C.19.1:1 - Problem frame

Bespoke computational heuristics can win locally while failing to exploit larger compute, data, or search budgets. General methods can improve across a declared scale window, but the empirical record is strongest for computational search, learning, and planning. A module organization, institution, work arrangement, or episteme does not inherit that empirical result by analogy. A project may still adopt a broader policy, but it must state the bearer-specific scale relation and evidence rather than treating all growth, reuse, or generality as one phenomenon.

Without this separation, teams either repeat the Bitter Lesson as a slogan or impose a costly machine-learning experiment recipe on bearers for which seeds, FLOPs, compute slopes, or data sweeps have no meaning.