Why we pivoted then

0267 closed F1.4's two-alphabet residue, the run that was going when F15 was added (#118, Alexis, 2026-11-04: "queued after 0267"). On the Priorities page, F15's table sits before the family tables with open rows (F1.13, F1.14, F6.1), so it is next.

What happened before

• 0253-0255 and 0267 decided every hand key schedule K4 fits, on one free alphabet and on two, by English against copies. In both, K4 sat inside its stand-ins (p 0.51, p 0.43).
• The pattern across F1, F6 and F13 is the same. Where the 24 known letters cannot decide a cell, the English decider finds a third to two thirds of plants, and K4 behaves like its copies. The cells are short of checks, not of compute.

• 117 found no published source for the two phrases, so the plaintext is Sanborn's own prose.

The change we believe is needed

Give the search more known plaintext from content rather than more compute. Read K1-K3 and Sanborn's statements as a model of what K4 says (F15.1). Write drafts under it (F15.2). Then measure, on held-out plants, whether the drafts as a prior find more plants than the 24 letters alone (F15.3, bar: at least 2 more of 12) before any K4 run.