FIELD STUDIES · PUBLIC METHOD
A live source can start the question. It cannot answer it alone.
Three releases each week join current evidence to baseline data, mechanism research, counterevidence and a disclosed decision model.
THE RESEARCH CHAIN
Twelve gates before publication.
- 01
Fix one current subject
One unused primary record is scored for evidence depth, novelty, decision consequence and modelling potential. Generic word matches cannot widen it.
- 02
Search five live lanes
Current development, official market data, mechanism research, contrary evidence and implementation evidence remain distinct.
- 03
Recover a missing lane safely
A blocked, thin or non-readable result is discarded and search continues. Every lane beyond the retained trigger still requires a separately retrieved and verified outside source.
- 04
Retrieve and fingerprint
Every cited external page is retrieved through a public-address, size-bounded HTTPS path before use and receives a content fingerprint.
- 05
Test the evidence packet
Five lanes, distinct identities, source families, a current primary trigger and a current external market date must exist before drafting begins. A source family is the disclosed host or official collection shown in the register; it is a diversity check, not proof of independence.
- 06
Normalise time, place and units
Monthly, quarterly and event data retain their native reference periods. A first-verification date is never presented as a publication date.
- 07
Build the claim ledger
Every material statement—including the closing interpretation—points to a source or a disclosed calculation and receives one evidence-state label.
- 08
Construct a bounded scenario
Canonical inputs, formulae, assumptions and outputs remain inspectable. The result is an educational comparison, not a forecast.
- 09
Rewrite without changing evidence
A failed treatment may be revised three times against one frozen source digest; evidence is never swapped merely to rescue prose.
- 10
Run an adversarial audit
A separate pass looks for unsupported causality, wrong dates, incompatible geographies, source mismatches, invented scenes and missing counterevidence.
- 11
Stage one recoverable release
Exact record links, evidence routes, calculations and media are digest-checked inside one durable publication transaction. Committed immutable files are checked again during recovery and health monitoring.
- 12
Publish or retain
A failed study does not become filler. The last valid edition remains public while the existing observatory windows retry later.
EVIDENCE STATES
These labels describe the route by which a statement entered the study. They are not confidence theatre and are never collapsed into one opaque score.
SCHEDULE AND RECOVERY
Monday · Wednesday · Friday.
The existing observatory service checks at 09:15, 13:15, 17:15 and 21:15 Europe/London. The first passing release becomes permanent; later runs remain recovery windows. No tenth production timer was created.