Study publication (003-v1)

Publication of the upload-frequency study: report, methodology, binned aggregates, and a de-identified row dataset with a self-contained replication notebook.

Headline (pre-registered, read once on a sealed 20% holdout): posting more does not lift each video (a small −10% per doubling, too small to act on). Total reach scales roughly one-for-one with output, and that result survives trimming the biggest videos. All correlational; direction unestablished.

Reproduce from dataset/replicate.qmd.

Pre-registration (003-v1)

The locked, publicly-timestamped analysis plan for the upload-frequency study, tagged before any frequency-vs-views number was queried. This is the audit trail: the design spec and pre-registration as of this tag are what the sealed-holdout headline was read against.

See design-spec.md and preregistration.md at this tag. The published results live in the 003-v1-published release.