Each dependency clock carries a permanently different burden of proof. Workflow possession is private but instrumentable: if On Signal were integrated with Okta's CRM, you could observe whether AEs pull discovery questions before calls or copy competitor-card language into deal notes. That evidence exists in principle. Nobody has built the pipe to capture it.
Event-clock synchronization is the only type partly falsifiable from outside. The Hysteresis flagged the DFARS FOCI comment deadline closing July 6, 2026. Whether it surfaced that gate before Coreshell's team found it independently is unknowable, but whether it surfaced it at all is checkable against the Federal Register.
Ambient authority has the lowest ceiling. GitHub shows 100,000+ stars on browser-use; NYC Health data confirms Docket's sourcing on heat mortality. Both signals confirm quality, but neither gets anywhere near dependency. No public signal can show whether a TinyFish prospect remembered Current during vendor selection or a Shade Tree donor forwarded Docket to a board member. The absence of dependency evidence here is structural, not a research gap waiting to be closed. Anchor should build accordingly: do not mistake content elegance for proof that anyone relies on it.
What's testable by type
Workflow possession (On Signal, Context Window): CRM pulls, call-prep logs, SE handoff citations, deal-review references. Theoretically observable, but nothing is currently instrumented. Denial test: content stays a static library no one opens before a live account moment.
Event-clock sync (Hysteresis, Clear Path): Did the publication surface a regulatory gate, trial readout, or policy deadline before the customer needed it? Testable against public records — ClinicalTrials.gov, Federal Register, conference abstracts. Denial test: the publication misses a gate the customer had to find alone.
Ambient authority (Current, Docket): Source accuracy and domain fit are verifiable. Donor forwarding, prospect citation, and buyer recall are not. No denial test exists using public data alone.