Autonomy Realms
Public Commons Claim Realm Docs

AI Analysis

Analysis is how Autonomy Realms extracts structured metadata from your content. The AI reads your signal (or cluster, or shape) and generates fields like title, summary, tags, themes, narrative arc, and more — based on configurable prompts you control.

Two-Pass Analysis

Analysis runs in two passes, each targeting different fields:

Surface Pass

The surface pass generates core metadata:

  • Title — a clear, descriptive title
  • Summary — concise overview of the content
  • Tags — relevant keywords for categorization
  • Date — when the content was created or occurred
  • Location — where it took place
  • Energy — the emotional or energetic quality of the signal

Structure Pass

The structure pass extracts deeper patterns:

  • Themes — the central ideas and concerns
  • Narrative Arc — the story structure (setup, development, resolution)
  • Symbolic Elements — archetypal or metaphorical content
  • Entities — people, places, organizations mentioned
  • Key Quotes — significant passages worth highlighting
  • Emotional Tone — the overall emotional register

Not all fields are available for every signal type. A photo signal, for example, has different fields than a transmission.

Field Selection

You do not have to analyze every field at once. Before running analysis, you select which fields you want the AI to process. Only those fields are included in the prompt — unselected fields are omitted entirely. This lets you run targeted analyses or re-analyze specific fields without overwriting others.

Questions Define the Prompts

Each field has a corresponding entry in a questions.json file that defines three things:

  • question — what the AI is actually asked about that field
  • format — the expected JSON response structure
  • definition — a brief description giving the AI context about the field

These questions are assembled into the analysis prompt based on your field selection. You can customize questions per realm in Settings > Templates.

Works on Signals, Clusters, and Shapes

Analysis is not limited to individual signals. Clusters and shapes have their own analysis pipelines:

  • Cluster analysis — the AI reads summaries of all signals in the cluster, then generates cluster-level metadata (themes that span signals, overarching narrative arc, etc.)
  • Shape analysis — the AI reads summaries of all clusters in the shape, identifying patterns across the broader collection

Each level uses its own prompt templates and question files.

Re-Analysis

You can re-run analysis on any signal, cluster, or shape at any time. This is useful when:

  • You have added new content (e.g., more signals to a cluster)
  • You have changed the prompt templates
  • You want to analyze fields you skipped the first time
  • You want a fresh interpretation from the AI

Re-analysis overwrites the selected fields with new AI-generated values. Fields you do not select are left unchanged.