Dynamic Causal Geometry Predicts Finite-Propagation Attractor Sensitivity

When timing changes the outcome

A delayed message does not always change a system's eventual behavior. This study investigates whether the influence of information at a particular moment predicts when delay matters better than a fixed map of network connections.

The work compares a dynamic causal predictor with a static baseline in recurrent-network simulations. It reports a held-out predictive advantage in its confirmatory study, measuring prediction quality rather than faster computation or a general reduction in AI errors.

Status and next steps

This is an independent technical preprint, version 0.9.1. The result is simulation-based and has not been peer reviewed. Physical validation remains pending, and the publication distinguishes research-phase numerical summaries from independently audited archival measurements.

The broader question connects to distributed systems and physical computing: how should systems place and communicate information when propagation itself becomes part of their behavior?