Hyper-Novelty and the Adaptation Gap
Civilisational fragility rises when change rate exceeds evolutionary bandwidth
Hyper-novelty, a concept developed by evolutionary biologist Bret Weinstein and his collaborator Heather Heying, describes the condition where the rate of environmental change exceeds the evolutionary bandwidth of any organism or system built to adapt to it. All creatures are well-built for the environments in which they evolved, but adaptation is always lagging rather than leading change. As the pace of technological and social change accelerates, humans and their institutions end up progressively more poorly adapted to the world they actually have to navigate.
The framework extends from individual cognition to institutional behaviour. Human beings have an unusually long developmental period precisely because acquiring adult competency requires a long apprenticeship in a stable environment. That program breaks down when the environment in which lessons are learned bears no resemblance to the environment in which adult decisions must be made. The same logic applies to organisations — universities, newsrooms, public health agencies — that were built for a slower-moving world.
The practical implication is a proliferation of risk vectors. Each year of accelerating change adds uncovered surface area that no adaptation cycle has reached. Applied to AI specifically, Weinstein argues that humans have zero evolutionary preparedness for a world where machine-generated narrative out-competes human narrative, since narratives are foundational to identity, coordination, and social trust. The framework predicts not one catastrophic failure but a widening field of simultaneous vulnerabilities.
- Adaptation always lags selection pressure; no organism or institution is prepared for what has not yet happened.
- The longer the required developmental period, the more brittle the system becomes when the environment changes faster than the development cycle.
- Uncovered risk vectors accumulate: each year of unmatched acceleration adds new failure modes that no adaptive cycle has reached.
- AI risk in narrative space is uniquely dangerous because narrative is the substrate of human coordination and social trust.
- Regulatory asymmetry compounds the problem: compliant actors are constrained while non-compliant actors absorb the capability gains.
- Map your rate of change against your adaptive bandwidthIdentify the domain you are analysing — an industry, an institution, a personal skill set. Estimate how rapidly the environment is changing versus how quickly the system in question can update its competencies, norms, or tooling. A gap here is a fragility signal.Pro tipFocus on the second derivative: it is not just that change is happening, but that the rate of change is itself accelerating that makes adaptive lag structural.
- Enumerate uncovered risk vectorsList the threat categories the system was not designed for. For an individual this might mean AI-generated misinformation, synthetic media, or automated job displacement. For an institution it might mean platform disintermediation or regulatory arbitrage by non-compliant actors.WarningAvoid anchoring on the most salient risk — the framework's power is in cataloguing the full surface area, not optimising against one scenario.
- Distinguish adaptation-possible from adaptation-impossible threatsSome risk vectors can be addressed through deliberate learning and structural change; others cannot be adapted to within any realistic timeframe. The former warrant investment; the latter warrant mitigation or avoidance. Weinstein's own heuristic: invest in a cognitive toolkit that works for a future you cannot yet imagine — generalist, interpersonal, and adaptive skills.Pro tipFor AI risk specifically, Weinstein suggests mandating reasoning trail logging so that when failures occur, the causal chain is recoverable rather than opaque.
- Apply diminishing-returns preparation logicFocus action on the steep face of the benefit curve — high-impact, low-cost interventions regardless of probability estimates. Skip preparations designed for scenarios that are not survivable. The goal is increasing capability to respond, not eliminating all risk.WarningOver-indexing on a single catastrophe scenario (bunker thinking) can crowd out the high-ROI general-purpose resilience building that helps across a wide range of outcomes.
Weinstein describes what he calls the Cartesian Crisis: AI-generated content is dissolving the chain of logic, evidence, and reason that allows factual consensus to form. When deepfakes are indistinguishable from authentic footage and when institutional journalism has already lost credibility, the capacity to establish shared facts — the basis of coordinated social action — breaks down entirely.
Humans require the longest developmental period of any species, during which they absorb the environmental context needed for adult competency. This period is calibrated to a relatively stable environment. When the world changes faster than the development cycle — when what you learn as a child has no bearing on what you need as an adult — the entire adaptive investment produces maladapted adults.
Weinstein and Heying developed the hyper-novelty concept through their evolutionary biology research and their experience at Evergreen State College, where they observed institutional systems fail in real time under rapid social pressure. Their thesis is grounded in standard evolutionary theory — adaptation lags selection pressure — but extrapolated to civilisational scale as a diagnostic for why modern institutions appear to malfunction simultaneously rather than in isolation.
Weinstein articulated the AI dimension of hyper-novelty on this Diary Of A CEO episode, arguing that the production of compelling machine-generated narrative represents a domain for which there is no evolutionary precedent. Unlike tools that augment physical or computational capacity, AI that competes in narrative space attacks the mechanism by which humans establish shared reality — a function that evolved over hundreds of thousands of years in stable social contexts.