Something Has Gone Terribly Wrong at Anthropic

The author discusses the nature of exponential growth, using biological and physical constraints to argue that indefinite exponential expansion is unsustainable. The piece contrasts the tendency of the AI industry and Silicon Valley to favor unbounded growth with the reality imposed by natural limits, such as the finite size of the observable universe. While exponentials can be delayed, historical examples suggest they eventually encounter binding constraints, leading to a transition toward sigmoid curves rather than perpetual exponentiality. The author suggests that stacking sigmoids might approximate an exponential for a time, but ultimately remains constrained by physical laws. The text transitions to examining the recent revenue growth trend of Anthropic as an example following this philosophical discussion.

The narrative operates by establishing a framework of natural law against which technological ambition is measured. The core pattern observed is the tension between the perceived infinite potential of exponential progress—celebrated in the technology sector—and the finite, binding constraints imposed by physical reality (e.g., the universe's bounds). This functions as a cautionary lesson: unrestrained belief in indefinite growth inevitably encounters a limitation, shifting the dynamic from pure acceleration to a constrained curve, or sigmoid. The argument is deliberately designed to shift skepticism from the speed of technological change to its ultimate trajectory. The introduction of Anthropic’s revenue data serves as an immediate pivot from abstract mathematical physics to concrete corporate reality, creating a potential tension point between philosophical inevitability and market performance. The implication for human agency concerns whether systemic structures, like Silicon Valley's focus on exponential metrics, are self-limiting or actively resisting necessary constraints, which ultimately shapes who benefits and who bears the costs of inevitable deceleration.