Zoom Out

Load /qi-layer if it isn't already loaded.

Orient in an unfamiliar area by exploring multiple angles — each covering a different lens — then synthesizing the findings into one orientation map. A single angle sees one slice; zoom-out covers structure and knowledge and history at once, so the picture isn't skewed by one lens.

Use when starting in an unfamiliar subsystem, picking up someone else's work, or when the user says "orient me" / "what's the lay of the land here."

Fan Out

Explore each angle independently, scoped to one question each:

• Structure — callers, dependencies, module boundaries, entry points, where behavior actually lives.
• Documented knowledge — what's been written about this area in the KB, guides, and reference docs — architecture, patterns, synthesized research.
• Decision history — why it's shaped this way — prior choices, rejected alternatives, constraints discovered. Skip when there's no relevant history.
• Domain concepts — the vocabulary in use (KB vocab pages, project guides, context docs).

Scope each exploration to its angle with concrete path or topic targets. Don't let one pass cover two angles — the value is independent coverage.

Synthesize

Read the findings and produce one orientation map, not four stitched-together dumps:

• What this is — the area's purpose and boundaries, in two or three sentences.
• Structure — the modules and interfaces that matter and how they connect (a mermaid diagram if the shape is spatial).
• Why it's like this — the decisions and constraints that explain the current shape.
• Vocabulary — the terms needed to speak about it precisely.
• Open threads — gaps, debt, or unresolved questions surfaced.

When findings contradict each other, resolve it rather than listing both — code is ground truth; a doc that disagrees is stale (note it). Keep the map tight: the goal is a reader who can now navigate the area, not an exhaustive transcript.