Large-Language Models as a Cognitive Virus

TEXT START: Large-language models (LLMs) are rapidly becoming part of human culture, reshaping how information is produced, transmitted, and used.

The Dissection

The paper models LLM adoption as contagion: transmission, coupling, persistence, recovery, reinforcement, tipping points, lock-in, and competence loss. Its useful insight is nonlinear adoption. Its strategic weakness is that it treats AI dependence primarily as a cognitive pathology rather than as a competitive response to superior productive machinery.

The framework relocates the crisis from ownership and labor displacement to individual cognition. It asks whether humans can remain mentally autonomous while largely ignoring who owns the systems, who captures the productivity gains, and whether non-users can remain economically viable.

The Core Fallacy

It mistakes dependence for the root mechanism.

Under the Discontinuity Thesis, people do not adopt LLMs merely because they are socially transmitted. They adopt them because AI increasingly delivers cheaper, faster, and more scalable cognitive output. “Immunization” may preserve unaided cognitive competence while making its recipients less competitive against AI-coupled workers and institutions.

The paper also risks treating cognitive decline as the decisive catastrophe. It is not. Post-WWII capitalism dies when AI severs the mass employment–wage–consumption circuit, even if users remain highly competent. A population can retain impressive unaided reasoning ability and still lose productive necessity. The paper models the infection’s symptoms while under-modeling the economic organ failure.

Hidden Assumptions

• Individuals can freely choose whether to couple to LLMs, despite employer, institutional, and market pressure.
• “Recovery” restores skills that still possess economic value after AI substitution.
• Adoption is driven mainly by social transmission and reinforcement rather than cost and performance superiority.
• Cognitive competence is clearly measurable and declines primarily because of LLM use.
• Reversibility remains practical after workflows, education, institutions, and infrastructure have locked in.
• Cognitive autonomy is materially available to people without ownership or control of AI capital.
• Users experience the transition similarly, rather than splitting into Sovereigns, indispensable Servitors, and economically redundant dependents.
• Reducing transmission can preserve human-only economic domains at population scale.

Social Function

Partial truth functioning as transition management and ideological anesthetic.

The viral analogy correctly identifies lock-in, nonlinear adoption, and the possibility that convenience becomes dependency. But its “immunization” language turns a structural displacement crisis into a problem of personal discipline and public-health-style intervention. That is a comforting mislocation of agency: it suggests the patient can resist the virus while the economic environment is being rebuilt around it.

For elites, the framing is convenient. It emphasizes users’ damaged cognition rather than concentrated ownership, control, and extraction. For institutions, it offers a vocabulary for managing dependence without confronting the impossibility of preserving stable human-only productive domains under P1 and P2.

The Verdict

This is a competent model of adoption dynamics and a partial diagnosis of cognitive degradation. It is not a complete account of the transition.

The decisive “virus” is not LLM use. It is AI-enabled capital severing labor from economic necessity. Cognitive immunization may preserve autonomy for selected individuals, but it cannot restore mass productive participation. Without modeling ownership, substitution, wages, and coercive competitive adoption, the paper describes the spread of the agent while missing the death of the host.