A world model for science.
We forecast clinical trials before they read out, and keep score in public.
Beat the odds on trials, and you know what to test next.
About nine in ten drugs that enter human trials never reach patients. A model that calls results in advance knows something real about biology, and can pick the experiment that would teach it the most.
26,682 forecasts sealed. None can be edited.
Each forecast gets a public fingerprint the day it is made, starting Sep 29, 2026. When the trial reads out, anyone can check it.
Trials reading out soon.
Randomized phase 2 and 3 trials from drug companies, finishing in the next four months. Eligible ones are sealed before the result.
Every miss becomes a question.
When a forecast misses, the model is wrong somewhere. That becomes an unknown, with the experiment that would settle it.
- Predicting what works
- Forecast every trial before it reads out, and keep score in public.
- Reading the literature
- Every claim, number and method from every paper, trial and dataset, read and linked.
- Knowing what's known
- One shared model per disease, every number pooled from all studies, with honest uncertainty.
- Picking the next experiment
- Ranked by how much each experiment would teach, per dollar.
Science forgets its misses. A world model keeps them.
Most drugs that reach human trials fail. Failed results often go unpublished, so the same mistakes get made again.
The model missed on pioglitazone. The miss is the lead.
A liver-fat model built from the literature called the result far too low. The gap points to a mechanism nobody has measured.