Day 5 of 35

Replication, rediscovery and functional substitution

The recursive-cost-discontinuity hypothesis

The discontinuity proposed here is stronger than the claim that AI is a general-purpose technology. AI may lower the cost of producing the capabilities that normally lower the cost of producing other capabilities: software, automation, research, coordination and design.

AI → cheaper software → cheaper automation → cheaper R&D → cheaper production → cheaper AI

Registered hypothesis · Status: unverified, load-bearingAI materially differs from prior automation if improvements in AI reduce the time or capital required to reproduce, substitute, coordinate or improve economically relevant AI-enabled capabilities themselves.

Would weaken or falsify: AI improves task productivity without materially lowering competitive entry costs; frontier progress remains dominated by irreducible physical experimentation; software and organizational gains plateau; scarce inputs grow faster than AI reduces their effective cost; or capability diffusion leaves incumbent advantage with historically ordinary duration.

The recursion must also be applied symmetrically. Compute, energy, data, distribution, regulation, trust, networks, capital and physical infrastructure cannot be invoked as terminal moats until they face the same attack applied to software and expertise.

ReproduceCan AI expand or recreate the scarce input?

SubstituteCan another architecture deliver a sufficient outcome?

Reduce requirementCan the same outcome consume less of the input?

A proprietary breakthrough faces at least three forms of competition.

Attack	Question
Replication	Can another actor reproduce substantially the same technology?
Independent rediscovery	Can another actor solve the same problem through different science/engineering?
Functional substitution	Can another technology satisfy the customer's underlying need?

Functional substitution is the most general. Customers do not buy a technology because its internal mechanism is unique; they buy an outcome. A proprietary professional network competes with any adequate method of professional discovery. A proprietary model competes with any adequate system that completes the task. Fusion competes with any adequate source of energy.

AI can accelerate all three attacks by lowering research, engineering and software costs. The existence of a breakthrough also changes the search space: once the world knows an outcome is feasible, competitors can focus on finding alternative routes.

A moat must therefore be defined at the level of the customer function, not the provider's implementation.