Skip to content

OPEN SOURCE

Proof of work, shipped in public.

Anito is a new practice, so the proof is not a testimonial wall — it is working code. These are the tools Anthony Garces is building in the open: verification machinery that makes software claims checkable. Read the code, the commits, and the honest statuses.

Every project links straight to its repository. Pre-release means pre-release.

Flagship · ranex.dev

Ranex

Deterministic governance for AI agents that build software. A kernel of ordinary, inspectable code stays outside the AI's loop and judges work by evidence and executable checks — never by model confidence.

The stated invariant: removing every model credential from the machine must not change a single verdict.

Status, said plainly: pre-release. The verdict path works and gates Ranex's own test suite; flow graphs and scenario compilation are designed, not built. Known gaps are listed in the README — not hidden in the marketing. Arxic and Leitir are folding in as Ranex features, and ranex-harness puts the kernel in charge of a real agent workflow.

Verdicts from evidence, not confidence

A verdict is a pure function of gate, evidence, subject, and approver. No evidence means FAIL — absence blocks, nothing passes by default.

Bound to the exact commit

Evidence is tied to a commit digest. A green light applies to the code it was produced from — not to whatever landed after.

An append-only, hash-chained journal

Every decision lands in a tamper-evident record with signed evidence. You can replay how any verdict came to be.

No self-approval

The agent that did the work cannot approve the work. The kernel stays outside the AI's loop — ordinary, inspectable code.

Why this matters to your project

The discipline in the repos is the discipline in your build.

Open source is not a side hobby here — it is how Anito proves the habits your project depends on.

Fail-closed gates in Ranex

No AI action touches your live systems without an explicit human approval step.

Hash-chained journals and signed evidence

Every automation ships with a structured audit log — you can see what ran and why.

Provenance-bound verification

Claims in your handover docs point at commits, configs, and logs — not at promises.

No self-approval, ever

You hold the credentials and the off-switch. Anito never grades its own work behind your back.

Want this level of rigor on your own system? Start with the free diagnosis.

See your engine map