Causal event replay

event index1

—event

—kind

—gate

—causal roots

Nearby event DAG

Frozen audit packet

Recursive evolution component test

Repeated feedback updates each node's response model. Hidden simulation truth is used only here to measure validation error; the runtime auditor cannot read it.

Horizon sensitivity

The same frozen event is audited at H=1…6. Unlike v0.5.0, horizon is causally connected to projection and uncertainty.

What v1.0 operationalizes

Causal history is a DAG with sparse vector clocks. Multiple independent roots can converge on one event. Primary audit, power audit, human gate, and postmortem are IPS outputs rather than invisible external judgments. The gate is non-compensatory: one severe Vow cannot be averaged away by the others. Initial node verdicts are unknown until a real event occurs. Constitutional lens changes are prospective epochs.

Complex-systems feedback

The Alan A. Cohen corpus constrains the research design toward multivariate state, modular/context-sensitive response, perturbation tests, architectural constraints, and anti-unitary interpretation. The canonical probe produces 15 distinct channel/intensity response signatures and a nonzero context-dependent response difference. It does not claim biological equivalence, a privileged health/alignment scalar, critical-transition inference, or resilience: this v1.0 engine has no calibrated autonomous recovery-to-attractor process.

Executable checks

Research boundary

All probabilities, latent propensities, transition functions, risk coefficients, Vow tensions, thresholds, and network geometry are synthetic. The experiments test implementation semantics and internal invariants only; they are not evidence of empirical safety performance.

Direct active sources

All references. Bibliographic references are included in the public ZIP.

Recursive causal governance · v1.0

Recursive causal governance · v1.0