Spellbook OS
Concept Alignment — 2026-05-21 — Post desk-flip fresh start
One-Line Concept
Turn one recording into a week of content
A content OS for solo creators. Ingest a pillar asset, extract the best moments, generate channel-native derivatives, and review them before anything leaves.
v1 Core Loop — Prism
Ingest
→
Extract
→
Generate
→
Review ✓
confirmed v1 = Prism only
confirmed Review is mandatory
confirmed Keep "Prism" name
Explicit Non-Goals (v1)
Foresight (planning)
Ritual (publishing)
Portals (distribution)
Scry (analytics)
Sanctum (dashboard)
Auth / Billing / DB
Multi-tenancy
Mobile
Value Wedge
Not "AI generates content for you."
Instead: "Your existing content produces 10x more output with your taste and judgment preserved."
- Source-to-output lineage — every derivative traces back to a moment
- PAC-aware adaptation — adapted to platform + culture, not just reformatted
- Mandatory review gate — creator stays in control
Target User
Solo creator-operator
Produces long-form content regularly. Uses 3-6 disconnected tools. Willing to pay when one tool replaces the stack. Buyer = user (same person).
hypothesis Needs ICP validation against narrower Prism-only scope
Riskiest Unknowns
- 1 Extraction quality — Can we reliably identify the best moments from a transcript?
- 2 Derivative acceptance rate — What % do creators approve vs. heavily edit?
- 3 Input format scope — Video, audio, AND text? Or transcript-first only?
- 4 BYOK fragmentation — Does model choice create inconsistent quality?
Lessons from v1 (Desk-Flip)
lesson
Prototype first. No DB, auth, billing, or infra until a user can click through the core loop.
lesson
One module. Ship Prism end-to-end before touching any other module.
lesson
No infra dependencies. If setup is blocked, build features that don't need it.
lesson
Specs ≠ progress. Cap spec work at the minimum to start coding.
lesson
Real tests. Tests must prove the product works, not that stubs compile.
lesson
No premature abstraction. Write the concrete thing first.
Next Step
Run ICP research against the narrower v1 scope
Primary: /pack install business-discovery then /icp
Existing ICP research in archive/research/icp.md is a valid starting point but needs re-evaluation for Prism-only scope.