01
The plan is to subscribe to publicly available newsletters through dedicated research inboxes, as a reader would. Today the inbound path stores the raw message for provenance and runs a light markup strip before extraction — HTML tags (including image-based tracking pixels) are removed from the text the model sees, but that is not automated PII redaction and the raw MIME is retained. We do not buy lists or resell anyone’s mail.
Inboxesdedicated to research
Raw storagefull MIME retained today
Extraction parsemarkup stripped; not PII redaction
PII handlingplanned, not yet automated
Tracking pixelsimg tags dropped in parse only
02
A real card will link to the rendered issue and the verbatim line behind its claim. If a line cannot be quoted from a source, it will not ship.
Verbatim or dropno quote, no claim
Confidence bandsHigh · Medium · Low (planned)
Model versionpinned per extraction
Source linkon every real signal
03
The planned brief will focus on sponsor frequency, category heat, and cadence shifts. Any single-title archive will need attribution and a citation trail, without subscriber data or PII in what we publish. We do not plan to sell personal information.
Aggregate feedsplanned for subscribers
Single-titlecited · attributed · auditable (planned)
PII / CCPApolicy: no consumer data sold
Brand controlssee below