As of mid-April 2026 there were three live campaign engines each with their own handbook, gist corpus, cadence, and voice: STRATT’s PHM product-marketing campaign, the Personal Dev / Escape Velocity brand narrative, and the devarno-cloud Org Content deck spanning all nine products. Each had a separate gist tree. Shared products showed up in multiple campaigns but the overlap was invisible. Timing conflicts — two campaigns targeting the same platform in the same week — had to be caught manually. Audience reuse was a Slack-thread problem.
Promoting “the set of all campaigns, their intersections, and the signals a sync produces” into a first-class data layer — not a spreadsheet, not a doc index, an actual relational schema with deterministic cross-reference detection — changes what’s possible. Shared products become edges. Timing conflicts become alerts. Audience overlap becomes a surfaceable signal. The same handbook content drives both the publishing pipeline (Programs → Trails → Posts) and the strategic view (the manuscript), because the latter sits above the former and decorates it rather than replacing it.The load-bearing decisions that made the layer cheap to build were: (1) Notion + Airtable stay the source of truth, so editors never leave tools they already know; (2) structural diff is deterministic and free to re-run, so LLM can be additive rather than critical-path; (3) STUDIO is a stateless proxy, so the write surface never has to reconcile with HUBBLE’s DB.
A campaign strategist with four live campaigns used to spend Monday mornings asking “what overlaps this week?” and getting a subjective answer. Once THE MANUSCRIPT is seeded, the answer is a URL with three visual modes and a ranked list of marketing signals. The strategist spends Monday acting on the overlap rather than finding it.The deeper leverage: every new campaign adds its handbook to the manuscript, and the cross-reference graph grows denser automatically. Network effects on internal content. By the fifth campaign, the xref graph is discovering reuse opportunities a human wouldn’t spot — “this post from PHM shares three products and two platforms with the new Personal Dev launch; run it as a spinoff trail”. The system starts suggesting trails rather than just organizing them.This is the same idea that made the STRATT schema-first approach compound (atlas/learnings/2026-03-28-schema-first-foundation-before-cli.md): treat the canonical data shape as infrastructure, and the tooling around it becomes easier to build, not harder.
