Skip to main content
Nothing on this site should be read as documentation of a working production language. The mzizi-lang README states the position plainly: “Status: prototype front end… The Phase 0 benchmark has not run, so nothing here has yet been measured against the charter’s kill criteria. Contract bodies parse but are not evaluated.”

​
What actually exists

Every row below is checkable in mzizi-dev/mzizi — that is the point of writing it as a table rather than a paragraph. The compiler crate is version = "0.0.0" with publish = false. There is no cargo install mz, no published crate, and no release. You build it from the repository or you do not have it.

​
The kill criterion, and why it has not fired

The charter gives Phase 0 a success criterion that is also a stopping rule:
A defined benchmark where an LLM agent authors N equivalent components in Mzizi’s syntax vs. raw Dioxus/Leptos, measured on tokens consumed, iterations to a clean compile, and defect rate. If this doesn’t show a measurable advantage, nothing downstream matters — don’t build Phase 1 until Phase 0 has a real number attached to it.
RFC-0001 §6 sharpens it further: “If Mzizi doesn’t beat both on at least two of three metrics, the thesis is wrong and Phase 1 does not start.” RFC-0002 §5 adds that the benchmark must include a small open-weight model arm, because the frontier arm alone cannot validate a thesis whose whole point is helping the models that a frontier arm does not represent. No arm of that benchmark has run. So every claim on this site about why a design decision is right is a stated rationale, not a measured result. The two things that have been measured are narrower, and they are reported as such below.

​
What has actually been measured

These come from compiler/tests/ir_measured.rs, over the nine primitives plus the corpus example — ten files, which is a tiny corpus and should be read as such. RFC-0003 §7 flags the first number honestly, and so does this page: “The sharing number is honest but small… At the scale of the real 571-component registry… the ratio should improve substantially — but that is a prediction, and it stays labelled as one until the registry is lowered.” The test suite is 75 tests across compiler/src and compiler/tests. CI gates the directory on cargo fmt --check, cargo clippy --all-targets -D warnings, cargo test, and a mz check over every primitive.

​
Where the RFCs and the implementation disagree

Worth knowing before you copy an example out of an RFC and expect it to compile.
RFC-0001 and RFC-0002 both carry the header “Status: draft for review — nothing here is implemented”. That was true when written and is no longer true: the front end exists. RFC-0003’s header is current (“draft; core implemented in this PR”), as is RFC-0004’s. Treat the headers as historical.
RFC-0001 §1 writes view attributes without an equals sign — text state.label, strip slot "nyuchi-connectivity-bar". The parser that exists requires name = value (compiler/src/parse.rs), and every .mz file in the repository is written that way. Syntax documents the implemented form.
Every primitive sets portal = "https://mzizi.dev/components/<name>", pointing back at the registry entry for the component it ports. It is not in RFC-0001’s attribute list.
RFC-0001 §4.3 describes mz fix applying all exact fixes in one shot, and RFC-0003 §5 lists refs, path, patch and diff under “Designed here, next in implementation order”. The binary dispatches exactly four commands today: check, outline, hash, ir. See The compiler.
RFC-0003 §4 shows a compact outline for button with enum variants on one line and contract 5. The emitter in compiler/src/outline.rs lists variants one per line and emits an empty contract block. The measured cost claim in §7 is from the real emitter; the printed sample is not.

​
The private repository does not exist yet

RFC-0004 specifies a private held-out benchmark repository in the mzizi-dev org, and then argues at length that it should not be created yet — because a private runner that exists before the public harness does will shape the harness around itself, inverting the dependency rule the whole RFC is built on. The trigger is stated: “the first held-out task. Not before.” The public half of the mechanism is implemented and inert: .github/workflows/mzizi-lang-benchmark-dispatch.yml reports “not configured” because the MZIZI_HELDOUT_REPO variable and MZIZI_DISPATCH_TOKEN secret do not exist. That is the intended steady state, not a failure.