Before a test window opens we publish the hypothesis, a hash of the frozen configuration, the universe, the horizon, the minimum sample and the exact pass/fail thresholds. When the window closes the result is published unconditionally — a failure appears here with the same prominence as a success.
The reason this matters: running forty variants and publishing the three that worked is indistinguishable, after the fact, from having been right. A hypothesis fixed and hashed before any outcome exists cannot be selected afterwards. A window that closes without a result is shown as OVERDUE here and fails our build — silence is not an available option.
No window has been opened yet.
The register is empty — not filtered, not pending review. Opening a window is a public commitment, so it is made deliberately rather than generated. When the first one opens, this page will show its hypothesis, config hash, universe, horizon, sample floor and thresholds from that moment on, and its result the moment the window closes.