Decision rule
Continue only when grounding errors are absent from publishable packets and reviewers find the research agendas useful enough to justify the added burden.
Future research ยท unvalidated
Earn the next claim in the right order.
Human review and workflow usefulness come before broader generation or investment-performance analysis.
Prioritized sequence
Five evidence gates
Each stage must be frozen prospectively and evaluated before the next begins.
- 1. Blinded human reviewRate factual support, novelty accuracy, numeric interpretation, mechanism quality, falsifiability, usefulness, and instrument fit without future outcomes.
- 2. Analyst workflow pilotCompare disclosure-only, generic summary, grounded hypothesis, grounded hypothesis plus skeptic, analyst-edited, and human-created research agendas.
- 3. Prospective hypothesis freezingTimestamp hypotheses, evidence, confirmation conditions, falsification conditions, and permitted market data before outcomes occur.
- 4. Broader case evaluationOnly after the human-review gate, consider a separately authorized bounded expansion. The existing 60-case phase remains blocked.
- 5. Investment-performance analysisOnly after prospective freezing should a distinct study evaluate information coefficient, Sharpe ratio, hit rate, signal decay, or realized-versus-implied volatility.
Stop rule
Stop or redesign if hypotheses remain generic, require excessive edits, obscure uncertainty, or produce weak safe-rejection behavior.
Claim rule
Do not convert technical test coverage, reviewer preference, or a workflow metric into an investment-performance or commercial-validity claim.