01 · Collect
Ingest the evidence base.
Trials, research text, molecular databases, de-identified records, and insurance claims — continuously.
Sectors / Medical
Medicine knows more than any clinician can hold — and it contradicts itself.
Trials, research text, molecular data, and insurance claims that quietly disagree — and evidence skewed toward populations that don't represent everyone. The answer a clinician needs is buried in the noise.
Research
Medicine generates more knowledge than any human can hold — trials, molecular data, clinical text, and hundreds of millions of insurance claims — and much of it quietly contradicts itself. Retractions, underpowered studies, and publication bias leave a literature that is vast, authoritative-sounding, and unreliable.
So we collect the data, the research text, and the claims, and ask whether what we believe is actually true. Is a disease chemical, environmental, or genetic? Does a result hold across ancestries, or only in the group that was studied? Racial and demographic diversity isn't a footnote — a treatment proven in one population can fail in another. Answering that is a knowledge-engineering problem.
Questions Worth a Clean Answer
Ask hard. Answer with clean data.
- Q01A landmark Alzheimer's paper was built on fabricated data and shaped a decade of drug development. How many other foundational results have never been independently verified?
- Q02Most clinical trials enroll predominantly white, male populations. For a drug approved on that basis, what is the actual evidence it works in the patient sitting in front of you?
- Q03Insurance claims capture real-world outcomes for hundreds of millions of people. When claims data contradicts a curated trial, which is telling the truth about how a treatment performs?
- Q04Two meta-analyses of the same intervention reach opposite conclusions depending on which studies they include. What would a reconciled, weighted analysis — across all available evidence — actually show?
The Method — A Continual Loop
Collect, refine, hypothesize, test — repeat.
02 · Refine
Keep what replicates.
Weak power, contradiction, and retracted findings stripped out. What survives agrees with itself.
03 · Hypothesize
Surface the implied links.
The core proposes drug–gene–phenotype links that no single study reported but all of them imply.
04 · Test
Check against fresh reality.
Predictions validated on new trials, held-out cohorts, and real-world outcomes.
05 · Refine
Grow the medical core.
Results fold back in. Every cycle it gets more predictive and less credulous. Continual.
The Cascade
Drug Repurposing Cascades.
How compounds discovered for one indication cascade through their real mechanisms into unexpected therapeutic uses — and reshape entire markets. A map of serendipity turned into pipeline strategy for 2026.
Compound / insight
Mechanism
Repurposed use
Industry shift
Select any node to trace its chain. Left to right: Compound / insight → Mechanism → Repurposed use → Industry shift.
What the Core Delivers
Knowledge you can act on.
- A private research core for a lab or hospital — its own data, refined against the world's.
- Real-world signal from insurance claims, not just curated trials.
- Evidence checked for who it actually studied, so findings hold across populations — not just one.