01
Structure the evidence
Research records are normalized by intervention, outcome, population, study design, dose, duration, source, and evidence grade—not stored as undifferentiated text.
StackTerminal evidence intelligence
From fragmented biomedical evidence to decisions people can inspect.
StackTerminal is developing evidence and decision-support infrastructure for digital-health and preventive-health providers. It connects structured research with personal context and changing real-world signals while preserving the sources, rules, and uncertainty behind every output.
314
structured interventions
1,018
evidence records
897
source-linked records
314
review-dated interventions
System architecture
A closed evidence-to-decision loop
The defensible system is the loop and its structured data—not a general-purpose language model or a list of supplements.
02
Resolve the context
Goals, constraints, body metrics, labs, and wearable-derived patterns establish where evidence may or may not transfer to an individual context.
03
Generate a traceable decision
The decision layer ranks candidates, scales doses within curated bounds, checks interactions, and returns citations, alternatives, and explicit uncertainty.
04
Re-evaluate over time
New wearable or lab signals are compared with the context captured at generation. Material drift can trigger a proposed revision instead of silently changing an output.
Operational today
What is already implemented
- ✓Structured evidence records with source links, populations, outcomes, doses, and grades
- ✓Deterministic dose bounds and interaction/risk checks around AI-assisted reasoning
- ✓Context fusion across profiles, uploaded lab reports, and supported wearable providers
- ✓Evidence-linked recommendations with alternatives and uncertainty
- ✓Longitudinal drift detection and reviewable stack revisions
- ✓Claim extraction and evidence comparison for social-health content
Validation roadmap
What must be proven next
Evidence fidelity
Measure extraction accuracy, citation validity, and grading agreement against expert-curated reference sets.
Decision quality
Benchmark relevance, safety-rule recall, calibration, and abstention across predefined user scenarios.
Longitudinal value
Test whether signal-driven re-evaluation improves decision stability and user outcomes in prospective pilots.
Clinical boundary
Define intended use, risk classification, quality management, and regulatory pathway before any clinical claims.
Supplements are the proving ground, not the ceiling.
The consumer application proves the evidence-to-decision workflow. Our infrastructure direction is to help digital-health and preventive-health providers build transparent, governed experiences—while requiring every new domain to earn its own evidence coverage, safety controls, validation, and regulatory assessment.
Current product: educational decision support. Not a medical device and not medical advice.