Corpus Algebra: Composable Vocabulary Intelligence at Scale

Per-source frequency tables turn documentation into a cheap map of vocabulary gaps, team silos, and knowledge drift. Because term counts merge by simple addition, each repo becomes a composable atom: build task-specific unions, compare them against a baseline, and diff them through git history. The result is corpus intelligence that answers questions embeddings usually blur: which concepts are overrepresented, which are missing, and which teams are using different words for the same work.