RETURN TO OBSERVATORY →
PROTOCOL VISIBILITY · PRECOMMITTED RULES
PROTOCOL PUBLISHED

THE GRAVITY PROTOCOL

ONE PAGE. EVERY ACTIVE RULE THE INSTRUMENT RUNS ON. EACH VERSION PUBLISHED BEFORE THE SCANS IT GOVERNS.

CURRENT VERSION: GRAVITY WEEKLY SCAN RUBRIC v2.1 · PRIOR VERSIONS RETAINED IN THE ARCHIVE

READ THE PROTOCOL ↓ RETURN TO OBSERVATORY →
PUBLIC SOURCES · PRECOMMITTED CLASSIFICATIONS · CORRECTION POLICY DISCLOSED · HISTORICAL GAPS LABELED
THE RULES
01WHAT GETS SCANNED

WHAT GETS SCANNED

Each week, one domain from the receiver rotation. Public sources only (Gallup, Stanford HAI, ITU, Edelman, World Bank, and named equivalents). Source-linked reporting begins with Signal 06. Signals 01–05 are preserved as early historical records with disclosed gaps; missing source links, forecasts, and current-rubric classifications are labeled NOT RECORDED rather than reconstructed.

02THE THREE CHECKS

THE THREE CHECKS EVERY SCAN ANSWERS ALL THREE

  1. 1ASSIGNMENT — Was completion assigned/promised in this domain's sources within the predeclared source window? (Products, programs, narratives promising arrival through more next.)
  2. 2OUTPUT SHAPE — Did the domain’s output remain continuation-shaped—more steps, upgrades, revisions, or new gaps—or did the public record contain credible evidence of stable, durable completion-side output?
  3. 3FALLOUT — Did fallout appear? (Churn, displacement, burnout, continuation without arrival.)
03THE RESULTS

THE RESULTS

Precommitted — these rules were published before the scans.

04CONTRARY DATA

CONTRARY DATA

Counter-evidence is logged in every report, by name, in its own labeled section — including when it does not cross the MISS threshold. Sub-threshold movement accumulates visibly in the record.

05WHAT THE INDICATORS MEASURE

WHAT THE INDICATORS MEASURE SCOPE, STATED PLAINLY

The indicators — engagement, loneliness, trust, adoption, and equivalents — are proxies, declared as proxies. The scan reads the misassignment and its fallout: the observable layer. It does not measure "arrival" itself — by the detection's own terms, arrival is off-axis and is not measured by anything. The scan tracks the fallout. It does not prove the detection.

06THE AUDIT TRAIL

THE AUDIT TRAIL

From Signal 06 forward, each rubric-governed scan record carries: the forecast logged before scoring, the full source list with release dates and links, the classification, the timestamp, the rubric version, and its production method, disclosed. Signals 01–05 predate the current reporting standard and are preserved with disclosed gaps rather than retroactively reconstructed. When automated production is used, the record additionally carries reproducibility controls: the model and version, fixed generation settings, and SHA-256 hashes of the exact prompt and source snapshot. Manual-era records are explicitly labeled as manual. The archive follows a public correction-disclosure policy: corrections are labeled in the affected record. Prior versions are retained where available; the public site does not claim cryptographic immutability.

07FORECASTS

FORECASTS

The weekly forecast number is computed from the archive itself — empirical transition probabilities and recurrence rates — and published with its basis (e.g., "estimate basis: N-scan record") and confidence tier. The instrument displays what it computes.

08ROLLING 12-SCAN REVIEW

ROLLING 12-SCAN REVIEW

After every 12 scored Signals, Gravity conducts a rolling review of the instrument record. The review does not retrospectively rewrite previous reports or rescore historical Signals.

The review examines: forecast calibration; result distribution; receiver coverage; source quality; compliance with the contrary-data requirement; and missing audit fields or record gaps.

The allowed review outcomes are NO CHANGE, CALIBRATION NOTE, or RUBRIC VERSION CHANGE.

If the review produces a rubric change, that change applies prospectively only. The published change must record the new rubric version, the first Signal it governs, the effective date, and the rationale. Earlier reports remain governed by the rubric version under which they were originally scored.