Governance: see PRINCIPLES.md.
- code-translation-ir.md — LMPL is a semantic pivot for code translation, above compiler IR and below prose.
- llm-procedure-language.md — LMPL is an operational script for LLM runtimes, without requiring a deterministic interpreter.
- agent-workflow-orchestration.md — LMPL describes agent workflows independent of any one orchestration framework.
- runtime-self-description.md — Semantic compression only holds when LMPL remains self-describing at runtime with minimal extra context.
- latent-logic-substrate.md — LMPL works by tapping latent logic and control structures learned from code, deeper than language-specific syntax.
- structured-pseudocode-as-semantic-compression.md — Structured pseudocode should compress complex model-mediated work better than unrestricted natural language.
- testing-strategy.md — LMPL's empirical claims should be tested across translation, procedure-following, token efficiency, workflow portability, and human reviewability.