Policy Comparison

Cross-jurisdiction comparison is exactly the kind of wide, shallow sweep that agents do well and humans do resentfully. The risk is that the output becomes a confident matrix nobody can check. CueCrux keeps the matrix honest: every cell, every "jurisdiction A requires X, jurisdiction B does not", carries a receipt to the specific provision or guidance it rests on, so a reader disputes the source, not the summary.

Token budgets make the economics work. Every retrieval call carries a hard budget and the daemon trims to fit; on the ScoreCrux benchmark, budgeted retrieval returns far fewer tokens than naive top-K with near-identical answer quality. A twelve-jurisdiction sweep is a routine job rather than a quarterly indulgence.

For the regulatory text itself, WikiCrux is live as a receipted knowledge hub: answers cite their sources and the citations open. Comparison work built on cited answers inherits their provenance.

Multinational compliance teams use the receipted matrices to justify where a single global policy suffices and where local variation is legally forced, and the justification survives because it is checkable.