Content Migration — Biggest Risk¶
Source: data-model call 2026-06-29 (raw/call-2026-06-29-data-model.md).
Content migration is the single biggest risk and sequencing challenge for the 14-week build. The CMS is the centerpiece — the app is essentially a styled content engine — so nothing downstream (curation, journey-building, graph engine) can start until a content subset is in the CMS.
The risk¶
- No raw content received yet. Format unknown — could be PDF, JSON, spreadsheet rows. Can't design the importer until we see it.
- AI pipeline planned to process the corpus: extract passages, generate AI interpretations, apply tags.
- Script-based approach preferred over pure AI — to avoid hallucinated quotes. Lewis's exact words are sacred; extraction must be faithful, not generative.
- Content entry is explicitly excluded from Fueled's statement of work. Fueled builds the pipeline + CMS; the client/editorial side owns getting content in.
Update — IA stand-up 2026-06-30¶
Client raw content was collected on 2026-06-30, and the team confirmed it won't map cleanly to Lina's CMS schema as-is. Plan: manually shape a few sample pieces into the model first, to validate the content model before committing to a bulk importer. Reinforces the "start on a subset" approach below.
Implications¶
- Start on a content subset (a manageable set of letters/passages) to unblock CMS seeding and everything after it.
- The importer is bespoke to whatever format arrives — keep it isolated/replaceable.
- Faithful extraction (script) + AI tagging/interpretation (build-time, editor-QA'd) are separate stages; don't let AI touch the quote text itself.
Related¶
04-mvp-scope.md— Phase 1 CMS priorities + sequencing reality.05-feature-backlog.md— tag taxonomy is the upstream content-design blocker.00-constraints.md#3 — AI tagging is build-time only.