Project Hermes Just Got Auditable: Satellite Line-of-Sight, Spatio-Temporal Corroboration, and Reproducible Hashes

For most of the history of UAP reporting, the question "could it have been a satellite?" was answered with a count. Was there a satellite roughly overhead? Yes? Then maybe. That's not a test — that's availability bias with a database behind it. Project Hermes just replaced it with something more honest.

SAT-LOS-01: From "something was up there" to geometric proof

The new SAT-LOS-01 rule propagates TLEs (Two-Line Element sets, the standard orbital parameters cached daily from CelesTrak) through the sgp4 propagator locally, and asks a strict two-part question: was a catalogued satellite geometrically consistent with the witness's reported bearing and elevation at the reported time, and was that satellite sunlit while the observer was in darkness? Only when both conditions are true does Hermes eliminate the case as a satellite.

This matters because a sunlit-observer or an eclipsed-satellite produces no visible flare, no reflected trail. The old SAT-OVER-01 rule was essentially a presence check. SAT-LOS-01 is a visibility check — and visibility is what a witness actually saw, or didn't.

CORR-01: Corroboration without confirmation

The second new rule, CORR-01, scans the native case store for independent reports within 30 minutes and 50 km of the subject case. But here is the discipline that separates Hermes from sighting databases that inflate counts into claims: CORR-01 never eliminates a case, and it never increases anomaly confidence. It is explicitly framed as evidence of reporting behavior, not of anomaly. Two people reporting the same thing at the same time in the same place tells us something real, but it does not promote a sighting from "unexplained" to "true." Hermes refuses to make that leap, and it refuses on the record.

Deterministic audit hashes

The quietest change is, in some ways, the most important. Every audit response now carries an audit_hash keyed on methodology version plus per-rule hashes, excluding wall-clock timestamps. Two evaluations of the same inputs against the same methodology version yield the same 16-character identifier. Every time.

This is what reproducibility looks like. A paper that cites a Hermes case can now include the hash, and a reviewer five years later — running the same methodology version — will get the same verdict or know exactly where the discrepancy is. That is a foreign concept in UAP research, where the same case has historically been described three different ways by three different investigators and nobody could say who was right.

Why it matters for the study of UAP

The field's credibility problem has never been a lack of reports. It has been a lack of checkable reports. A case that cannot be re-analyzed by a third party against a stated method is a story, not data. SAT-LOS-01 replaces a vibe with geometry. CORR-01 acknowledges coincidence without exaggerating it. And the audit hash makes the whole thing auditable by outsiders — which is the only property that can eventually move UAP analysis from "intriguing" to "citable."

The new rules are live now. You can examine any case's rule-by-rule audit trail at projecthermes.tech, or pull the JSON directly from /api/audit/<case_id>.

Project Hermes and UFO Index are affiliated projects.