What this stage does
Translate the study concept into an executable protocol: schedule of assessments, inclusion/exclusion, and operational detail.
Where AI applies
Draft generation from templates, consistency checking across sections, and forecasting operational burden per visit.
Under the Hood · 4
4 capability-angle deep-dives generalizing this essay's pattern for AI and software engineers.
GovernanceValidationHypothesis
The Citation-Gated Drafting Loop
The system can be wrong, but it can never be unaccountable.
EvaluationGovernanceHypothesis
Your LLM Doesn't Need a Judge. It Needs a Linter.
Verdicts gate. Opinions advise.
GovernanceHypothesis
Designing the "No"
Boundaries drawn on consequence, not on capability.
Data EngineeringData CatalogsHypothesis
When Documents Become Data
The highest-leverage AI investment is usually not a better model. It's a schema.