What these 146 are
The model ranks 2.6 million untouched crossings; the full live ranking, subfield by subfield, is on the Radar. From it we took the top ten of each view in each of the four domains (4 × 4 × 10 = 160 places), after removing by hand, one by one, those that were not real methods or made no sense. Some forecasts sit in more than one view at once, so the places hold 146 different forecasts, sealed with a date that cannot be changed. They are not all the forecasts: they are the strongest and best-checked ones, and the ones we will be judged on.
Pick a forecast and try it
Every card opens the full sheet: what the method is, where it attaches to the field, the papers to read first and a plan in four steps. Being first on a crossing is a paper nobody else has written.
A bet that cannot be edited
Each line has a SHA-256 fingerprint and the rule is written below. In 2029 anyone can count how many came true, including the ones we got wrong.
Some we study ourselves
We are turning some of them into studies, with the papers on the Research page. If you start one, tell us and we will link it.
The rule, written before
A forecast “method M will arrive in subfield S” comes true if at least 3 works whose primary subfield in OpenAlex is S, published between January 1, 2026 and December 31, 2028, use M in their title or abstract. Each forecast below had zero such works up to 2025 and none in the part of 2026 already indexed. Every line is fingerprinted with SHA-256 and a secret salt, so its date can be proven without anyone being able to change it.