The marketing operating system that turns user requests into deliverables. Three components, one flow. Built so the orchestrator can compose without re-deriving anything from scratch.
Aleph isn't a single thing. It's a knowledge base, a catalog, and a router — each with a job, in a specific order. The router governs the process. The catalog tells the router what to pull. The knowledge base holds the actual marketing intelligence.
Each component is a distinct artifact. Delete one and the others stop being useful. They're decoupled but not independent.
The actual marketing intelligence. Chapters, frameworks, matrices, decision rules across fifteen disciplines. This is where the substance lives.
Lives at /revised/{discipline}/ with templates at /revised/_templates/.
The lookup table. Maps user requests to chapter sets, matrices, and templates. Names what to pull when. Without it, every request would re-derive its composition from scratch.
Lives at /revised/DELIVERABLE-CATALOG.md.
The governor. Reads the user's request, detects the archetype, consults the catalog, pulls chapters, runs matrices, fills templates, emits the deliverable with a decision trace.
Lives at /revised/GNESIS-MARKETING-ROUTER.md.
User asks: "Build a brand thesis for our B2B SaaS company." Here's what happens, step by step. Most steps are silent; the user just sees the final deliverable.
Identifies trigger phrase "brand thesis" and archetype signal "B2B SaaS". No archetype override from the user; the router uses business-model signature to confirm.
B2B SaaS confirmed. Router pins master-lifecycle-b2b-saas.xlsx as the lifecycle reference, even though this isn't a lifecycle deliverable — it informs voice and persona defaults downstream.
Looks up Deliverable 1. Catalog returns: required brand context (audience, category, founder POV, anti-positioning), primary chapters, matrices to run, and the template variant to fill (greenfield, repositioning, or sub-brand).
Router picks brand-thesis-worksheet-greenfield.docx based on archetype + brand-context cues (no existing thesis on file). If unclear, it would ask the user.
Checks if 70%+ of required context is present. If not, pauses and asks the user for the missing fields in one batch. Otherwise proceeds and tags any inferred fields with their synthetic-anchor source.
Router pulls primary chapters: brand-strategy/01-foundation-and-brand-thesis.md, framework chapters, and the templates pack reference. Loads them into composition context.
Two matrices fire: brand-tier classification (challenger vs. category-creator) and cultural posture (apolitical / values-aligned / cultural-leader / cultural-disruptor). Outputs feed the thesis.
Router fills the template using brand context for variable substitutions, framework outputs for structural decisions, and matrix outputs for posture / tier choices.
Self-check: any unfilled variables? Any synthetic-brand placeholders that should be brand-specific? Tier flag attached: this template is currently Tier-B; output carries a "spot-check" annotation.
User receives: filled brand thesis (.docx), decision trace (which matrices ran, why), chapter trace (sources cited), open questions (where context was thin), and template-tier flag.
The router can't operate without the catalog; the catalog can be consulted manually without the router. That makes the catalog the authoritative artifact. Every decision about routing logic is downstream of how the catalog is structured.
Every router provision has an escape hatch and a visible flag when it fires. A broken catalog reference surfaces in the output, not just in logs. A thin brand-context check pauses and asks rather than guessing. The router fails legibly so the failures get fixed.
The router doesn't try to be smart. It follows a fixed sequence: detect archetype → consult catalog → pull chapters → run matrices → compose → flag. Smart routers fail in surprising ways. Disciplined ones fail in legible ways.
Eight lifecycle archetypes (B2B SaaS, D2C subscription, D2C one-time, services, fintech, ecommerce, marketplace, media) act as scaffolds. The closest archetype anchors voice, persona, and operational defaults; divergence from the archetype is surfaced explicitly.
Every template is scored on seven dimensions (end-to-end coverage, multi-channel per row, operational rigor, persona branching, AI tagging, compliance integration, build-out layers). Outputs inherit a tier label so the user knows what they're looking at.
Most deliverables come in variants (greenfield / repositioning / sub-brand for brand thesis; eight archetypes for welcome sequences; B2B / consumer for voice axes). The router picks based on archetype + explicit user signal + brand-context cues. When unclear, it asks.
Aleph is built in scaffold form with a runnable Python skeleton. The catalog has migrated to YAML (machine-readable). Five end-to-end simulations pass. Platform implementation (the runtime that wires the skeleton to LLM composition) is the remaining piece.
| Component | Sub-piece | Status |
|---|---|---|
| Knowledge Base | 16 disciplines, 513 chapters total | Done |
| Knowledge Base | 8 lifecycle archetype masters (Tier-A) | Done |
| Knowledge Base | Pass 1-3 quality lifts (paid-media, brand-strategy, positioning) | Done — 88%, 77%, Pass-3 bespoke |
| Knowledge Base | Reference-depth chapter index (REFERENCE-INDEX.yaml) | Done — 91 cataloged + 422 reference |
| Knowledge Base | Integrity audit + structural sweep | Done — 0 missing files, 0 thin chapters |
| Catalog | 147 deliverables across 31 categories | Done — up from 56 / 11 at start |
| Catalog | Depth audit: 0 thin · 49 medium · 98 thick | Done — 67% thick |
| Catalog | YAML structured form (DELIVERABLE-CATALOG.yaml) | Done |
| Catalog | 35 compliance-aware deliverables with explicit regime mapping | Done — GDPR, FINRA, HIPAA, etc. itemized |
| Catalog | 5 artifact composites + 10 engagement composites | Done |
| Catalog | Engagement-readiness pre-flight (P0/P1/P2) on 13 composites | Done |
| Catalog | 30 category template packs (Pass-4 quality) | Done |
| Catalog | 34 bespoke Pass-5 templates spanning brand-foundation, campaigns, email, social, launch, GTM, measurement, ABM, content, customer-marketing, retention, martech, AR/PR, partner, pricing, research, and engagement composites | Done |
| Catalog | 8 sample synthetic outputs across all 9 anchors + Aleph-on-Aleph self-demo | Done |
| Catalog | Avg trigger phrases: 8.1 per deliverable | Done |
| Router | Specification (§1-§23 of GNESIS-MARKETING-ROUTER) | Done |
| Router | Python implementation (aleph_v2.py) | Done |
| Router | JavaScript port (in runtime UI) | Done |
| Router | 19 patches applied (A through V) | Done |
| Router | Regression suite — 12 waves, 106+ scenarios | Done — 95%+ pass |
| Router | End-to-end demo: fintech EU regulated launch | Done |
| UI | Dependency graph (D3 force-directed) — live | Done — aleph-deps-gnesis.netlify.app |
| UI | Aleph-on-Aleph self-demo — live | Done — aleph-selfdemo-gnesis.netlify.app |
| Catalog | Synthetic anchor profiles — 9 anchors with full audience/channel/KPI depth | Done |
| UI | Catalog explorer with live router preview + 7 filter chips + cross-link to runtime | Done — aleph-catalog-explorer-gnesis.netlify.app |
| UI | Runtime UI (chat-style) with brand-context library, persistent state, 2-deep composite expansion, engagement-readiness pre-flight | Done — aleph-runtime-gnesis.netlify.app |
| Ops | Maintenance + audit loop (depth + integrity + regression + drift) | Done — GREEN |
| Roadmap | Filling remaining 49 medium-depth deliverables to thick | Pending |
| Roadmap | Real Aleph runtime as a server-side product (multi-user, persistent storage, SSO) | Pending |