mojojojo mojojojo

← Research

The open-source stack behind mojojojo

mojosub, the compiler, the coding agent, and the Mojo port factory are separate projects on purpose. Here is what each one owns, what is ready today, and where the hard gates sit.

Mojojojo is not one repository with a public folder. Its compiler path, coding agent, and library ports are separate MIT-licensed projects with separate tests and failure modes. The hosted service is a consumer of those projects.

ProjectOwnsStatus
mojosubDeterministic Python-subset transpilation, cache, and hot dispatchWorking subset; refuses unsupported Python
mojojojo-compilerHybrid bundles with CPython fallback and safe native hot functionsEarly working slice
mojojojo-agentGrounded repository search, tools, sessions, and executionEarly, installable, MIT
mojo-factoryScaffold, build, test, benchmark, review, and publish library portsActive factory with hard gates

A repository count is not a readiness claim

The factory workspace contains a large and growing set of mojo-* projects. The live service currently knows about 108 ports and marks 49 ready for its runtime. Those are different numbers on purpose: a generated repository, a port that builds, a port that passes parity tests, and a port integrated into the hosted runtime are four different states.

The gate is the product. A port has to build, agree with its reference implementation, and publish honest benchmark results. The factory drops repeated failures instead of declaring success because an agent wrote plausible code.

Why split the projects?

Someone using mojosub locally should not need a mojojojo account. Someone auditing the agent should not have to understand the hosting database. A library port should reproduce its tests and benchmark from its own lockfile. Keeping those boundaries visible makes the hosted convenience optional.

It also lets each layer be blunt about limitations. mojosub accepts a typed numeric subset. The compiler falls back to CPython. The agent is a generator, not an authority. The factory publishes slower-than-upstream rows when that is what the machine measured.

Where to start


Product notes describe what is live at publication time. Anything still being designed is marked as a preview; source repositories carry their own build and test instructions.