experiments
api unreachable
The research API did not answer, so experiments cannot be listed. Nothing is shown in its place — an empty list is correct here, invented rows are not. What experiments would look like is below.
503 Service Unavailable
what this page shows
- —each question run against its own dataset, hashed so the run can be repeated exactly
- —every row that could not be built is counted under a named reason, and that list is usually larger than the dataset
- —the critic's verdict can only be stricter than the deterministic one, never lighter
hypothesis
#12 — volume-burst-pool-2h
unit of analysis
token-measurement
rows
226 built · 1,904 excluded
excluded, by reason
no_reading_at_horizon 1,502 · window_too_sparse 402
dataset hash
9f2c…41ab
critic
NEEDS_MORE_DATA — sample_size, independence
Every row that could not be built is counted under a named reason. Silent filtering is what makes “found nothing” indistinguishable from “looked at nothing”, and the exclusion list is usually larger than the dataset.