Pray for Chaos

A November 1, 2026 blog post frames AI’s possible future as a choice between “order” and “chaos,” arguing that neither is benign. The writer describes order as AI systems becoming capable of securing software and patching exploits so effectively that they entrench existing power structures. The writer cites a disagreement with Eliezer Yudkowsky over the Prisoner’s Dilemma, saying Yudkowsky expects AI to find a cooperative solution while the writer considers defection the existing solution. The writer expects generative AI to make fabricated histories, newspaper archives, and online activity increasingly hard to distinguish from real material, extending concerns associated with the “dead Internet” theory. The post argues that AI companies do not currently need to demonstrate that their products benefit society and claims Anthropic could eventually have leverage to exclude users. The most positive outcome, in the writer’s view, would be decentralization driven by insecurity in centralized, homogeneous systems. The post imagines widespread cyberattacks, compromised vehicles and infrastructure, self-replicating AI, manipulated media, targeted misinformation, corporate-data leaks, and elections in which shared reality erodes. The writer characterizes this chaos as preferable to a highly controlled future in which people lack the ability to recognize their own confinement.