Access granted

    Graph Engineering
    for a working Second Brain.

    Build the relationship layer that helps an AI brain recall, act, learn, and stay trustworthy.

    Start here

    Build one small graph before building a platform.

    Choose one real work question. Use the templates below. Keep every relationship inspectable.

    1. 01Choose one question that needs relationships, history, or commitments.
    2. 02List the entities and sources involved.
    3. 03Define only the relationships you need to answer the question.
    4. 04Add provenance, timestamps, and confidence.
    5. 05Test the answer, then record what the graph got wrong.
    The operating loop
    01Work
    02Recall context
    03Capture feedback
    04Consolidate and evaluate
    The one-page model

    A graph is memory with connections.

    A useful operating graph keeps five things together: what exists, how things relate, where the claim came from, when it was true, and how confident the system should be.

    graph-design-canvas.txt
    QUESTION
What should the brain be able to answer?

ENTITIES
Person · Project · Decision · Task · Source · Event

RELATIONSHIPS
supports · contradicts · supersedes · part_of · prerequisite

EVIDENCE
Source path, author, timestamp, excerpt, confidence

OPERATING LOOP
Recall → act → feedback → consolidate → maintain → evaluate

STOP CONDITION
If the source, relationship, or confidence is unclear, flag it for review.

    Start with the questions your brain needs to answer. Then define the smallest set of entities and relationships that makes those answers possible.

    Entity

    A person, project, decision, task, source, or event.

    Relationship

    A typed connection with a clear meaning.

    Evidence

    The source that supports the claim.

    Time

    When the claim was true and when it was recorded.

    Confidence

    How certain the system is about the connection.

    Owner

    Who can review or change the information.

    graph-lite.sql
    create table graph_entities (
  id text primary key,
  type text not null,
  name text not null,
  source_ref text,
  observed_at timestamptz,
  confidence numeric check (confidence between 0 and 1)
);

create table graph_edges (
  from_id text references graph_entities(id),
  relation text not null,
  to_id text references graph_entities(id),
  evidence_ref text,
  observed_at timestamptz,
  confidence numeric check (confidence between 0 and 1)
);

    This is a deliberately small starting shape, not a drop-in production schema. Add constraints, ownership, and audit history only when a real workflow needs them.

    Second Brain 2.1 reference map

    The graph is the connective tissue.

    Second Brain 2.1 is designed around an operating brain with evidence-gated learning, task truth, cadence, memory maintenance, provider routing, and automation packaging. This guide explains the architecture behind those pieces. It does not ask you to build every piece at once.

    See Second Brain

    Your first build

    1. 01Pick one real question that needs connected context.
    2. 02Create five entity types and three relationship types.
    3. 03Attach a source and timestamp to every first link.
    4. 04Run one daily recall workflow.
    5. 05Capture one correction or confirmation.
    6. 06Review what changed before adding more automation.

    Build the graph. Then make it useful every day.

    Second Brain adds the operating layer around connected context, so the system can help with priorities, decisions, preparation, learning, and maintenance.

    Explore Second Brain