Acquisition
Social channels and pain-led entry materials
Distribution eventsIntelligence for the journey, not AI for decoration.
Compass combines an explicit context model, constrained AI production, behavioral evidence and a versioned control plane. The architecture keeps the learning loop useful, inspectable and reversible.
Five layers, one evidence spine.
Each layer can evolve without hiding what produced a public result.
EVIDENCE SPINEIDs · versions · receipts · costs · decisions
Context
Articles, semantic routes and awareness state
Reading-path transitionsIntelligence
Logitronics, Awareness Scale and model routing
Proposed next-best contentConversion
Context-aware CTA and qualified action
Lead or decision receiptLearning
Analytics, experiment policy and rollback
Verified system changeTwo methods shape how understanding develops.
These are operating lenses, not mystical scoring systems. They help decide what question the reader can usefully answer next.
Logitronics
Maps how concepts, distinctions and decisions depend on one another so content can resolve the right uncertainty in the right order.
- Concept dependencies
- Contradiction and gap detection
- Decision-oriented sequence
Awareness Scale
Models the reader's movement from felt problem to informed choice, changing depth, framing and CTA along the way.
- Problem recognition
- Context accumulation
- Readiness-sensitive action
Architecture claims stay bounded by evidence.
Production and evidence contracts
Versioned definitions exist for content, visual, publishing, distribution and recovery states.
Azure-backed AI execution
Existing producer runtimes use Azure model infrastructure; this product site is isolated from corporate ARCHYX runtime.
Complete content-to-lead path
The current milestone connects the working parts into a single physically verified journey.
Policy-bounded optimization
Autonomous low-risk experiments begin only after a baseline journey and rollback receipt exist.
Governance is part of the product.
01No paid or public action outside an approved boundary
02No model is called live without a runtime, route and canary
03No production state advances without a physical receipt
04High-impact claims and offers require owner approval
05Every experiment preserves the previous version and rollback
Choose a real commercial context for the first verified journey.
A pilot starts with one audience, one complex offer, one knowledge base and one qualified action.