{title}
+ {description &&{description}
} +{description}
} ++ Substrate turns the documents your organization already writes — + specifications, regulations, domain models — into a structured, + link-typed corpus that humans and LLMs can read, edit, and + reason about. The same structure that tools can check is the + context an LLM can be trusted to act on. +
+ ++ Every claim a corpus makes — types, dependencies, examples, + test cases — is mechanically checkable. Findings are reported + against the originating prose, not an opaque intermediate + form. +
++ Authors describe the domain in their own words. Links between + documents quietly accumulate a semantic knowledge graph that + tools can traverse, type-check, and slice for an LLM. +
++ Externalize the mental model. Register patterns and + technology mappings against the corpus and let the tooling + generate implementations directly — instead of re-prompting + an LLM for every change. +
+
+ The substrate CLI extracts focused context for an
+ LLM, validates the corpus for consistency and type
+ compatibility, refactors safely across files, and runs the
+ test cases embedded in the prose. Exits non-zero on failure —
+ the same shape as any other check in a CI pipeline.
+
context — extract a tree-shaken slice for an LLMvalidate — check structure, links, and typestest — run every embedded caserefactor — rename and move with references intact