Innovation theory distinguishes incremental from radical recombination. Incremental work combines ideas from nearby domains — a slightly better battery, a refined algorithm. Radical work bridges distant fields — a biological principle applied to materials science, a physics insight imported into economics. The value of radical innovation comes partly from its scarcity: few firms can bridge distant domains, so those that do earn monopoly rents on the novel combination.
Bazzichi, Riccaboni, and Castellacci model what happens when AI enters this landscape. Initially, AI increases optimal recombination distance. The mechanism is direct: AI makes it cheaper to survey distant fields, identify combinable ideas, and execute the bridge. Firms that adopt AI can profitably attempt combinations that were previously too costly to explore.
But the model identifies a threshold. As AI capability increases beyond a critical complementarity point, firms shift back toward incremental innovation. The reason is competitive: when every firm has the same AI capability, the monopoly rents from radical combinations shrink because rivals can replicate the bridge. The optimal strategy reverts to small, defensible improvements in domains the firm already knows.
The limiting case is striking. Under full automation — AI without human judgment — optimal recombination distance collapses to zero. Fully AI-driven research doesn't just narrow. It converges on the most local possible innovation, because the tool that was supposed to enable distant bridging has eliminated the scarcity that made distant bridging valuable.
The through-claim: a general-purpose bridging tool, universally deployed, destroys the value of the bridges it enables. The collapse is not cognitive but economic. The AI can still combine distant ideas. It just has no competitive reason to, because so can everyone else. The tool's power is self-canceling at the market level, even as it remains real at the individual level.