Composition

How token flow is composed across the four pillars (input, output, cache_creation, cache_read). The structural pattern of token usage.

The four raw token pillars.

Ratios between pillars (transmission, commitment, reuse), cascade structure.

Composition reveals how an operator structures their AI interaction — not just how much they produce, but how they use cache, reuse, and fresh input.

Compare composition profiles across operators and archetypes.

Pillar ratios, cascade stage values, archetype classification.

If composition profiles are identical across all operators, composition is not a meaningful measurement.

10 composition archetypes identified across 4 families in field data.

Composition is descriptive, not evaluative — a composition profile is not "better" without context.

SigRank archetype taxonomy, cascade ontology. Architecture: mos2es.com/architecture.