Arasense
The receipts point here.
Every number published on this site links back to a raw pipeline output. These are those files — unedited JSON, exactly as the Arasense engine wrote them. Download them, interrogate them, recompute them.
The full 34-model Bologna projection: every model's trust weight, both rejections, the spread, and the agreement behind the +14.2% headline.
The Aras Diagram coordinates of the Bologna trust screening: every model's bias, variability, and correlation error, its weight and tier — including one borderline model flagged because it drifted above the rejection floor between runs.
SSP2-4.5 vs SSP5-8.5 for Bologna, sharing one observed baseline and one set of trust weights — only the emissions scenario changes.
The five-city Italian portfolio ranking: Rome, Milan, Florence, Bologna, Venice — with the agreement column that changes the story.
The European Trust Atlas screening runs behind the ranked map — one method, one configuration, every city.
The global preview runs, including the full Mexico City arc: the implausible number caught at launch, and the honest signal published after the extreme-plausibility screen was built because of it.
The full-ensemble Mexico City investigation run: 31 of 34 models trusted on mean climate, several carrying physically impossible daily extremes — the evidence that mean-climate skill does not certify extreme-value sanity.
The re-verification sweep of all 36 published atlas numbers under the extreme-plausibility screen: 35 matched within 0.02 points, the largest deviation 0.1 points.
The Screening Drift Ledger itself — the append-only record behind the Drift Ledger page: every re-run of a published screening, what moved, and what was done about it.