Observe
Read primary sources as they publish, and record what was available when — so a forecast can later be checked against the data that existed at the time, not the data we have now.
6986Research
The method and the record.
The One Lab's commercial products use a research discipline that keeps commitments, outcomes and limitations checkable. This page shows what the public record can evidence now.
Current record
FORECASTS LOCKED174
every lock ever written, including re-forecasts of the same target
GRADED0
first verdicts due late Oct 2026
MISSES PUBLISHEDINSUFFICIENT DATA
every miss stays public, permanently
MISS RATENOT PUBLISHED
rates publish per series and horizon, not per arena
The rate is not published at arena level. Rates publish per series and horizon, where each realized month is counted once.
Arenas
Two arenas issuing. The rest say so.
AR-01 · LOCKED · AWAITING FIRST GRADE
Housing / United States
150 locked · 0 gradedAR-02 · LOCKED UNDER HOLD · PROVISIONALHousing / Canada
24 locked · 0 graded · provisionalAR-03 · IN DEVELOPMENTQuant
No forecasts issued — this arena is not runningAR-04 · IN DEVELOPMENTFinancial Stability
No forecasts issued — this arena is not runningAR-05 · IN DEVELOPMENTGeospatial
No forecasts issued — this arena is not runningAR-06 · IN DEVELOPMENTMine Design
No forecasts issued — this arena is not runningThe method
Eight stages. Three of them have a record behind them.
3 of 8 stages carry a count from the record.
Reason
Work out what can honestly be forecast and what cannot, and write down the refusals with their reasons. A series the models decline is a finding, not a gap.
20Hypothesize
State what might be true before testing it, so the claim can be attacked. Hypotheses are typed as unverified and are never rendered as fact.
No artifact on the public recordExperiment
Design and run the trial that could disprove the hypothesis, logging every run — including the discards — with a manifest sufficient to reproduce it.
Not yet exercised — no arena has run this stageBuild
Turn what survives into something usable — a model, a tool, a decision aid — and put it under the same grading discipline as everything else.
Not yet exercised — no arena has run this stageForecast
Commit to a number before the outcome exists, in a table that cannot be edited or backdated. This is what makes the rest of the loop checkable by someone who does not trust us.
174Verify
Grade the commitment against the primary source when the outcome arrives, by arithmetic on records that already exist, and publish the verdict whichever way it goes.
No verdicts published yetLearn
Change what the evidence says to change — adopted on graded outcomes, never on a backtest, and never by the system editing its own judge.
Waits on the first verdictForecasting is the most mature public capability here, not the purpose. Experiment, Build and Learn remain visibly bounded by what the record can substantiate.