Skip to main content

What Are Aspects?

Every fact in your CORE memory is automatically classified into one of 11 aspects. Think of aspects as filing cabinet drawers - when you need something specific, CORE knows exactly which drawer to open.

The 11 Aspects

AspectWhat it capturesExampleWhen it surfaces
IdentityWho you are (slow-changing attributes)“Manik works at Red Planet”Understanding your role, company, or background
KnowledgeWhat you know and understand”Expert in TypeScript and React”Explaining concepts or choosing complexity level
BeliefWhy you think that way”Values simplicity over complexity in code”Architectural decisions or design choices
PreferenceHow you want things done”Prefers pnpm over npm”Code reviews, setup, choosing tools
ActionWhat you actually do (habitual behaviors)“Reviews PRs every morning before standup”Suggesting workflows or understanding routines
GoalWhat you want to achieve”Launch MVP by Q2 2026”Planning features, prioritizing work
DirectiveHard rules and automation triggers”Always run tests before creating a PR”Always - directives are enforced rules
DecisionPast choices and their reasoning”Chose Neo4j for graph storage because of relationship traversal performance”Encountering similar technical choices
EventThings that happened (with timestamps)“Started working on authentication refactor on Jan 15, 2026”Temporal queries (“what happened”, “when did we”)
ProblemChallenges, blockers, issues”Hit rate limits with GitHub API when syncing 50+ repos”Debugging similar issues or planning around constraints
RelationshipConnections between people, teams, projects”Works with Sarah on CORE project”Understanding team dynamics or collaborators

Why Aspects Matter

When you ask “What are my coding preferences?”, CORE doesn’t search through all your conversations. It:
  1. Classifies your query as an Aspect Query for Preference
  2. Filters statements where aspect = Preference
  3. Returns only preference facts: “Prefers TypeScript”, “Uses pnpm”, “Avoids class components”
Result: 5 precise facts in 300ms, instead of 50 maybe-relevant chunks in 2000ms.

How They’re Used Together

What the agent is doingAspects surfaced
Writing codePreference, Directive, Knowledge
Code reviewPreference, Directive
Architectural decisionDecision, Belief, Problem
DebuggingProblem, Knowledge, Decision
Planning featuresGoal, Decision, Directive
Explaining conceptsKnowledge, Identity