Brand Brain

The Brand Record contract. One canonical, AI-readable brand record per brand, every engine reads it or fills it.

Spec v1, Draft

Brand Brain is the Record, not a bot. It is the data contract every engine (blog, identity kit, content, money, Sales Concierge) reads from, so brand truth lives in exactly one place instead of leaking into a dozen private copies.

The rule that makes it work

Engines never own brand truth. They read the Record, and write outputs + learnings back. Any tool storing its own private copy of voice, palette, or positioning is a bug.

The 13 sections

Every brand gets one Record. Three storage tiers per section: required for any brand (required), standard for active brands (standard), deep for content-engine brands (deep).

#	Section	Tier	What lives here
A	Identity & Vision	required	Brand name, one-line truth, category owned, mission, 12-month goal
B	Customer	required	Who buys, what they want, what they’re moving away from, objections
C	Voice	required	5 descriptor words + 3 banned words, do/don’t rules, pacing, voice routing
D	Offers & Ladder	required	Every offer: name, price, promise, who it’s for, next logical offer
E	Point of View	standard	Philosophy + Hot Takes, tier calibration (Additive / Spirited / Combative)
F	Evidence	standard	Stories (first person, tagged by lesson) + Proof (numbers, testimonials, credentials)
G	Frameworks & IP	standard	Named methods, each with steps, the problem it solves, where it’s taught
H	Topics	standard	4-8 verified content categories, drives blog taxonomy and AEO targeting
I	Visual System	standard	Palette, type, surface modes, style dials, component rules, imagery world
J	Cast	deep	Founder bio (verified facts only), team, mascots/personas, masthead identity
K	Boundaries	required	Compliance non-negotiables: income disclaimer, AI disclosure, banned claims
L	Back Catalog	deep	Index of everything published, Repurpose Score, Refresh Status
M	Snippets	deep	Drop-in components: bios, boilerplates, CTA blocks, framework explainers

Every entry carries the same trust layer: Confidence (HIGH/MEDIUM/LOW), Authenticity (LIVED/BUILDING/CURATED, prevents AI faking expertise), Source ref, Last verified, Record version.

Where it lives

Canonical home: Notion. Humans edit here, AI reads/writes via MCP. Sections A-K as structured entries, L-M as child databases.

Portable exports, regenerated on every material change:

brand-brief.md

≤10 KB. Any chat AI, custom GPTs, Claude projects, the universal carrier.

brand.json

Full record. Engines: blog pipeline, article writer, identity generator, agents.

tokens.css

Small. Web surfaces: blog templates, portals, landing pages, kits.

image-prompts.md

Small. Image generation: mood boards, thumbnails, article art.

The pro-tier move is a hosted export at a stable URL per brand (cdn/brands/{slug}/brand.json), version-stamped so a stale consumer is detectable. That’s what makes “thousands of articles per month” operationally sane.

The reader contract

A reader declares which sections it needs. Missing required section means the engine recruits it (the Phase 0 pattern) instead of generating generic filler. That single rule is the anti-slop mechanism at scale.

Engine	Must read
Publication / blog engine	A, B, C, E, H, K
Identity engine (kits, mood boards, sites)	A, B, I
Content engines (social, newsletter, decks)	A, C, D, E
Money engines (money-finder, win-back, follow-ups)	D, B, C
Client studio pipeline	A-K progressively, builds the record as it runs

The writer rules

Every write carries Source ref + Confidence + Authenticity. No anonymous truth.
Intakes (extraction, interview, pipeline, ingestion) may create and update sections.
Engines write outputs to their own surfaces, and learnings back to the Record.
Conflicts resolve by recency + confidence, with a human gate on section A, C, and K changes, identity, voice, and boundaries never drift silently.
90-day refresh cycle: feedback (visibility reports, analytics, reactions) triggers a review pass. Stale sections get flagged, never silently trusted.

Migration plan

Approve the spec

Adjust sections/tiers to taste. The field list is the whole decision.

Build the Record for Design Hacker first

Merge the DH 15-layer output, the DH Brand Library, and the BWA DH pack into one Record. Surfaces every mapping problem on a brand you can judge by eye.

Retrofit the blog engine

The bigger game and the hungriest reader. Article templates read brand.json + brand-brief.md, nothing else.

Point the intakes at the Record

Brain Builder v1.3 writes Record sections instead of its own 3-DB shape. Brand interview skill writes A-D. BWA export writes I.

Author engagements inherit everything

New author = Record + book ingestion into F/G + publication engine reading it.

Why it matters

Industrial blogging without slop: every article grounded in E/F/G with authenticity tiers, so the moat sentence stays true at volume. New engines get cheap to build, a reader of a known shape is a weekend, not a quarter. And the Record itself becomes sellable: “your complete brand brain, portable, yours.”