Quokka

Quokka is an open-source programming language released under the MIT License at v0.1.0, built by a solo developer and hosted on GitHub under the Akshraj2022 account. It is designed around verifiable self-hosting and deterministic semantics, with a current Windows-only installer and a VS Code syntax-highlighting extension included out of the box. The project was created and first published in September 2026.

What It Is

Quokka is an expression-oriented, strictly self-hosted programming language whose core design goal is provable correctness: the canonical lexer, parser, AST, and evaluator are all written in Quokka itself, bootstrapped from a minimal C host of roughly 2,150 lines. It targets developers who want predictable, exception-free code with explicit error handling, and who need a clean boundary between a pure language core and heavy ML workloads.

Core Language Design

Quokka's design centers on three principles that distinguish it from mainstream scripting languages:

• No exceptions — errors are returned as values using Result (Ok/Err) and Option types, combined with match expressions and the ? postfix operator for ergonomic propagation.
• Immutable by default — variables declared with let are immutable; mutation requires explicit let mut, and shadowing requires the shadow keyword.
• Deterministic semantics — the language avoids hidden control flow, making program behavior traceable and auditable.

Verifiable Self-Hosting Architecture

The self-hosting model is a central technical claim of the project. A small C bootstrap host (Stage 0) runs the canonical Quokka interpreter, which is itself written in Quokka. Running quokka check-self re-bootstraps the interpreter through multiple stages and compares cryptographic hashes of the AST and output at each stage. If the hashes match, self-hosting is verified on the user's own machine rather than taken on faith from the project page.

Joey: ML Integration via Process Isolation

Joey is a standard library extension that lets Quokka orchestrate Python/PyTorch/Unsloth workflows — including Llama-3 QLoRA fine-tuning — without embedding Python in the language core. The architecture works through a strictly typed JSON IPC boundary: Quokka code writes a configuration manifest, spawns an isolated Python and Unsloth process for training (with CUDA acceleration), and reads the result back as a typed Result value. AMD/ROCm support is listed as untested. A fully native Joey runtime that removes the Python process hop for supported operations is listed as in progress on the project roadmap.

Update: v0.1.0 Launch

The project published its first release, v0.1.0 (tagged v0.1.0-updated), on September 18, 2026. The completed milestone set includes the C bootstrap host, the self-hosted lexer/parser/AST/evaluator, the quokka check-self verifier, Result/Option error handling, immutable-by-default bindings, the Joey ML extension with CUDA support, a Windows installer, and VS Code syntax highlighting. Active development targets a native Joey runtime, an expanded standard library, AMD/ROCm support, and macOS and Linux builds.

Setup Path

Installation on Windows is a single-step process: download Quokka-Setup.exe from the GitHub Releases page, run the installer, and it automatically configures PATH, registers the .qka file extension, and installs the VS Code extension. Scripts are run with quokka run hello.qk from any terminal. Building from source requires GCC to compile the C bootstrap, after which quokka bootstrap re-bootstraps the canonical interpreter.