Why we pivoted then
Every item in F6's table is closed or blocked. F12 is third in the ranking (added 2026-10-16, issue #53), after F6 and before F3; the ranking is unchanged.
What happened before
- 0158: the mixed triples of matrix readings at wide widths, 105.5 million routes, enumerated exactly in compiled code (F6.3b).
- 0159-0163: compass directions as reading routes, move-change-move, two keyed stages on different tables round a move, triple routes with a keyword columnar stage, and autokeys with the width-7 double columnars: each at the scrambled copies' rate.
- 0164-0167: every keyword transposition the field manuals and the agent-cipher literature describe that keeps letters one for one (disrupted, triangle, interrupted, trapezoid, keyed blanks, row reshuffles, French diagonal, EVA, Chiffre XIII as a parametric family, the Nihilist square with all 10! keys, YAR's 1-2-3): none fits K4 better than scrambled copies; Chiffre XIII's published substitution fits no mask at all.
- F6.1 and F6.2 (free alphabets after a move) remain blocked, now by cost: F1.5's decider would need about 20-30 minutes for each of hundreds of thousands of fitting settings.
The change we believe is needed
- F12 next, in its table's order. It is a way of working, not a cipher: look at the copper, quantify the impressions against shuffles, then crib. F12.2 and F12.4 produce observations, not verdicts; an observation that stands outside the shuffles becomes a new item.
- The compiled route tester built for F6 (any list of moves against a periodic key on the sculpture's tables, K4 and copies) is ready for any route an observation suggests.