Written down first. Checked after.
Forecasts are logged before the outcome is known, and findings can be run again by anyone.
Most forecasts are judged by the people who made them.
A model that is never scored in public can claim anything. Most are tuned after the fact, and the misses fade from view.
We write every forecast to a ledger before anyone knows the answer, score it against what happened, and sign our findings so others can run them again from the same data.
Forecasts and findings that hold up when someone checks.
- Forecast ledgerForecasts logged before anyone knows the answer, and scored in public after.
- Signed findingsFindings signed so anyone can replay them from the same data and the same code.
- Large cataloguesLaplace searches ESA's Gaia DR3 at full scale and puts what it finds on a star map.
- Open codeLibraries and notebooks published as open source, runnable in the browser.
- QuantumMethods evaluated on quantum hardware, as in 26 experiments delivered to the University of Osaka.
How we work.
- Register
- We write down the question and the method before we see the answer.
- Run
- The same data, the same code, a signed result.
- Publish
- Method, code and score in the open, for anyone to check.
Figures.
- 1,376
- Forecasts logged before the outcome was known
- 1.8 bn
- Stars searched in ESA's Gaia DR3
- 220,656
- Candidate companions found
