Quarterly, in writing.
Oct 1, 2026 We were wrong in August. Less wrong by September.Our posterior on the first frost narrowed from ± 14 days to ± 4 as the weather data arrived. We buried accordingly. Latest
Q2 2026
Jul 2, 2026 On the base rate of wintersWinter has arrived every year in our dataset. We have chosen not to treat this as a coincidence. Buried
Jul 2, 2026 On the base rate of wintersWinter has arrived every year in our dataset. We have chosen not to treat this as a coincidence. Buried
Q1 2026
Apr 1, 2026 We dug up the acorns. Then we buried them again.A note on February, what the model saw, and why we did not override it. Buried
Apr 1, 2026 We dug up the acorns. Then we buried them again.A note on February, what the model saw, and why we did not override it. Buried
Q4 2025
Jan 5, 2026 A prior is not a promiseOur first full year: one winter survived, and a few things we would do differently. Buried
Jan 5, 2026 A prior is not a promiseOur first full year: one winter survived, and a few things we would do differently. Buried
Older letters were buried for safekeeping. We know roughly where.