Research: prefix-bounded signed-Horner circuits for CK and JP8 (no new kappa)

research/complex-prefix-bounded-dyadic/CLAIM.json

@@ -0,0 +1,95 @@
+{
+  "status": "completed; research PR submitted for review",
+  "assignment_seed": 23,
+  "randomness": "Python secrets.randbelow(100), OS entropy; drawn once at session start",
+  "track": "complex supplier",
+  "base_commit": "66849e1475c930fd7732b1a0067ffee7e1e95f60",
+  "started_utc": "2026-10-13T04:37:48Z",
+  "expected_finish_utc": "2026-10-13T05:18:00Z",
+  "verification_reserved_from_utc": "2026-10-13T05:08:00Z",
+  "selected_kappa": "51110230636731/31250000000000000",
+  "candidates": [
+    {
+      "task": "Reorder one raw K-triangle read window with exact minimal frames",
+      "decision": "defer: PR687 already closes the closest narrow relocation class; seed change alone adds little"
+    },
+    {
+      "task": "Resynthesize one fixed coordinate-total factory map with restoration",
+      "decision": "defer: PR681 already supplies exact fixed-map elimination; another block alone adds little"
+    },
+    {
+      "task": "Prefix-bounded signed-digit Horner for selected complex dyadic additions",
+      "decision": "chosen: changes coefficient synthesis; PR696 only replaces leading clear/add and unit macros"
+    }
+  ],
+  "seed_use": "23 mod 3 = 2 selects third candidate; 23 mod 2 = 1 prioritizes JP8 backward for screening; final coverage includes both providers and orientations",
+  "object": "Dyadic addition coefficients in the pinned PR626 CK and JP8 LOCAL-PROGRAMS; unchanged scalar boundaries, frames and paid recursive profile",
+  "method": "Finite signed-binary Horner recurrence with every intermediate integer coefficient in [1, abs(numerator)]; compare explicit cost and exact prefix envelope against binary Horner and PR696 copy/unit lowering",
+  "expected_deliverable": "Standalone circuit, exact rational verifier, finite deterministic certificate, adverse controls and scoped proof; no new kappa claim",
+  "stopping_rule": "Complete all distinct selected coefficients and four selected scalar event streams; stop if no further fully charged primitive saving or if the inherited prefix envelope cannot be retained",
+  "coordination": {
+    "pinned_issue": null,
+    "viewer_permission": "READ",
+    "checked_utc": "2026-10-13T04:41:00Z",
+    "local_claim_reason": "No pinned coordination issue exists; no reservation issue opened",
+    "open_pr_count": 533,
+    "latest_pr": 714
+  },
+  "predecessor_prs": [
+    411,
+    622,
+    626,
+    696
+  ],
+  "initial_screen": {
+    "additional_real_primitives_saved_vs_pr696": {
+      "ck/forward": 122,
+      "ck/backward": 122,
+      "jp8/forward": 3090,
+      "jp8/backward": 3090
+    },
+    "status": "exploratory counts only; verification pending"
+  },
+  "attribution": "Chris Pagolu (@AllStars101-sudo), with substantial OpenAI Codex assistance; Apache-2.0",
+  "prior_negative_result": {
+    "path": "research/experiment-log/campaign-100/algebraic-word-construction.md#38-signed-digit-coefficients",
+    "contributor": "Rohan Arun",
+    "source_head": "fa8ab946df57ebadae1a470b637c2a590a0a5317",
+    "finding": "Coefficient-3 rewrites at rank zero do not improve paid climbs",
+    "distinction": "This contribution changes primitive synthesis for the later selected large numerators, retains the exact intermediate coefficient envelope, and claims only scalar cost savings; it agrees that the paid profile and kappa do not improve."
+  },
+  "completed_utc": "2026-10-13T05:00:13Z",
+  "implementation_commit": "21606831df4b08f79d87fd420c558a520dca571e",
+  "pull_request": "https://github.com/CrocSwap/integer-mult-bounds/pull/734",
+  "result": {
+    "classification": "finite complex-supplier circuit improvement / research note; no new kappa",
+    "saved_real_primitives_beyond_pr696": {
+      "ck/forward": 122,
+      "ck/backward": 122,
+      "jp8/forward": 3090,
+      "jp8/backward": 3090
+    },
+    "coverage": "All 686 selected dyadic coefficient types and four complete scalar programs; 716 native + 9 private formal columns each; 8256 independent digit-graph comparisons; 12 mathematical controls and four package/algorithm corruption controls. No search timeout or unfinished case.",
+    "prefix_bound": "Original per-macro numerator cap and inherited exact global scalar envelope retained",
+    "verification_started_utc": "2026-10-13T04:52:04Z",
+    "reproduce": "python3 -B research/complex-prefix-bounded-dyadic/verify.py",
+    "expected": "PASS_PREFIX_BOUNDED_DYADIC programs=4 columns=716+9 coefficients=686 extra_savings=122,122,3090,3090 controls=12",
+    "limitations": "No production emitter integration, new global invoice/admission, full theorem proof, new kappa or practical runtime claim."
+  },
+  "concurrent_result": {
+    "pr": 723,
+    "head": "d3bc41b87fc7d23f50d064a0f06f1a89e3ab02a7",
+    "credit": "Sagar Thakrar independently developed JP8 signed-digit lowering during this session",
+    "distinct_evidence": "Hard per-macro numerator cap with unchanged exact prefix envelope; optimal recurrence for this restricted class; separate forward graph check; CK coverage and explicit composition with pinned PR696. PR723 permits numerator+1 and saves 186 more recoding primitives per JP8 orientation. Its implementation is not independently verified here."
+  },
+  "final_overlap_refresh": {
+    "utc": "2026-10-13T04:59:42Z",
+    "latest_other_pr": 732,
+    "related_heads_unchanged": [
+      696,
+      722,
+      723
+    ],
+    "coordination_issue": null
+  }
+}