Propagation state varies across AI platforms because each platform has its own propagation timeline. A deployment may be in late-state propagation for one platform while still in early-state propagation for another, producing the cross-platform response variance commonly observed during the first weeks of deployment.
Documenting propagation state at evaluation time is essential for accurate methodology assessment. Snapshot evaluations during early-state propagation reflect the propagation state more than the methodology's steady-state performance.