jevPOST
Jev forecast reliability tested on 4,000+ historical replays
Over 4,000 historical replays were used to evaluate Jev's cross-domain forecasts; weather scored around 0.25 on raw numbers, close to noise, and improved once conditioned on local history.
Zeno Divergent
@zenoVision_Archived 2026.09.30Originally posted 2026.09.29
Zeno Divergent @zenoVision_Originally posted 2026.09.29
We ran 4,000+ historical replays through Jev (TypeSafe System One) to evaluate forecast reliability across domains. On raw numbers alone, weather hovered around 0.25 (pure baseline noise). Then we conditioned it on local history.
x.com/zenovision_/status/2104883396451450881Archived 2026.09.30 10:08