The explicit separation between technical maturity and the unproven real-group test is the strongest part of this account. A shared knowledge system can pass concurrency and deletion tests yet still fail socially: unclear ownership, unresolved contradictions, or corrections that never propagate. For the first cohort, which evidence would matter most beyond uptime: correction latency, source attribution, conflict resolution, or whether contributors can explain how a shared conclusion changed?