Design lesson

Universal Decision Learning says that decision making extends partial behavior into new contexts by candidate generation and global consistency. Infinitesimal Decisions asks whether that extension remains coherent under typed local change. The answer is richer than a gradient of expected utility. It retains which half of the universal construction changed, which variations are behaviorally null, where nonsmooth switching occurred, and which finite decision tests must still be passed.

This is the general decision-theoretic role of tangent LINCS. It turns the local sensitivity of a decision system into an inspectable structural object without reducing all decisions to reinforcement learning or all failures to a scalar loss.