Explore: search the science and map a field

Put in a topic and a date. Get the labs working on it and what their papers claim, using only what was published by that date. Then test your own picks against what happened, and read what the full evaluation produces.

Search the science, as of a date

Every query word must appear in a paper's title or abstract. Labs are ordered by how many of their papers match, then by text match. Nothing is predicted and nothing is scored: this finds and describes.

The example topics are the fixed queries of the search evaluation, in its own order. None was chosen for how its results look.

Map a field: who had public evidence on a topic, as of a date

Labs, institutions and companies with public evidence on a target or topic by your date, and how the field filled up year by year. Other public names for the same target are searched too. Or start from the list of targets.

Topic: CAR-T, as of 2021-12-31. 239 labs at 36 institutions had a last-author paper matching the topic by that date. This is a count of public evidence. It is not a prediction, and it does not rank people.

Institutions, by number of labs with matching papers

Institution	Labs	Matching papers
University of Pennsylvania	41	84
Harvard University	30	41
University of Washington	17	36
Cornell University	14	27
University of North Carolina at Chapel Hill	13	19
University of California, San Francisco	10	15
University of California San Diego	10	12
Stanford University	9	15
The Ohio State University	8	11
Emory University	7	7
Case Western Reserve University	6	9
University of California, Los Angeles	6	9

Three immunology targets, chosen by ruleFrom a fixed list of 12 immunology targets, the three with the most labs publishing on them by 2021-12-31.

IL-6

567 labs at 40 institutions had a matching last-author paper by that date.

Institution	Labs	Matching papers
Harvard University	51	57
University of California San Diego	26	32
University of Colorado Anschutz	26	32
University of California, San Francisco	24	26
Johns Hopkins University	22	28

IL-2

148 labs at 35 institutions had a matching last-author paper by that date.

Institution	Labs	Matching papers
Harvard University	10	11
University of Pennsylvania	9	9
University of Pittsburgh	9	9
Emory University	8	8
University of Washington	8	8

IL-17

126 labs at 35 institutions had a matching last-author paper by that date.

Institution	Labs	Matching papers
University of Pittsburgh	8	17
Emory University	8	11
University of California, San Francisco	8	9
University of California San Diego	6	8
Harvard University	6	6

Pick your own, then see what happened

It is 2022-12-31. Each set shows 10 young health companies with their names hidden and everything public about them that day. Pick the ones you think raise a seed round or larger within 24 months. Then the page shows what happened, with the model's picks and the blind AI's picks beside yours.

The sets are easier than real life and the page says so: 3 of every 10 came true here, against 7% of 1,097 companies in the full test. Over many sets drawn the same way, the model's top picks find 1.5 on average, the blind AI's 1.6, and random picks 0.9. There are 12 sets, all from one test: new company to seed round. Other stages have no deck yet. The model and the blind AI also saw one line you do not: whether an officer's name matched a lab head. A name match is not shown anywhere on this site, so it is left off the cards.

Evidence briefs, built ahead of time

A brief is what the full evaluation produces for a topic or a lab: the search above finds the papers, a sealed reader checks each extracted claim against the abstracts and must return two exact quotes, and code checks the quotes and sorts. It needs reader agents, so this page cannot run it. These 6 were built by the real pipeline on the dates shown.

Across them: 100 of 103 claims were verified against their own paper. The reader found 0 contradictions and 0 cases where a second, independent lab supports a claim. A brief is a checked, quoted listing of the evidence.

Explore: search the science and map a field

Search the science, as of a date

Map a field: who had public evidence on a topic, as of a date

Pick your own, then see what happened

Evidence briefs, built ahead of time

antibody-drug conjugate

in vivo CAR-T

KRAS inhibitor

targeted protein degradation

lab of David R. Liu

lab of Jay Shendure