The B.L.U.E. System - MVP

Overview

The B.L.U.E. System is a decentralized learning architecture that transforms distributed user-generated content into structured knowledge.

This MVP is built around three core principles:

• Data is owned at the edge (PDS)
• Structure is derived, not stored centrally (Indexer)
• Presentation is independent of data ownership (App layer)

MVP Goal

This MVP exists to prove two core ideas:

• Content can be transformed into structured understanding
• Structured understanding can be derived across independent PDS nodes without central control

Achitecture Diagrams

Core MVP Architecture

Architecture at Scale

System Components

1. Personal Data Server (PDS)

Each PDS:

• stores only user-owned data
• uses immutable ATProto-style records
• emits append-only record streams
• contains no global state or awareness of other nodes

Stores:

• blue.article
• blue.method
• blue.vote

2. Indexer

A read-only computation layer that aggregates distributed PDS data.

Responsibilities:

• consume records from multiple PDS nodes
• group content by subjectArea and difficultyLevel
• compute derived learning hierarchies
• link methods to articles
• aggregate votes
• validate structural consistency

MVP Constraint

• Only one Indexer instance exists
• It is not authoritative
• It can be replaced or recomputed at any time

3. App / UI Layer

The App layer:

• queries Indexer outputs
• displays structured learning views
• renders subject => level => article navigation
• shows methods and vote signals

It contains no structural logic.

Data Model

The system defines three core record types:

• blue.article - procedural learning guide (how-to format)
• blue.method - alternative explanation of an article
• blue.vote - independent signal for aggregation

All records are:

• immutable
• CID-addressed
• stored in PDS nodes
• globally referenceable via URI

Key System Properties

Decentralised

• User-owned data (PDS)
• Content creation
• Identity (DID-based)

Centralised (MVP only)

• Indexer instance
• Derived hierarchy computation

Stateless Layers

• App layer
• Indexer (regenerable from PDS streams)

What This System Demonstrates

The MVP validates that:

• structured learning systems do not require central ownership
• hierarchical knowledge can be derived from distributed data
• multiple independent data sources can form a unified learning graph
• ATProto-style record streams are sufficient for educational structure systems

What is NOT in MVP

• multi-indexer federation
• relay / firehose infrastructure
• trust or reputation systems
• blockchain anchoring
• cross-PDS storage syncing

These are defined in /docs as future extensions.

Documentation

Full system design is split into:

• /docs/architecture.md
• /docs/data-model.md
• /docs/indexing.md
• /docs/trust-and-federation.md
• /docs/vision.md

Design Philosophy

B.L.U.E. is built on a separation of concerns:

• PDS - owns data
• Indexer - interprets structure
• App - presents structure

And critically:

No single layer controls data, meaning, and presentation simultaneously.

MVP Success Criteria

The system is valid if:

• Articles from multiple PDS nodes appear in a unified view
• Content is correctly grouped by subject and difficulty level
• Methods correctly attach across PDS boundaries
• PDS nodes function independently of Indexer
• The hierarchy can be fully recomputed from raw PDS streams

Summary

B.L.U.E. is not a traditional learning platform. It is an experiment in whether structured knowledge can emerge from distributed, user-owned data without central authority.