viralcode/machino

Machino is a programming language built specifically for AI agents to write in, rather than for humans. It compiles down to WebAssembly, a portable format that can run the same way in browsers, servers, phones, and edge computing platforms. The idea is that when AI systems write software today, they inherit a lot of ambiguity and undefined behavior from languages like Python, JavaScript, or C++ that were designed for people. Machino tries to remove that ambiguity by giving every piece of code exactly one correct way to be written, and by building contracts and tests directly into the language's grammar instead of treating them as optional extras. Every function in machino can declare conditions that must be true before it runs and conditions that must be true after it finishes. If either is violated, the program stops with a clear, structured error describing exactly what went wrong. Errors from the compiler and test runner come back as data an AI agent can read and act on directly, rather than plain paragraphs meant for a person to interpret. The project also uses a capability system, meaning a machino program can only reach the network, files, or other outside resources if it explicitly declares that it needs them. The whole toolchain, including the compiler, an interpreter, and two different ways of producing WebAssembly output, is written in Rust with no external dependencies, and the codebase is small enough that someone could read through it in a weekend. Machino programs can be run directly, tested, or compiled into a single WebAssembly file that then runs anywhere a WebAssembly host exists, including tools like wasmtime and Cloudflare Workers. The project ships with an example that is a complete HTTP server written in machino, showing that the language can be used to build real command line tools and network services, not just toy programs. Because it is a brand new language, there is very little existing code written in it for AI models to learn from. The project addresses this by including a tool that generates random machino programs, verifies each one actually compiles correctly, and adds it to a training corpus, alongside a compact guide document meant to be pasted into an AI model's context so it can write the language with no prior training.