live · v0.26

A harder test of when correction earns its place.

Three model variants, three independent seeds, and one global decision-time split. The campaign starts from a strong forecast and asks when Zeno should preserve it—or improve it.