Same input, same result
Every run is checked by hash, and snapshot and replay must agree. When something breaks, you can roll the exact scene back and reproduce it.
In development
Liv-ay World is a runtime for creating, editing, running, and verifying worlds. What a person places by hand and what an AI plans and builds run on the same rules, and every run is checked by snapshot and replay to confirm the same input still yields the same result.
The world kernel, WorldTask execution, and performance gates are in development.
The moment several people and an AI edit one world, an unreproducible result means nobody can say what went wrong. Liv-ay World treats reproducibility as a premise, not a feature.
Every run is checked by hash, and snapshot and replay must agree. When something breaks, you can roll the exact scene back and reproduce it.
The AI never touches the world directly; it plans and calls the same editing tools a person uses. So what the AI did can be undone like anything else.
Frame and memory budgets run as CI gates, so regressions are caught before release.
You state the goal; planning, execution, and verification follow.
Say what you want to build.
The AI lays out the editing steps.
The world is edited with the same tools.
It lands only when hash and replay agree.
Ask us about Liv-ay World's progress and collaboration.