Data Bank
Everything the desk has collected, and everything it has computed from it. 0 rows of primary record. The two layers are kept apart on purpose: a thin collection makes everything below it decoration, while a stale derived layer is a job that has not run — a different problem with a different fix.
Collected
What came in from outside. Every row here was written by a job, and every job is named — because “0 rows” is only useful next to what to run about it.
Synthesised
What the desk built out of the archive. All of it is derived, so nothing here is lost when it goes stale — it is recomputed by the job named on its row.
Per city
every active city is completeA total hides the thing that decides whether a city is tradeable at all. A city with no forward forecast has no price to disagree with; a city observed under twelve times a day has an understated daily maximum, and every feature built on it inherits that.
Is any of it true
The frozen record, asked the only two questions that matter: does a band priced at 30% settle 30% of the time, and did a claimed edge ever show up in money.
Export proprietary record
Download a bounded, integrity-verifiable copy of the desk's accumulated research and source evidence. This adds an exit path for your data without exposing the private research tables to the browser.
Immutable research captures, corrections, quality flags, and integrity manifests.
Read-only export. Source rows are never changed. Each file ends with its row count and SHA-256 manifest. Exports are limited to 31 days and 50,000 rows so the app does not create an unbounded server job.