A method here, a problem there, separated by a corridor, a journal, a vocabulary. History shows that many breakthroughs happen when the two finally meet. Singularity finds where they will meet next — so that it happens years sooner, on purpose instead of by chance.
Each tower is a field; each rising dot, a paper. Each arc is a real crossing from history. Hover a tower to see the ideas that left it or reached it.
A century ago a physicist could read most of physics. Today a researcher cannot keep up with a single sub-field. Knowledge piles up in towers that rarely talk to each other, each with its own journals, conferences and words. The method that would crack a problem in medicine may already be standard in astrophysics — and no one in the hospital has ever heard of it.
This is not a lack of intelligence. It is a lack of a map.
Look at the history of science and the same story returns: a method matures in one field, then crosses into another and changes it.
How long the idea waited between the field where it was born and the field it changed. The waits are getting shorter. Our goal is to bring them close to zero.
Singularity turns that pattern into a forecast. It reads the scientific record, learns how every method in it travelled from field to field, and names the crossings most likely to open next: this method, in that field, by that date.
Replayed on fifteen years of history across all of science, its top forecasts on crossings nobody had touched came true more than two hundred times more often than chance, and more than ninety in a hundred got at least a first work. It does not replace scientists. It tells them where to look, before anyone else does.
“At least three works in experimental and cognitive psychology will use consensus clustering by the end of 2028.”
T3v3-2026-0038 · SHA-256 · 2026-11-05
Today this site is a window: it shows where progress is about to happen. A forecast says “take this method there” — it does not say how. That is the next part of the work, and we have already started it.
Dated, fingerprinted forecasts on all of science — 252 subfields — with a track record replayed on fifteen years and confirmed twice with rules sealed in advance.
We took our own forecasts and turned them into studies, from a black-hole formula to a test for cognitive subtypes and a platform trial design for eczema.
For the strongest forecasts, a first concrete experiment: the question to ask, the data to use, the test that would settle it — carried out by us and by anyone who wants to try.
The registered forecasts are judged by the public record, hits and misses alike, and every result goes back into the model.
Forecasting where progress will happen is the first half. The second is making it happen sooner — and learning from every result.
The whole scholarly record, every year, every field.
The crossings most likely to open, with a deadline and a strength measured on history.
Turn the strongest forecasts into first experiments — ourselves, with partner labs, and increasingly with automated research tools.
Every success and every failure goes back into the model. Each turn of the loop makes the next forecast better.
Today we close the loop by hand, one study at a time. The goal is to close it for every field of science, every year.
Six frontiers where the next decade will be decided — and where the missing piece may already exist somewhere else.
Batteries, superconductors and catalysts designed on a computer with methods borrowed from quantum physics and AI.
Diagnoses that know when not to trust themselves; drugs found by searching chemistry the way search engines search text.
Readable equations for how cells and molecules work, instead of black boxes.
Old questions about shapes and numbers attacked with tools that were born in physics, economics and computing.
Cause and effect in the climate system, measured with methods first built for brains and markets.
Gravitational waves, black holes and the early cosmos read with the newest interpretable AI — as our own black-hole formula already showed.
See the crossings the model sees coming, frontier by frontier →
The rules we will not bend, because the value of a forecast is only as good as the trust in it.
Singularity is a starting point, not the last word. This is how we want to work, and how we would like you to work with us.
Every forecast, the rule that judges it and how we test the model are public, free for anyone to read, use and check. We look for funding to keep going and to grow; the forecasts stay open to everyone.
A forecast is a probability, not a promise: about half of the strongest come true, and about one in five of the hardest. Our studies are first steps, often simulations, and others may have reached the same idea elsewhere. A crossing that is likely does not happen by itself: someone has to do the work. Every study is signed by Singularity: we take responsibility for it, and if you find a mistake, tell us and we will correct it in the open.
You do not have to join anything. If you think a forecast is wrong, have a better idea or a different way to test it, hold the data that would settle it, want to try one in your lab, or simply prefer another direction: write to us. Every contribution, large or small, critical or not, makes the map better. Progress is something we make together.
Researchers who want to be first in their field. Labs that can run the experiment. Funders who want their money where the next breakthrough will be. Students choosing what to learn. The map is public; the discoveries are still to be made.