Biological evolution provided the hardware — the brain, the vocal apparatus, the social instincts. Cultural evolution provided the software — language, institutions, accumulated knowledge. Computational adaptation now provides a third layer: systems that traverse, synthesize, and extend the knowledge landscape at a speed that makes cultural evolution look glacial. A large language model can be trained on the accumulated output of human culture in weeks. It can be retrained in hours. The binding constraint has shifted.
The institutional consequence is severe. The rate at which AI systems generate knowledge now exceeds the rate at which human institutions — universities, regulatory bodies, professional organizations, editorial boards — can evaluate that knowledge. The gap is not closing. It is widening with each capability improvement. The institutions were designed for cultural-evolution timescales; they now operate in a computational-evolution environment, and the mismatch produces specific failure modes: regulatory frameworks that address last year's capabilities, curricula that train for skills the machine now performs, certification systems that credential competencies the market no longer values.
The framework illuminates You On AI's central frustration with institutional response. The EU AI Act, American executive orders, emerging frameworks — all address supply-side questions (what companies may build) on cultural-evolution timescales while the computational evolution of capabilities outpaces them. The demand-side questions (what citizens, workers, students, parents need to navigate the new environment) receive even less attention, because demand-side institutions — education systems, professional bodies — are even slower to adapt.
The solution is not to slow computation. That horse has left the barn. The solution is to accelerate the adaptive capacity of human institutions — but institutions are themselves products of cultural evolution and carry the inertia of their history. A university does not redesign its curriculum in a quarter. A certification body does not redefine its standards in a year. The structures that shape how humans relate to knowledge are deeply embedded and resistant to rapid change. Which means the AI transition's most consequential challenge may be not technical but institutional: whether existing structures can adapt faster than their history suggests is possible.
The three-timescale framework draws on Joseph Henrich's cultural evolution synthesis, extending it to computational adaptation. Agüera y Arcas's specific articulation emerges from his Santa Fe Institute work on complex adaptive systems.
Three qualitatively different systems. Biological, cultural, and computational adaptation operate on different mechanisms, not merely different speeds.
The mismatch is structural. Institutions designed for cultural timescales cannot keep up with computational ones, and the gap widens rather than closes.
Evaluation, not generation, is the binding constraint. The species that defined itself by accumulating knowledge now faces the challenge of evaluating knowledge generated faster than any institution can process.
The demand side is neglected. Policy focuses on what companies may build; citizens must navigate the new environment largely without institutional support.