EXPERIMENTS IN TRUST
Source-aware
confirmation.
Agreement can come from shared error. Explore when a second source is worth its cost—and when confidence is misleading.
FROZEN SOURCESChecking
CHECKPOINTS VERIFIED—
MODELS IN MEMORY—
EXPERIMENT TYPEExploratory
01 / DESIGN A TRIAL
What should we trust?
02 / INSPECT THE EVIDENCE
Awaiting a trial
Run an experiment to inspect its purchased observations and final predictions. No demonstration results are prefilled.
Final belief by group
These are final predictions and confidence, after all observations. They are not intermediate scores.
Observation trail
Inherited reports carry no new charge. Every purchased observation consumes the displayed resource budget.
| Step | Source / charge | Group | Observed label | Predictive probability | Resources left | P(bias = 0) |
|---|
Exact receipt and source binding
03 / PERSISTENT RECORD
Recent local trials
Reading saved receipts…
04 / THE FROZEN BENCHMARK
Results with their limits.
Checking for the completed evaluation.