Programming language & developer tools / Case study

Kelft

An AI-native, intent-first programming language with a self-hosted native compiler and a growing developer toolchain.

Who it serves
Developers exploring typed native programs and an intent-first approach to AI-assisted software development.
What I built
Created Kelft and its developer tooling, spanning language design, a self-hosted compiler, public specifications, native distribution, and editor support.
Availability
Experimental alpha
Experimental alphaKelft · Native compilation · Rust / C hosts · VS Code
KELFT / COMPILER ARCHITECTURE
  1. Typed Core source
  2. Check and lower
  3. Emit native objects
  4. Link executable

Architecture walkthrough: source becomes checked intermediate forms, native objects, then a linked executable. The host provides platform mechanisms at a separate boundary.

Read about self-hosting and native execution boundaries

The problem

AI-assisted development still needs explicit requirements, controlled permissions, and evidence for accepting changes. Kelft explores these concerns through a programming language and its tooling, while supporting directly authored typed programs.

How it works

Engineering decisions

The result

Published compiler 0.1.0-alpha.2 with native release assets, public Core and Intent specifications, source-build instructions, and editor tooling.

Current status

Kelft is available as an experimental alpha. The native compiler and broader AI-assisted authoring workflow are at different stages of development. Check the release notes for available platforms and supported features.

Explore the project

Working through a similar problem?

Discuss your project ↗