T002 STRUCTURE / EXCEPTION CONSTRAINTS HANDOFF

Most useful independently reviewed findings

1. Without guessed plaintext, the26-ordinary-gap interval relaxation needs exactly88 exceptional events among1036 observed outputs. The lower bound uses an exact per-message parity count, and an explicit88-event timing witness attains it. This is a necessary relaxation, not a valid full encryption key. Claims C001/C002; review T003-R003.

2. A single hypothesized equal plaintext action, W1[38]=W1[68], sees consecutive ciphertext return gaps8 and9. Their allowable endpoint-action sets intersect for exactly35 of3321 fixed off-top exception pairs. No ordinary-alphabet bound or shared initial deck is needed. Adjacent pairs(a,a+1), a=1..7, and pairs b-a>=75 are the complete survivor description. Claim C005; review T003-R009. A short delta-based classification proof is also recorded by the reviewer in T003-E010/analytic_classification.txt.

3. The parent complete conditional M002 exclusion needs only two29-action context equalities plus that single W1 link. The other16 equalities in the W1 self-context are unnecessary. Claim C007 is independently reviewed by T003-R014; fresh35-pair replay and four complete final boundary branches.

Latest independently reviewed result

4. Both long contexts can be reduced to27 actions: E4[70:97]=W4[73:100]=E5[71:98], plus W1[38]=W1[68]. All35 pairs are excluded with independent arbitrary starting decks and arbitrary nonzero ordinary action slots. Claim C008 is independently reviewed by T003-R017. The fresh independent replay excluded all35 pairs in20.68 seconds and retained39,996 true positions across12 planted models, including all four boundary assignments. Producer census is in repaired_census.json. These ranges are zero-based and end-exclusive. The equal-action assumptions are not established plaintext facts, and no minimum-context claim is made.

Failed compression is preserved

An18-action triple context excluded the final pair(1,82), but a complete census left pairs(1,80) and(3,82). Both are relaxed fixed points only, not encryption keys. The final27-action context restores enough earlier constraints to exclude them. The compression search was adaptive, and no mathematical minimum-context claim is made.

Other records

E006's special(1,82) event/endpoint relaxation has a rational LP certificate proving at least98 exceptions and an independently producer-verified100-event witness, hence a98..100 optimum bracket. Its bounded MILP timed out; there is no exact-optimum claim. The full conditional model exclusion supersedes this branch for the stated plaintext contexts.

E003/C004 preserve the older30/30/18-action hypotheses. D003 explicitly corrects these: a30-observation ciphertext isomorph naturally motivates29 subsequent shared actions. Old findings must not be silently described as applying under the corrected weaker hypotheses. E005 is the corrected endpoint census.

All8 T002 experiments have plan/result records. The16 records passed a required-field check. No processes remain active. No further compression searches are planned.
