Exaptation in AI Development
Exaptation, a term Gould and Elisabeth Vrba introduced in 1982, names the phenomenon by which a feature that evolved for one function (or no function—a spandrel) is subsequently co-opted for a different one. The feather evolved for thermal regulation in flightless theropod dinosaurs and was exapted millions of years later for flight. The swim bladder (buoyancy control) was exapted into the lung. Jaw bones became middle-ear bones in mammals. In each case, the most transformative function was not the original function. Applied to AI, exaptation explains why the most consequential uses of language models are not the uses they were designed for. Claude was optimized to predict text. But creative collaboration, architectural brainstorming, serving as intellectual sounding board—the uses constituting Segal's orange pill experience—are exaptations. These capabilities emerged as structural consequences of fluent context-sensitive generation and were discovered by users who co-opted them for purposes the designers never targeted. If the most important consequences of AI are exaptations, they cannot be predicted by examining designed purposes—the industry optimizes for adaptations while transformation comes from exaptations that optimization cannot target because it cannot foresee them.
In the AI Story
Gould argued that exaptation might be more important than adaptation for understanding evolutionary innovation. Adaptation refines existing functions—making organisms better at what they already do. Exaptation creates new functions—opening ecological possibilities that did not previously exist. The feather-as-insulation is adaptation refined over millions of years. The feather-as-flight-surface is exaptation opening the entire domain of powered aerial locomotion, previously inaccessible to vertebrates for three hundred million years. Adaptation makes organisms better; exaptation makes them different. The differences produce the most consequential changes.
The Trivandrum engineer who built a complete user-facing feature in two days despite never writing frontend code was exapting a text-prediction system for unintended purposes. The backend engineer discovering Claude could hold architectural conversations was exapting a token predictor into cognitive prosthesis. Segal's collaboration with Claude—producing insights neither could achieve alone—was literal exaptation: text prediction co-opted for intellectual partnership. Each represents a feather-to-flight transition: capability arising as prediction byproduct, repurposed for function prediction was never designed to perform.
The pattern is visible across technology history. The internet was designed as military communication infrastructure (survive nuclear attack by routing around damage). Its exaptation into the World Wide Web, e-commerce, social media produced consequences dwarfing original military application. GPS was designed for military navigation, exapted into consumer mapping and location services with greater economic impact. In every case, designed function is the swim bladder; unintended use is the lung. The exaptation is more consequential than the adaptation—the pattern Gould's framework predicts for any complex branching system.
The forecasting implication is humbling. If exaptation theory is correct—and evidence from biology, technology history, and LLM deployment supports it—the most consequential AI uses in 2035 are probably not uses anyone in 2025 has imagined. They will emerge from intersections of structural properties with unarticulated human needs, in uncreated contexts, through co-optations that cannot be anticipated. The roadmap maps adaptations. The territory is shaped by exaptations. The map is useful but radically incomplete—it cannot include what it cannot foresee.
Origin
The exaptation concept emerged from Gould and Vrba's 1982 effort to provide evolutionary biology with vocabulary for features serving functions they did not originally evolve for. The term was deliberately chosen to complement 'adaptation'—parallel structure, distinct meaning. An adaptation is shaped by selection for its current function. An exaptation arose for a different function (or as a spandrel) and was subsequently co-opted. The distinction is fundamental: it reorients how functional structures' histories should be understood, revealing that many of evolution's most consequential innovations (vertebrate hearing, terrestrial respiration, powered flight) emerged through co-optation rather than direct adaptation.
Key Ideas
Most transformative function was not original function. The swim bladder did not evolve for breathing; jaw bones did not evolve for hearing—co-optation was contingent, unpredicted, improbable from original perspective.
Creative collaboration is exaptation. No one designed Claude for intellectual partnership—capacity emerged as spandrel of fluent generation, discovered by users and co-opted.
Industry optimizes for adaptations. Benchmarks, roadmaps, funding priorities target intended capabilities while most transformative consequences are exaptations emerging in the wild.
Exaptations cannot be anticipated. They emerge from intersection of structural properties with specific unforeseeable needs in specific circumstances—the flight that the feather discovers.
Historical pattern is consistent. Internet, GPS, and now AI exhibit the same dynamic—designed function is adaptation, unintended use is exaptation, exaptation is more consequential.
Appears in the Orange Pill Cycle
Further reading
- Gould, S.J. and Vrba, E.S. 'Exaptation—A Missing Term in the Science of Form' Paleobiology 8:1 (1982)
- Gould, S.J. 'Exaptation: A Crucial Tool for Evolutionary Psychology' Journal of Social Issues (1991)
- Larson, G. et al. 'Exaptation: A Missing Term in Design Vocabulary' Design Issues (2012)
- Andriani, P. et al. 'Exaptation in Innovation' Research Policy (2017)