General-purpose platforms make capabilities borrowable, widening the range of viable startups and eventually displacing incumbent-led commercialization; yet large legacy capability systems and platform governance (fees, access rules) delay incumbents' adaptation, creating persistent hysteresis in industry structure.

Main Finding

General-purpose digital platforms that make productive capabilities callable through standardized interfaces (platform-mediated horizontal capability access) fundamentally change industry evolution. As platform quality improves, more entrepreneurial projects become commercially viable; platform-mediated commercialization can displace incumbent-mediated collaboration at a threshold of platform quality; incumbents’ inherited routine–capability systems create a gap between evolutionary selection and productive efficiency and generate rational hysteresis; and platform governance (fees, access rules) shifts the thresholds at which entry and regime change occur.

Key Points

• Conceptual innovation
  • Defines "platform-mediated horizontal capability access": using capability functionality supplied via standardized platform interfaces without owning underlying asset stocks or negotiating project-specific transfers.
  • Distinguishes three acquisition modes: internal accumulation, bilateral access, and platform-mediated horizontal access.
• Core theoretical results (from the model)
• Platform improvement expands the set (measure) of entrepreneurial projects that can be profitably commercialized (an extensive-margin increase in organizational variety).
• There exists a platform-quality selection threshold: below it incumbent-mediated collaboration is evolutionarily fitter; above it, platform-mediated independent commercialization outcompetes incumbents.
• The evolutionary selection threshold can occur before the platform route produces greater total surplus because incumbents’ outside options in bargaining protect them—so selection can shift earlier than aggregate productivity gains.
• Incumbents with larger or more specific inherited routine–capability systems convert later (higher platform-quality threshold), producing rational hysteresis between population-level selection and incumbent adaptation.
• Role of institutions
  • Platform rules (fees, API terms, data rights, eligibility) materially affect which organizational variants are feasible and the platform-quality thresholds for entry and regime shift.
• Theoretical framing
  • Uses a history-friendly evolutionary modeling approach to link micro-level commercialization behavior and meso-level industry structure changes.
  • Emphasizes that platforms change the recombination operator available to entrepreneurs by decoupling use from ownership for modular capabilities.

Data & Methods

• Model structure
  • Analytical evolutionary model with a continuum of heterogeneous entrepreneurial projects, each with intrinsic value and capability requirements.
  • A general-purpose platform provides capability of quality q(t) that exogenously improves over time (q̇ = g(q)).
  • Entrepreneurs choose between commercializing via the platform or via incumbent-mediated collaboration; incumbent cooperation is modeled as a bargaining problem in which platform commercialization is the entrant’s outside option.
  • Incumbents decide when to convert inherited routine–capability systems into platform-native architectures; conversion cost rises with the size and specificity of the incumbent’s inherited system.
• Solution approach
  • Derivation of viability/fitness conditions for projects under both commercialization architectures, identification of quality thresholds, and comparative statics on governance variables (fees, access conditions).
  • Numerical illustrations to map formal results to industry-like settings (e.g., sports analytics, API-driven AI applications); proofs collected in appendix.
• Assumptions and limitations
  • Platform quality path treated as exogenous to isolate downstream transmission mechanisms (section discusses endogenous feedback as an extension).
  • Applies to modularizable capabilities that can be supplied through standardized interfaces; tacit/firm-specific capabilities remain outside scope.
  • Stylized (history-friendly) rather than fully empirically calibrated — intended to clarify mechanisms and generate testable predictions.

Implications for AI Economics

• For AI-enabled industries and startups
  • APIs to foundation models, cloud ML services, and other callable AI capabilities lower the cost of assembling viable product-market combinations, raising entry and innovation by independent entrepreneurs.
  • Access to improving platform-level AI capabilities changes the frontier of feasible AI applications without requiring startups to own large model weights, data centers, or specialized infrastructure.
• For incumbent firms and industrial dynamics
  • Incumbents with large, specific legacy systems (data, pipelines, customer relationships) face higher conversion costs and therefore may persist longer even after platform-mediated forms become prevalent — producing observable hysteresis in market structure.
  • Platforms can alter the mix of competition vs. collaboration: initially reinforcing incumbent roles via bargaining, later enabling displacement as platform quality rises.
• For welfare, policy, and governance
  • Evolutionary selection for platform-mediated entrants can precede welfare-improving adoption, because incumbent outside options (bargaining power) slow efficient reallocation — suggesting a role for interventions that align private and social incentives.
  • Platform governance (pricing, interoperability, data-access and IP rules, eligibility/discrimination policies) materially shapes industry evolution — regulators and platform designers can meaningfully bias which organizational architectures become dominant.
  • Policy levers that lower platform access frictions (interoperability, fair API terms, data portability) can accelerate productive reallocation; conversely, exclusionary/platform-fee strategies raise thresholds and slow transitions.
• Testable empirical predictions and recommended measurements
  • As platform capability (e.g., model quality, latency, cost per inference) improves over time, entry rates of independent AI firms relying on platform APIs should increase, and a compositional shift toward platform-based commercialization should occur after a measurable threshold.
  • The selection threshold (observed market shares shifting toward platform-native entrants) can occur before aggregate productivity/welfare gains are realized—measure incumbents’ bargaining terms, licensing fees, and outside options to test this wedge.
  • Incumbents with larger, more specific legacy assets should show later conversion/adaptation timing and greater persistence post-platform-improvement.
  • Potential data sources: API usage and pricing histories (OpenAI, AWS, Google Cloud), firm-level entry and exit records, partnership/licensing agreements, platform governance changes or fee shocks as natural experiments.
• Broader research directions
  • Endogenize platform investment and feedbacks: study how downstream adoption shapes platform quality trajectories and strategic platform governance.
  • Empirical case studies: mobile app ecosystems, API-driven fintech (Stripe), communications (Twilio), and foundation-model ecosystems to validate thresholds, hysteresis, and governance effects.
  • Welfare-focused work to quantify the gap between evolutionary selection and social optimum and to evaluate policy interventions (interoperability mandates, fee caps, mandated data portability).

If you want, I can: - Extract concrete empirical hypotheses and propose datasets and identification strategies for testing the model. - Draft a short literature map linking this paper to recent work on foundation models, platform ecosystems, and evolutionary industrial dynamics.