How it works
From raw data to inferred motivations
01
Observe
The Observer reads windows of raw interactions and prompts an LLM to infer what the user thinks and feels — grounded in named behavioral evidence.
02
Synthesize
Observations are grouped by session and synthesized into insights with context on when each pattern applies.
03
Layer
Insights from multiple sessions are recursively merged into higher-order patterns
04
Explore
Navigate the resulting lattice interactively — click any node to read its full text and trace it back to its supporting observations.
Explore