Five ways teams put the compiler to work. Each page gives the audience, the path from source to World, what the workflow leaves behind, and the limits that come with it.
AI-ready knowledge
For: AI and platform engineers
Teams repeatedly clean, chunk and index the same sources for each model or application. Identity, relationships and provenance drift between projects.
Compile document collections into a versioned World before retrieval, assistants or agent workflows consume them.
Document intelligence
For: Document and operations teams
A clean text dump hides the printed order, the uncertainty and the exact source geometry needed to review an extraction.
Move from difficult PDFs and scans to reviewable structure, without losing the exact source location each result came from.
Knowledge graph
For: Data and knowledge architects
A graph that cannot show why an edge exists is difficult to review, govern or trust downstream.
Turn document facts into stable semantic objects and evidence-bound relations inside a versioned World.
Grounded assistants
For: Application and agent developers
An answer can sound confident while depending on stale, conflicting or untraceable source material.
Ask, API and MCP consume the same current World and return evidence from the same version.
Knowledge operations
For: Knowledge owners and security reviewers
Automated extraction becomes operational risk when updates silently replace the knowledge used by production systems.
Separate candidate compilation from the active World, preserve change history and keep human decisions explicit.
Next: the guides, samples and research in the resources hub, or the exact product contract in the documentation. These five pages describe how the compiler is used, not completed customer engagements.