Why we pivoted then
D1 cannot go further without a decision. Design 2's power for one free alphabet is flat in the number of runs, so one run is within sampling noise of ten, but the pre-registered rule asks for power "at least" the ten-run power, which a noisy comparison meets only at ten runs (0229). Whether a stated margin is allowed is Alexis's call (#100). D2 waits on D1. The queue's next table is F13 (#85), whose first item is a desk count.
What happened before
- Stage 2, step 4 closed: A6.2 (0226) and A6.3 (0227). Sanborn-style plaintext costs the climbs little; masks that dilute the message (an X every kth letter, the message in every nth place) are not found at power.
- D1's two measurements (0228, 0229) redrew design 2's power with more plants and found it under the bars that closed F1.1's Beaufort 23 and 26 and F1.2's 13. Those cells are open and wait for Alexis (#100).
The change we believe is needed
- F13.1: count, for each family, the key bits against what 97 letters of English can decide with no known letter, so that F13.4 runs only where English alone can decide.