Punctuated Equilibrium in Tech Frontiers: 77.8% of Capability Advances Occur in Single Years Across 1.74 Million Records

The arXiv preprint by Dawoon Jeong and colleagues manually codes 29 capabilities then scales to 204, showing that median frontier product share is only 1.46% while commercial variety remains broad. Forward out-of-sample tests reveal that manufacturer competition, population dynamics, and product-composition signals add no predictive power beyond frontier-history baselines, contradicting demand-pull expectations common in innovation economics. This pattern implies that application-specific valuation and switching costs, rather than observable market aggregates, gate which technical opportunities scale. The work therefore isolates the minimal data required to test whether price-quantity observations at the application level can explain why most frontier gains remain economically narrow.

Punctuated-equilibrium theory, originally developed for paleontology, here maps onto technology by separating rare, concentrated pushes from the long tail of incremental offerings. The authors note that stylized switching-cost calculations identify candidate mechanisms linking scale economies to the economic relevance of jumps. Their proposed language-model extraction of dated demand documents offers a replicable route to validate these mechanisms against independent market data, addressing a gap in prior qualitative histories of semiconductors and displays.

Policy and R&D planners should therefore treat frontier forecasts as history-dependent processes rather than market-signal aggregates. The finding that additional commercial indicators fail to improve accuracy suggests resources spent on broad market surveillance may yield lower returns than targeted tracking of the narrow set of leading specifications. Future validation requires matched price and quantity series at the application level to test whether the observed concentration arises from valuation thresholds or production indivisibilities.