Clear jobs

Ring0 starts with three jobs: move the information, run the work, check the result.

That sounds trivial until you notice that almost every blockchain fuses the last two. A validator runs the work in order to check it, which means every machine on the network repeats every computation that has ever happened on it. The check is the repetition. That is why these systems are slow, and it is also why they close over time: if checking is expensive, only expensive machines get to check, and the set of people who can hold the network accountable quietly shrinks.

Ring0 splits them. A producer runs the work and attaches evidence that it ran correctly. A validator checks the evidence on ordinary hardware, and never re-runs anything. Doing the work stays expensive. Confirming it does not.

Evidence travels. Computation does not.

A record that holds

A network needs one answer that survives everything else changing: what makes a result part of the same network?

Ring0 answers at the level of the record, and the record holds exactly four things. State, which is what the network currently recognizes. Authority, which is who may change that state and within which limits. Obligations, which are transfers or work that started and have not finished. History, which is the accepted path explaining how the present came to be.

Those four are the entire contract. A later version of Ring0 may execute differently, store differently, prove differently, and finalize differently. What it may not do is silently create permission, lose a pending duty, duplicate value, or make an earlier claim impossible to inspect.

Private work and long-term custody sit outside that contract on purpose. They are there when an application needs them and absent when it does not. Nobody should pay for a guarantee they never asked for.

Change without confusion

This is a first engine, not a final machine. It says so in the name.

Every design can be wrong, and every machine eventually gets old. A network that cannot replace its parts becomes a museum. A network that replaces its rules by announcement is asking you to trust whoever controls the announcement. Both are common failures, and from the inside they look nothing alike.

ERAS is the rule for the space between them. A proposed successor is a claim to be tested, not an upgrade merely because somebody named it one. It is accepted only when the exact candidate is authorized, ready, continuity-correct, finalized through one unambiguous handoff, and activated inside its approved window. Before that boundary a defective candidate holds no authority at all. After it, a mistake is corrected by moving forward through another valid transition, never by quietly restoring the old one and hoping nobody kept a copy.

An Era may change how work is executed, stored, proved, or finalized. It may not break the continuity that makes its results part of Ring0.

The engine changes. The network remains.