# Summary: #brand-pipeline-spec

## Overview
This channel documents the specification for Curious Endeavor's **Brand Pipeline**—a human-in-the-loop production system that compresses brand creation from weeks of work to 6 structured decision moments (~2 hours total). The system is not autonomous; it amplifies human taste by handling execution between 6 strategic touchpoints where creative judgment genuinely matters.

## Key Decisions & Outcomes
- **Product scope:** Human-in-the-loop system (not autonomous AI brand generator). AI handles execution; humans own taste at 6 designed gates.
- **8-tool pipeline defined:** Brief Enricher → Research Engine → Strategy Maker → Territory Finder → Style Forge → Asset Generator → Deck Builder → Page + Social
- **Time target:** Under 2 hours human time; 3-4 days elapsed time per brand
- **Taste as infrastructure:** Taste board is a reference signal database (tagged by category, timing, CE resonance), not a mood board. Tool 04 queries it to generate 3 visual directions.
- **Quality layer:** Vision model runs QA criteria before images reach human selection (T5)—removes technical failures, preserves taste gate.
- **Phase 4 (productization) is conditional:** Only viable after 10+ runs with calibration data showing Anton's scores correlate with Assaf's at meaningful rates.
- **Data model:** All work flows through one evolving Brand Brief object with folders for intake, enrichment, positioning, visual direction, style guide, asset library, and decision logs.

## Active/Open Items
- **PHASE 0 in progress** (design-only, no code):
  - [ ] Design visual-style.json schema (urgent—blocks everything downstream)
  - [ ] Design taste board tagging schema (controlled vocabulary)
  - [ ] Forensic map of Phat project (signal → decision → inference chain)
  - [ ] Define what makes a "good brief" vs. merely complete
- **Open decisions:** Which semi-cold brief for Sprint 1 test (Assaf owner); Phase 4 pricing model (later); who operates this besides Assaf (later)
- **Gate:** No code starts until all Phase 0 artifacts exist

## Key Links & References
- **Core reference:** Phat project (used as real example to validate visual-style.json schema)
- **Decision corpus location:** `brand-brief/decisions/` folder (logs all human choices + rationale for ML calibration)
- **Build plan phases:** Phase 0 (design, ~1 week) → Sprint 1 (2-3 weeks) → Sprint 2 (3 weeks) → Sprint 3 (2 weeks) → Phase 4 (conditional productization)