# Job 2612 pre-registration (route 100), written before any sweep row runs Object: A = K*(P,R), B = K*(Pp,R), drop = A - B (#1267 definitions; instrument kstar.c from #1812, gated on #1267 (30,11,{7,13,19,23}) -> 12,10 and #1246 (30,13,{7,11,19}) -> 9,8 before this file). Bound under test: D(P,p,A) = max over start phases of min over translates (dbound2612.py); drop <= D is proven. Population (sweep2612.py): P=2, R subsets of primes 3..43 with 2*prod(R) <= 2e8; P=6, R subsets of primes 5..47 with 6*prod(R) <= 5e8; |R| >= 2. For every R: A. Then for every prime p <= 61, p !| P, p not in R, with D(P,p,A) >= 2 and P*prod(R)*p <= 1e10: B. G (guard, stop): any drop > D => instrument or proof error; report, do not interpret. S (success): a row with drop = D >= 3 (first drop-3 witness), cross-checked by kfork.c. Record: per shape (P,p,A,D) the maximum observed drop; shapes with D>=2 attained vs not attained in scope. Not attained in scope is a measurement over this population only, never a bound. ## Amendment 1 (before any B row completed) Attempt 1 (MPMAX 1e10, all D>=2 rows in one batch: P=2 gave 16,292 B rows) was stopped by me after 10 min with no row output (nothing read). New plan: rows with D>=3 and Mp <= 2e9, plus rows with D=2 and Mp <= 1e8; all rows sorted by ascending Mp, run in chunks, each chunk's rows appended to sweep2612.jsonl; wall budget 20 min; report the largest Mp completed per base. Hypotheses G and S unchanged.