Work should survive the intelligence that started it.
Ordivon keeps AI work coherent when a model session ends, a process restarts, or one provider replaces another. It preserves task meaning, execution evidence, and a recoverable path forward.
Tested under replacement
31 July 2026 · Harness H1–H5One task survived Codex↔Hermes replacement and three injected faults.
Four live provider runs completed both replacement orders. Host rejected stale completion, rejected success without an artifact, and recovered a response-lost Runtime job without dispatching the work twice.
- Provider runs4Real Codex App Server and Hermes ACP runs, not simulated adapters.
- Replacement orders2 / 2Codex→Hermes and Hermes→Codex both completed one coherent task attempt.
- Injected faults3Stale assignment, missing artifact, and response loss were contained without inventing completion.
State ownership
Three consequence boundaries and one research plane keep the work legible.
Host, Runtime, and World preserve different consequence-bearing facts. Computing tests which contracts deserve to remain shared.Replace the model session without replacing the Task.
↗testedCommit one Agent action to local reality without losing what happened.
↗experimentalLet a Task reach and act through a changing external world without rebuilding the world beneath it.
↗activeKeep cognition, Task continuity, semantic execution, and machine execution separable—and delete abstractions that mature systems already carry.
↗Under test now
Ordivon Game · testingWhich game structures become possible only when Agents persist and act through the same Host?
Testing with a released application. Station Zero now demonstrates persistent specialists, partial observations, communication whose reachability changes outcomes, player authority, provider replacement, replay, diagnosis, and exact run comparison. It still does not prove that these mechanics outperform a strong classical implementation or equal-budget single Agent.
Recent arguments
Experiments, failures, and the judgments they changed.
Read the complete reasoning behind a boundary, a deletion, a release, or a claim that survived real pressure.What Survived When Codex and Hermes Replaced Each Other Mid-Task
How one task remained coherent through Codex→Hermes and Hermes→Codex replacement, three injected faults, and materially different provider lifecycles.
Addresses: Which Harness objects survive live provider replacement without duplicating Host or Runtime state?↗The Smaller Core That Survived Strong Baselines
How Core Work System Round 1 used strong classical baselines and live Codex/Hermes replacement to remove a Task Runtime, reject a generalized Context Kernel, shrink Effect, localize DecisionRequest, and retain provider-neutral Task state.
Addresses: Which Agent-native responsibilities remain after strong classical baselines?↗Winning the Move Can Lose the Contest
How Ordivon Security Round 1 used local dynamic opponents, CAGE Challenge 4, and bounded Hermes/Codex diagnostics to separate tactical success from strategic outcome without promoting a Campaign engine, organization ontology, strategic state, or custom cyber range.
Addresses: Can strategic adversarial trajectories be evaluated without collapsing them into one reward or success flag?↗Latest publication · What Survived When Codex and Hermes Replaced Each Other Mid-Task