The Conditions of an AI Capability Explosion: How Research Can Sustain or Inhibit Its Own Advance

This hypothesis/perspective examines how AI research can form, maintain, transmit, and reconstruct the conditions of its own continuation. Its direct proposition identifies changes in practitioners' problems, methods, knowledge and judgment, divisions of work, and tools that are actually taken up in subsequent practice as a possible source of sustained research capability growth. Humans themselves change in this process and are not assumed irreplaceable. The revised counterintuitive proposition is conditional: stronger solving and cheaper judgments may increase genuine local success while, under particular rules for selecting and advancing research, making the path fit existing success forms and impairing conditions needed by actual successors. Reduced change alone is not inhibition. Adequate existing methods, productive concentration, equivalent substitutes, and compensation can defeat a claim of damage. The paper distinguishes product, applicability, actual use, and condition reformation; compares close work on AI breakthrough exhaustion, scientific monoculture, opportunity costs, and reward overoptimization; and treats evaluation-driven discovery as a substantive constructive alternative. Bounded mathematical and archival illustrations preserve local positive facts, failed regenerations, unknown execution, and competing arithmetic explanations. A restricted identification argument and an unrun cap-set factorial design specify what discriminating evidence could look like. An appendix gives C01–C13 precise statements, assumptions, evidence identities, defeaters, and meaning-preserving paraphrases. No new experiment, sustained capability explosion, field-wide net harm, or global priority is claimed. This independent manuscript extends three earlier MaoField preprints, identified by their specific version DOIs: 10.5281/zenodo.23331705, 10.5281/zenodo.23332354, and 10.5281/zenodo.23353640. It does not assign the earlier positive hypothesis's dates to the later inhibition mechanism.