Skip to content

EXPLAINME.adoc

Latest commit

 

History

History
319 lines (251 loc) · 11.7 KB

EXPLAINME.adoc

File metadata and controls

319 lines (251 loc) · 11.7 KB
Raw

Next-Gen Languages — Show Me The Receipts

Table of Contents

The README makes claims. This file backs them up with architectural context, honest status per language, and enough structural detail for an external reviewer to understand the relationship between the languages, their -iser counterparts, and the rest of the hyperpolymath ecosystem.

Important
 This is a pure coordinator repository — no language implementation code lives here at all. Each language is its own standalone canonical repo at hyperpolymath/<lang>, referenced from here, not vendored. (The submodule pointers were removed 2026-06-13; there is no .gitmodules.) This repo holds cross-language coordination only. See the COORDINATOR BOUNDARY in 0-AI-MANIFEST.a2ml and EXTRACTION-MANIFEST.md.
Note
 tentacles-agentic-syllabus is NOT a programming language. It is an educational framework using My-Lang’s dialects as a curriculum vehicle. Never audit it as a language implementation.

Claim 1: "A family of languages covering the full spectrum of modern software challenges"

A coordinated effort to design and develop a family of programming languages that collectively cover the full spectrum of modern software challenges, from AI-native development to formally verified real-time systems.

— README

How it works

The family currently spans fourteen languages plus the KitchenSpeak DSL — the canonical list is .machine_readable/LANGUAGES.a2ml. They are designed to complement each other, not compete, each occupying a distinct cell in the design space (each maintained as a standalone hyperpolymath/<lang> repo). The table below shows the representative core:

Language Primary Paradigm Key Technical Mechanism Implementation Language

Solo/Duet/Ensemble (my-lang)

Imperative → AI-assisted → AI-native

Progressive dialect family: Solo → Duet → Ensemble. @synth and @verify annotations in Duet. AI<T> effect type in Ensemble.

Rust compiler

Phronesis

Declarative, logic-based agent ethics

Agent., Values:, EVALUATE(…​) spec constructs. Formal AI ethics specification language.

Rust (per roadmap)

Eclexia

Declarative, constraint-driven

(energy budget …​), (resource …​) constraints propagated through type system. Green computing / IoT.

Rust

Oblíbený

Turing-incomplete for deployment, metaprogramming for authoring

(forbid recursion), (bounded-for …​). Designed for HSMs and secure enclaves where non-termination is a security violation.

OCaml (per language design conventions)

Anvomidav

Functional, concurrent, formal (linear/session types)

task @sched(EDF), Linear<T>, Π (…​) . T. Hard real-time: avionics, autonomous vehicles, robotics.

Rust + Idris2 proofs

WokeLang

Imperative, natural-language

only if okay "…​", attempt …​ or reassure. Human consent as a first-class language primitive.

OCaml parser, Rust evaluator

AffineScript

Affine types targeting WASM

Affine type system (use at most once) for WebAssembly. Companion to Ephapax (which is dyadic: affine + linear).

OCaml compiler

Ephapax

Dyadic linear type system

let x = …​ (affine) vs let! x = …​ (linear). Targets WebAssembly. Gossamer backend. Rust compiler, Coq proofs.

Rust (17 crates), Coq proofs
Note
 The table shows a representative core. The full family also includes betlang (probabilistic), JtV (systems / Harvard architecture), Haec (trope-particularity), error-lang (pedagogical), Tangle (topological), and 007 (a private agent meta-language) — see .machine_readable/LANGUAGES.a2ml for the canonical registry.
Note
 KitchenSpeak — an experimental applied-type-theory teaching DSL (seven-type kitchen orchestration: Tropical, Linear, Choreographic, Echo, Dyadic, Ceremonial, Primitive) — is NOT one of the core languages above and is no longer carried in this repo. It is canonical at the standalone hyperpolymath/kitchenspeak repository (v2.0: physics-in-the-types, reference OCaml compiler, machine-checked Agda proofs). Per ADR 0004 its Echo type attaches to the Linear/Dyadic consumption types on the echo-types substrate.

Most mature: Ephapax

hyperpolymath/ephapax is the most advanced implementation — a Cargo workspace with 17 crates:

  • ephapax-lexer — tokeniser

  • ephapax-parser — AST construction

  • ephapax-analysis — type-checking and linearity analysis

  • ephapax-desugar — surface → IR desugaring

  • ephapax-ir — intermediate representation

  • ephapax-interp — interpreter

  • ephapax-cli — command-line entry point (ephapax run, ephapax check)

  • ephapax-lsp — Language Server Protocol implementation

  • ephapax-repl — REPL

  • ephapax-package — package management

It also carries two arXiv paper sources (arcvix-code-as-matter.tex and arcvix-dyadic-language-design.tex) and Coq soundness proofs. See that repo’s README.adoc and ROADMAP.adoc for current status.

Honest caveat

The language family is at different stages. Ephapax has a working type checker, interpreter, LSP, and REPL with ~307 tests. WokeLang has a parser and evaluator but is pre-production. Anvomidav has its type system designed but the compiler is early. Oblíbený and the my-lang dialect family are active. An external reviewer should consult each language’s own repo (its CHANGELOG and ROADMAP) for current implementation completeness rather than assuming the table above means "production ready."

Claim 2: "Languages vs. -iser tools — implementations in their repos, extractions in developer-ecosystem"

Each programming language has full compilers, parsers, type checkers, and interpreters in its own repo. The -iser repos in developer-ecosystem let you use each language’s unique power without learning the full language.

— README

How it works

Each language has a corresponding -iser tool in developer-ecosystem/:

-iser Tool Based On What It Extracts

ephapaxiser

Ephapax

Linear type checking for Rust code — Ephapax’s linear types as an external linter

wokelangiser

WokeLang

Consent annotations and well-being checks for any codebase

eclexiaiser

Eclexia

Resource budget analysis for any project

anvomidaviser

Anvomidav

Formal verification harness for real-time code

oblibenyiser

Oblíbený

Security hardening — Turing-incompleteness guarantees as an external tool

affinescriptiser

AffineScript

Affine type linting for WASM projects

phronesiser

Phronesis

AI ethics audit for agent codebases

The design principle: learn the language when you want its full power; use the -iser when you want one specific guarantee applied to existing code.

Honest caveat

Most -iser tools are earlier stage than their parent languages. The split between a language and its -iser is architectural policy, not evidence that either side is complete. Check developer-ecosystem/ for the current status of each -iser before assuming it is deployable.

Dogfooded Across The Account

Technology Role in Next-Gen Languages Also Used In

Rust

Primary compiler implementation language for Ephapax (17 crates), WokeLang evaluator, Eclexia, Anvomidav. #![forbid(unsafe_code)] policy enforced.

gossamer, verisim, gitbot-fleet, maa-framework

OCaml

AffineScript compiler, WokeLang parser (per language design convention: OCaml is the allowed language for compiler front-ends)

developer-ecosystem/ocaml-ecosystem/

Idris2 ABI (per ABI/FFI Universal Standard)

Anvomidav type proofs; Ephapax has ephapax-proven and ephapax-proven-ffi crates with Coq proofs

Gossamer, Groove, Stapeln, VeriSimDB

Coq proofs

Ephapax linearity and termination proofs (arXiv paper backed)

maa-framework/absolute-zero, echidna (Coq is one of echidna’s prover targets)

Julia

language-status-tracker.jl — batch script tracking implementation progress across all languages from this coordinator

developer-ecosystem batch processing, data analysis

Hypatia scanning

.hypatia/ CI workflows

Every RSR repo in the account

Stapeln containers

Account-wide layered container tool; the coordinator ships no build container (it contains no language code to build)

All containerised services in the account

Where Each Language Lives

This coordinator does not contain language code; it references it. Each language is a standalone canonical repo:

Language Canonical repo

AffineScript

hyperpolymath/affinescript

Anvomidav

hyperpolymath/anvomidav

betlang

hyperpolymath/betlang

Eclexia

hyperpolymath/eclexia

Ephapax

hyperpolymath/ephapax

error-lang

hyperpolymath/error-lang

JtV

hyperpolymath/jtv

My-Lang (Solo/Duet/Ensemble)

hyperpolymath/my-lang

Oblíbený

hyperpolymath/oblibeny

Phronesis

hyperpolymath/phronesis

Tangle

hyperpolymath/tangle

WokeLang

hyperpolymath/wokelang

KitchenSpeak (experimental DSL)

hyperpolymath/kitchenspeak

What This Coordinator Holds (and what it does not)

Path What it is

README.adoc, EXPLAINME.adoc

Research overview and these receipts (cross-language, not per-language).

language-status-tracker.jl

Julia script tracking implementation completeness across all languages. Authoritative cross-language "what is actually done."

TOOLING-STATUS.adoc, TEST-NEEDS.md, PROOF-NEEDS.md

Cross-language readiness / test / proof trackers.

docs/disambiguation/

Cross-language anti-confusion notes (e.g. ephapax-vs-affinescript).

wiki/ (overview, tooling, tutorials, frameworks)

Language-agnostic education only. Per-language pages are NOT kept here.

.machine_readable/

A2ML coordination metadata (STATE, META, ECOSYSTEM, AGENTIC, NEUROSYM, PLAYBOOK; LANGUAGES registry; contractiles; bot_directives). Canonical AI session state.

hooks/

Repo-hygiene validators, including validate-coordinator-boundary.sh which fails the build if language-implementation content leaks back in-tree.

extraction-queue/

Per-language docs that were pasted into this coordinator and are staged for relocation to their language’s own repo. See EXTRACTION-MANIFEST.md. Temporary.

(NOT here)

Compilers, grammars, proofs, language source, per-language specs/design/wiki — those live in each hyperpolymath/<lang> repo.

Governance & CI Posture

CI governance (the governance/ jobs) and the Hypatia neurosymbolic scan are consumed from hyperpolymath/standards as *pinned reusable workflows — the enforcing rulesets live there, not in this repo. Two consequences for readers and for automated agents:

  • The Hypatia Security Scan comment on PRs is advisory and non-blocking. The blocking gates are the governance/, scan/, analyze, semgrep, validator, dogfood (including the coordinator-boundary job), and dogfood checks.

  • Recurring advisory findings that have been triaged and accepted are recorded once in .machine_readable/6a2/NEUROSYM.a2ml under [accepted-findings]. Agents consult that registry and do not re-fix accepted findings on every PR. Changing the enforcing ruleset itself must be done in standards.

Questions?

Start with TOOLING-STATUS.adoc for the current state of each language. For the most complete implementation, start with the hyperpolymath/ephapax repo. For language design philosophy, see each language’s own repo and the arXiv paper sources carried there.