The Standard Model simulates how a patient’s disease unfolds over time — so it seemed fitting to let a simulation explain it.
In this 23-minute audio deep dive, two AI-generated hosts walk through how Standard Model Biomedicine is building The Standard Model™ V-2: a JEPA world model of the patient that predicts biological state across molecular, tissue, organ, and patient scales rather than predicting the next word in a medical record.
What’s covered
0:00 — Why the patient isn’t a document: world models vs. language models
6:12 — Turning messy clinical data into a continuous patient timeline
11:52 — Protege’s double-blind evaluation: competitive performance at 36–170× lower cost per prediction
15:21 — Why a small model that runs inside the hospital firewall matters
16:47 — Patient matching for a Phase 3 trial: 95.3% criteria-level accuracy
18:35 — TrialFront™: simulating clinical trials before the first patient enrolls
21:37 — Digital twins and the future of the control arm
Go deeper
The Standard Model™ V-2 → — the world model behind the conversation
TrialFront™ → — AI co-statistician for clinical trial design
About this recording: both voices are AI-generated, and the conversation was produced with AI from Standard Model materials. Every figure discussed comes from our published work and independent evaluation.



