Settings
Spatial & World APIs
Credentials go directly to encrypted server-side storage. Stored credentials are never loaded back into the browser or written into the world-state graph.
V134 is permanently retained as failed prospective evidence: task labels were never opened, no second paid retry was started, and the generated world can never be retroactively counted as a V134 invariance PASS.
V136 fixes the proof protocol, not the historical result. It cannot spend credits, cannot reuse the failed V134 world as fresh evidence, and cannot raise the AGI/world-model scores before a later full external pass.
V135 is shadow-only and promoted:false. A constraint being verified at source is no longer enough: it must survive a separate applicability probe, negative control and freeze-before-validation gate. AGI heuristic remains 86%.
V137 is not centrally promoted yet. It stores no target feature IDs, probe IDs or outputs. Fresh World Labs external validation is still pending; AGI heuristic remains 86% until a later evidence/readiness gate authorizes a score change.
One fresh World Labs world, one 1,580-credit generation, no retry. The AGI heuristic remains 86% because a single generated virtual world is not AGI or physical-world proof.
V139 is development-only and promoted:false. When several structures are behaviorally indistinguishable it reports the equivalence class and chooses the minimal sufficient explanation instead of inventing certainty. Official meters remain 92.6% world-model, 95.8% learned-world and 86% AGI heuristic.