Complete, runnable examples: a grammar, the guide that walks you through compiling it on b3u.dev, and real code that consumes what comes back.
This use case publishes no code. It publishes the 29 step specifications and 29 retrospectives that produced a working ~39,000-line application, so you can read the input and judge the method for yourself. New here? Read Adaptive Programming — a user's guide in the documentation section first: it explains what a step file is and where its content comes from, using this use case's own step.1.txt as the worked example.
A complete, minimal CCS use case: json.sgr describes JSON in SGDL (Source Grammar Definition Language); compiling it on b3u.dev yields ccsjson, a native JSON compiler frontend that turns JSON files into CCS SCB instances your own code can load and walk.
The companion to the ccs_json use case. That Z ends where your frontend writes .bfgr SCB instances; this one shows what consuming them from C++ looks like in practice — with real, runnable code and honest numbers against nlohmann/json and simdjson.
MeTTa is not an ALGOL-family language. It is s-expression based, its symbols contain characters an identifier may not, and it carries two sigil-prefixed namespaces — $variables and &space-refs — that have no counterpart in SGDL's predefined token classes. It is, in other words, exactly the kind of language a compiler-compiler built for C-like syntax should struggle with.
The companion to the metta use case. That one ends where your frontend writes SCB binaries; this one picks up there — what it costs to produce those binaries once and read them many times, with runnable C++ that does the reading.
The published MeTTa benchmark reports its headline numbers from a Rust consumer, and says plainly that the consumer language is the dominant variable. The metta_bench use case teaches the C++ one. This use case closes that gap.
A working transformation tool for CCS grammar binaries, shipped as an open platform: the point is not the algorithm we wrote, it is that you write your own. Around 350 lines of C++ give you a loaded runtime, a place to transform it, and a re-serializer — plus the two oracles that tell you whether your transformation was honest.
A working Python web service was translated into a working TypeScript one by a pipeline, not by hand. You can run the translation yourself and then run a test suite nobody here wrote against the result. One command does both. New here? The companion paper Cross-Language Translation in the documentation section is the whole story; this use case is the executable half of it.