Pipeline
From raw revision history to a source-linked trajectory.
NextConsensus estimates the current state and future transitions of consequential claims across institutions. Four stages, each one auditable: extract the trajectory, classify what kind of change it was, decide how it should be reviewed, and deliver it with the evidence attached. Forecasting is a research program built on top of it. The full method is on its own page.
The pipeline
Extract, classify, prioritize, deliver
The classify step carries the weight. Without it, a burst of editorial housekeeping looks identical to a claim genuinely gaining ground, and a system that cannot tell those apart will confidently rank the wrong things. Where actors disagree, the record shows the disagreement before it summarises it.
Extract
Refract reconstructs claim trajectories from public revision history. Every wording change, citation, certainty shift, and propagation event is dated and verifiable.
Classify
Each change is labelled for what it actually was. A new trial citation attaching to a benefit claim is evidence movement. That same wording turning up in a guideline entry is authority language spreading. A reformatted table or a renumbered reference is neither. Without the distinction, housekeeping reads as momentum.
Prioritize
Direction and strength of movement are the features under test: gaining stronger support, becoming harder to dismiss, spreading across influential sources, losing qualifiers, appearing in regulator or guideline language. They are the features for ranking review attention.
Deliver
The current program returns the trajectory with its source trail, ready for review. A useful lead-time result remains part of the research program.
Interactive Dissection
Breaking a broad question into units the record can settle
“Will payers cover our drug?” is where most teams start. The pipeline breaks a question like that into four conditions before it looks at any evidence: a named authority, a defined action, a target population and a cutoff date. We call a question in that form resolvable: the named body either takes the action by the date or it does not. Try one.
Deterministic Claim Decomposition
Transforming ambiguous commercial and evidentiary questions into resolvable, time-bounded propositions with named authorities and explicit verification criteria.
“Will payers lift restrictions on GLP-1s for diabetic kidney disease?”
“When will bispecifics move to second-line DLBCL standard of care?”
“Will European health authorities approve and reimburse complement therapies for GA?”
Where it sits
The twelve-rung ladder
A claim climbs from first publication through replication and expert consensus to the moment an institution acts in public. Rungs 1–4 are the claim moving through the literature: that is what a claim history records. Rungs 5–7 are a guideline body, a regulator or a payer acting: that is what a forecast is registered against. Each rung on the ladder below names what is observable at that step.
Formal recommendation change by a recognized guideline body: grade shift, new/retracted endorsement, population expansion/restriction.
The ladder guide gives each rung its signals and its role; the method gives the rules a forecast is registered and scored under.
Provenance
Every event is traceable to a public revision
Every event in a claim history has a dated public source and a reproducible hash, so a trajectory can be recomputed from the public record and checked.
The system maintains the same claim across paraphrase, wording changes, broadening, narrowing, qualification, relocation, removal, reintroduction, and citation replacement.
A public source may change because of new evidence, delayed editor attention, imported regulator language, citation cleanup, editorial judgment, or routine maintenance. The system labels each — it does not automatically claim "evidence strengthened."
Every trajectory is byte-for-byte reproducible from the public record. Registration hashes let anyone verify that the trajectory has not been altered.
Research program
What the research program is testing
The forecasting layer is a hypothesis under prospective test. None of the four questions below is settled yet. An earlier proof cycle appeared to show a relationship between public revision activity and later authority actions; the result was invalid, because the positive cases had full revision histories and the negatives did not. With the negatives correctly ingested, no signal remained in general activity. If there is one, it is more likely in specific claims and specific patterns of movement. That is the next experiment.
Scores appear only after a registered forecast resolves. The protocol is tracked in the Verification Ledger.
Whether trajectory features improve on base-rate forecasts.
Whether the system surfaces movement earlier than the workflows teams already run.
Whether results from one therapeutic area, authority, or event class transfer to another.