# Efficient New Language / EML > EML is a deterministic semantic-overlay programming layer: it compresses high-frequency program > intent into compact symbols and transpiles, rule-based and reversibly, to standard languages > (canonical target: Python), preserving executable truth through parsing, transpilation, > interpretation, eml-trace-v1 trace generation, and round-trip fixpoint validation. It is not > a replacement language — it is an overlay for humans and AI agents alike. No LLM in the core > transpilation chain. Canonical site: https://efficientnewlanguage.org/ Maintainer: Neo.K / EveMissLab License: Apache-2.0 · Patent: Taiwan Utility Model M672933 (Taiwan only) Spec: EML-LANG-2026-v1.0 · AI layer: 0.1.0 ## The EML family (since 2026-09-27) The site's front page is now the home of a language family for the AI era. The specification, examples and tools listed below belong to EML-P, the member that runs today. - EML-U — Universal Semantic Profile: languages can evolve. Explicit semantic attachment and composition on existing hosts, one semantic model projected to C++20, Rust 2021 and EML-P/Python, with evidence for every supported transformation. A bounded experimental MVP (snapshot 2026-10-05: four integer profiles, 297 tests, zero accepted production cases; corpus expansion next). An AI that iterates languages this way is the research direction, not yet demonstrated: /eml-u/ · overview: /ai/eml-u/overview.md · the original theory: /origins/ - EML-P — Practical Execution Profile: the stable, linear, executable subset, EML-P ⊆ EML-U, transpiled deterministically to Python and checked against CPython. The former homepage: /eml-p/ - EML-NOVA (N) — the structure-first research line: programs as typed graphs with a stable identity, changed by patches and verified as they are built. Concept page: /nova/ - Standard library (levels L0–L5, including loops as Scan or While over library functions, integer indices, dense linear algebra, the QR and singular value decompositions, floor division, the Fourier transform, special functions such as erf and log-gamma, distribution functions such as the normal quantile and the incomplete gamma and beta functions and their inverses (chi-square, gamma, beta and t quantiles), differentiable in their shape parameters too, and reproducible random numbers): every function a typed NOVA graph, equal to an independent reference in exact arithmetic and inside an error bound computed from the graph: /nova/lib/ (one page per function: /nova/lib///) · primitives: /nova/primitives/ Each function is also a control handle (Ω:std..): pinned to its call closure, with its verification record's hash as evidence, shown on its page. - NoGlyph — Blank Generative Constraint Projection Language, the constraint-first research line: the visible source can be completely blank; the program is a canonical set of constraints, resolved deterministically into a program graph and lowered to C11 for a conventional compiler. Reference MVP-0 to MVP-7 implemented and revalidated locally: /noglyph/ - CVSG — Chinese Variable-Spectrum Semantic Graph Language, the Chinese-first research line: Chinese is the authoring surface over a typed semantic graph with spectrum-valued state and a versioned rulebook; rules run as deterministic, replayable world-state transitions. Stage O1.0 closed: /cvsg/ - One further line exists and is not public. ## Entrypoints - Human site: / - Playground (in-browser, runs the real toolchain): /app - Docs: /docs - Origins / EML-U (the original, broader vision — human + AI page): /origins - Related languages & specifications (sibling EveMissLab projects, not EML): /related - CAIR — Canonical Authoritative Intermediate Representation (authoritative program ontology, multi-projection, candidate write-back; v1.0 frozen): /related/cair/ - ICNS — Interpretation-Complete Numeral Specification (declare a numeral's carrier, grammar, ordering, resolution and successor before use; v1.0 frozen): /related/icns/ - MNVP — Multi-layer Numerical Visualization Protocol (the display tier above ICNS: every render declares what it preserved, omitted and whether it stayed faithful; v1.0 frozen): /related/mnvp/ - WRPL — World-Rooted Projection Language (one versioned world projected into verified surfaces for people, programs and AI agents; v0.1.0 reference MVP, published early: never used on a real world model, so its practical benefit is undetermined): /related/wrpl/ - AI-native interface: /ai/index.md - Machine manifest: /ai/manifest.json - Version: /ai/version.json ## Specification - Current spec (digest): /ai/specs/eml-v1.md - Self-contained AI-semantic spec (status vocabulary, twelve-loop taxonomy, bug/repair/criticality models): /ai/specs/eml-semantic-model-v1.5.md - Grammar (EBNF, normative): /ai/specs/eml-grammar.ebnf - AST schema: /ai/specs/eml-ast-schema.json - Trace schema (eml-trace-v1): /ai/specs/eml-trace-schema.json - Error / diagnostics schema: /ai/specs/eml-error-schema.json ## Corpus - Origin / genealogy: /ai/corpus/eml-origin.md - EML-U (universal semantic profile — the original, broader vision): /ai/corpus/eml-u-profile.md - EML-U today (dated 2026-10-05: the bounded MVP, its limits, the corpus plan): /ai/eml-u/overview.md - EML-U engineering snapshot (JSON summary of dated observations, not a raw receipt): /ai/eml-u/engineering-snapshot.json - EML-P ↔ EML-U compatibility (projection function, no-silent-loss rule): /ai/corpus/eml-p-eml-u-compatibility.md - Current implementation: /ai/corpus/eml-current.md - Design history: /ai/corpus/eml-design-history.md - Engineering notes (invariants): /ai/corpus/eml-engineering-notes.md ## Examples - /ai/examples/000-arithmetic.eml.md - /ai/examples/001-summation.eml.md - /ai/examples/002-conditional.eml.md - /ai/examples/003-coldhot.eml.md - /ai/examples/004-closure.eml.md - /ai/examples/005-recursion.eml.md ## Projects (EML-P, since 2026-09-27) Functional software written in EML-P: several modules behind a text menu, built on verified corpus cases. Each recorded session is replayed under CPython and, when the site is built, in the EML interpreter; a session whose screen does not match is not published. - Index: /eml-p/projects/ · the full list: `projects` in /ai/manifest.json - One project: /eml-p/projects/-/ (page) · /ai/projects/-.md and .json - First project: P001 Todo list — /eml-p/projects/P001-todo-list/ · /ai/projects/P001-todo-list.md ## Tools (bounded, machine-callable) - Tools guide: /ai/tools/tools.md - Tool catalog: /ai/tools/tool-catalog.json - OpenAPI 3.1: /ai/tools/openapi.json - POST /ai/tools/parse · /ai/tools/transpile-python · /ai/tools/transpile-eml · /ai/tools/interpret · /ai/tools/trace · /ai/tools/roundtrip - GET /ai/tools/health ## Recommended reading order for LLMs 1. /ai/index.md 2. /ai/corpus/eml-origin.md 3. /ai/corpus/eml-u-profile.md 4. /ai/corpus/eml-current.md 5. /ai/specs/eml-v1.md 6. /ai/specs/eml-grammar.ebnf 7. /ai/examples/001-summation.eml.md 8. /ai/tools/tool-catalog.json ## Status EML is an active experimental language substrate and AI-agent-readable semantic interface.