When a person learns a new domain, what they already know changes the cost. This paper formalizes how: dissolved knowledge — skills and patterns practiced to zero conscious processing cost — forms a structural codebook that generates candidate matches when encountering unfamiliar material. Where genuine processing structure is shared between domains, candidates survive filtering against reality and the new domain dissolves faster. Where structure is not shared, candidates fail and the domain costs full price. The mechanism is hypothesis generation through the reduction pipeline, not pattern matching. Transfer is proportional to structural overlap, an empirical property of each domain pair. The paper defines six measurable levels of domain knowledge from None through Master, each separated by a testable boundary transition. It identifies four conditions required for dissolution — rapid feedback, manageable iteration cost, context consistency, and sufficient repetitions — that predict which domains yield to acquisition and which resist regardless of prior knowledge. It formalizes the operational method: name, simplify, connect, compare to reality, formalize or build, test adversarially, fail and restart without hedging. Two disciplines underpin the method — failable design, where all internal logic fails loudly on violations, and testing range, where practice against what defeats you maps validity envelope boundaries before they produce unrecoverable failure. Six falsifiable predictions are stated with explicit test methods and falsification criteria. The framework is mechanistic, empirical in posture, and makes no claim of universality beyond what its predictions survive.
Geoffrey Howland (Mon,) studied this question.