touches a deeper problem: when a great power's technology debate is compressed into two preset tracks, the thinking that is truly valuable is usually not on the tracks at all.
Dissecting the Two Preset Tracks
Track One: The Apocalypse Prophets
Represented by Sam Altman and Anthropic's Dario Amodei. In this narrative, the endpoint of AI development is a god that stands above everything — humanity will either be saved or be judged. The narrative is inherently bound up with fear of superintelligence, anxiety over the "alignment problem," and appeals for regulation. Its power lies in mobilizing humanity's deepest existential anxieties — and thereby gaining authority over the conversation.
Track Two: The Toolists
Represented by Jensen Huang. AI is compared to the steam engine — it will bring disruption, and it will bring limitless possibility. The core metaphor of this narrative is that "compute is the new horsepower." It acknowledges disruption but stresses the instrumental character of technology. This discourse naturally serves the suppliers of compute and the players of infrastructure, and it requires the AI industry to keep expanding in order to sustain itself.
The Shared Limitation of the Two Tracks
Shen Yi's core insight is that although these two tracks stand opposed, they share a single premise: both presuppose the tracks themselves. The apocalypse prophets manufacture fear; the toolists manufacture demand — and both narratives serve the discursive order of particular interest groups rather than genuinely exploring the multiple possibilities of AI.
"These two narratives are different, yet both ride on tracks that have been laid in advance — everyone else can only walk along the road."
"Mature tracks decide the distribution of existing interests; new tracks decide the redistribution of power."
— 2026-06-03 23:00
The Third Possibility
Shen Yi does not directly define what the third road is, but he offers a directional judgment: "Only by refusing Altman's AI panic and Jensen's compute ode can one embrace limitless possibility."
The core of this judgment is not negation but derailment — breaking free of the cognitive framing imposed by the two narrative tracks. Only when the apocalypse prophecy and the toolist ode fail together can the discussion of AI return to directions that are concrete, grounded, and in service of real needs.
Case Verification
Case One: Narrative Positioning in Regulatory Policy
On June 3, 2026, Trump signed an executive order regulating AI models, with Anthropic identified as the targeted object of the controls. On the surface this is a policy event, but seen through Shen Yi's framework: the regulatory discussion runs entirely inside the narrative track of the "apocalypse prophets" — controls in the name of safety naturally borrow the rhetoric of AI fear, while the camp opposed to regulation borrows the freedom narrative of the "compute ode." The real questions — what kind of regulation would promote the broad sharing of technology and prevent the concentration of power — are obscured by the two discourses instead.
Case Two: The Cost Reckoning of Enterprise AI
In the same period, enterprise-grade AI in the United States is undergoing a shift from the "cash-burning race" to the "inflection point of cost rationality." Microsoft cancelled Claude Code licenses, Uber questioned the input-output ratio of its AI spending, and average monthly enterprise AI expenditure has passed $500 million. Read through the "apocalypse prophet" narrative, these signals say: AI is expanding out of control and needs intervention; read through the "toolist" narrative, they say: the inevitable fluctuations of cost. But step outside both tracks, and what emerges is this: the real value of large-scale enterprise AI deployment is now being tested — a problem that neither fear nor ode can solve.
The analytical force of this framework depends on one precondition: that the AI discussion of the subject under analysis is genuinely confined by the framing of two tracks. It is best suited to analyzing the American AI policy debate, the discourse of the technology media, and investment narratives. For discussions of AI in China, the discourse of the open-source community, or conversations at the level of concrete technical research and development, there may be more than two tracks — and they must be identified separately.
The opposition between the two tracks is not accidental — together they serve an implicit purpose: to keep the discussion of AI forever spinning inside a preset frame. The apocalypse prophets lock public attention onto the distant threat of "superintelligence," while the toolists confine the conversation to the supply-side dimension of "compute." The result of this collusion is that no one pursues the questions that matter: the real distribution of AI's social costs, the risks of power concentration, and the viable paths toward democratizing the technology.