the line's destination is the acceptance statement: an updateful endpoint with integration and one matched advance, and updatelessness as the ideal extension

Agent-decided, reversible (the LI/BRIA specification round, third pass, on the maintainer's stated position). The target a candidate is checked against is projects/decision-theory/rounds/2026-09-29-li-bria-synthesis-spec/PROBLEM_STATEMENT.md §0: the supporting milestones (computable existence with named classes; logical induction on the active trajectory; BRIA recovery on a declared recovery domain with witness estimates that need not be public; coverage as accountability with menu safety named), the integration obligation (market information provably improves decision performance under explicit accounting, the compatibility relation stated at least in its established one-sided form, and a formal result that the integrated process excludes a precisely specified failure mode of the belief–decision coupling), the recovery and integration rows of the suite, and one advance on a recognized obstruction by a diagnostic meeting the round's qualification standard — the belief-disagreement troll is one proposed, provisional instance, and one theorem may discharge both the failure-mode result and the advance. Logical induction is required on the active trajectory; passive recovery is a weaker supporting milestone. The endpoint may be updateful; updatelessness is an ideal extension whose recovery domain is an open specification question (queued in Awaiting the author). Architecture is not specified; a modular implementation qualifies; the continuation-BRIA block contract is one route. A candidate with the milestones and the performance result alone is an integration theorem, not the endpoint. Rejected alternative: a single unified criterion from which every benchmark verdict follows, found vacuous (satisfied by a logical inductor and a BRIA side by side) or false (unit-level coverage of the full class with cross-unit commitment, P1′); requiring updatelessness for success; and treating belief–decision compatibility as the whole of integration, which the third pass separates into three achievements.