Praxis Labs · A Research Laboratory for Overlooked Medicine
Founded 2026
Praxis Labs uses computing systems to read the whole published record on common diseases, to find the treatments it already points toward but no one has yet tested, and to test the most promising of them quickly and inexpensively in established contract laboratories.
Fig. 1
Each problem we take up passes through the same four stages.
Schematic · simulated values, not data
The reading ends in a single point, which is the finding.
Our systems study every relevant paper, trial and public dataset on a problem, noting where the findings agree, where they conflict and where they point toward something untried.
Each candidate is weighed by the strength of its evidence, by the cost and speed of testing it, and by the number of people it could help.
The strongest candidates go to the bench in contract laboratories, often as a cell-based assay that costs thousands of dollars rather than millions and returns its answer in weeks rather than years.
Every result is published, including those that fail, together with a plain account of what was learned.
The four stages are captioned beneath the screens. Buttons I to IV jump to each act, and the pause button holds the current frame.
Fig. 2
A problem becomes a candidate for our attention only when it meets all four of these conditions. Each row applies one condition to the problems that survive the row above it.
Condition 1, A common condition
It affects a great many people.
Condition 2, An inexpensive test
A simple experiment can give a clear answer.
Condition 3, An untaken step
Published findings point toward it, yet no study has examined it directly.
Condition 4, An open field
No large, well-funded laboratory is already pursuing it.
Candidates for our attention
Only the problems that pass every row remain.
Fig. 3
Every study we undertake will be listed here with its technical report and a plain account written for general readers, whether the result is favorable or not. The first reports will appear once our initial experiments are complete.
Index of reports · no report has yet been issued
Numbers are reserved in sequence
Form of entry
Specimen · not a report
PRX-001 1
PRX-002
ReservedPRX-003
ReservedPRX-004
ReservedPRX-005
ReservedPRX-006
ReservedPRX-007
ReservedPRX-008
ReservedPRX-009
ReservedPlain language
Each technical report will be paired with a plain account of what was tested, what was found and how confident we are, written for patients, funders and the interested public.
Fig. 4
Praxis Labs was founded in 2026 and is at the very beginning of its work. The pink point marks where the work stands today.
Origin
Founded 2026
01 · Now
Surveying the literature and selecting the first problems.
02 · Next
Commissioning the first laboratory experiments.
03 · Then
Publishing the results, whatever they show.
We would be glad to hear from scientists, physicians and laboratories who share an interest in this approach, and from those who may wish to support the work.
Correspondence address
A correspondence address will be published here shortly.
Fig. 5
The mark of Praxis Labs, a letter P composed of seven lines of text, drawn out across the page.
The Approach in Full
Praxis Labs is a new research laboratory with a single purpose, which is to find promising treatments for common diseases that the published record already points toward but that no one has yet put to the test. We take our name from the Greek word praxis, meaning the putting of knowledge into practice, because we believe the distance between what has been learned and what has been tried is wider in medicine than in almost any other field.
Our method joins two kinds of work that have rarely been done by the same hands. The first is reading, carried out by computing systems capable of studying many thousands of papers, trials and datasets on a single question and of reasoning carefully about what they contain. The second is experiment, carried out in established contract laboratories, where the most promising of those ideas can be examined quickly, inexpensively and honestly.
Every year the biomedical literature grows by more than a million articles, and no physician, scientist or institution can hope to keep pace with it. Good ideas are published, cited a few times and then set aside, not because they were disproved but because nobody was in a position to connect them with the other findings that would have made their importance plain.
It has recently become possible for machines to read that literature in its entirety and to reason about it with genuine care. In mathematics, such systems have already settled problems that waited years for attention, and there is every reason to expect that biology and medicine contain discoveries of the same kind, lying within reach and waiting only for someone to look in the right place and then to run the experiment.
Reading. We gather every relevant paper, clinical trial and public dataset bearing on a chosen problem, rather than a convenient sample of them, and our systems study the whole of it, noting where findings agree, where they conflict, and where they point toward something that has not been tried.
Ranking. Each candidate idea is weighed by the strength of the evidence behind it, by the cost and speed with which it could be tested, and by the number of people it might help if it proved correct, so that our limited resources go first to the questions most likely to repay them.
Testing. The strongest candidates are taken to the bench, using contract laboratories and the simplest experiment that can give a clear answer, which is often a cell-based assay that costs thousands of dollars rather than millions and returns its result in weeks rather than years.
Reporting. Every result is written up and published, including the ones that fail, and each technical report is accompanied by an account in plain language, so that those who fund, follow or depend upon the work can understand exactly what was learned.
We are deliberate about where we look. A problem becomes a candidate for our attention only when it satisfies four conditions, each of which exists to keep our work useful, affordable and honest.
It must affect a great many people, because common conditions have the largest literatures to study and the greatest benefit to offer when something works.
It must be testable by an inexpensive experiment that gives a clear answer, so that a promising idea can be confirmed or set aside without years of delay.
The evidence must point toward a step that no one has yet taken, which is to say that several published findings suggest the idea while no study has examined it directly.
It must not lie directly in the path of a large and well-funded laboratory, since in that case they are likely to arrive first and our effort is better spent elsewhere.
We hold that research undertaken in the hope of relieving illness ought to be understandable by the people it is meant to serve.
For that reason each technical report we publish will be paired with a plain account of what was tested, what was found and how confident we are in the finding, written for patients, funders and the interested public rather than for specialists alone.
Praxis Labs was founded in 2026 and is at the very beginning of its work. We are now surveying the literature and selecting our first problems, and we expect to commission our first laboratory experiments once those choices are made. We make no claims about results we do not yet have, and when we do have results, whether favorable or not, we will publish them here.
Surveying the literature and selecting the first problems.
Commissioning the first laboratory experiments.
Publishing the results, whatever they show.
We would be glad to hear from scientists, physicians and research laboratories who share an interest in this approach, and from foundations and individuals who may wish to support the work. A correspondence address will be published here shortly.