H2-F: the framed transfer does not use the H2-sem no-target-feedback clause (F4 admissible; no new kappa)

research/r439-h2sem-cond2-framed-transfer/CERTIFICATE.json

@@ -0,0 +1,40 @@
+{
+  "status": "PASS_R439_H2SEM_COND2_FRAMED_TRANSFER",
+  "assignment_nonce": "2276dce171053917945d67a485ce77f3",
+  "task_id": "bit-helper-contract",
+  "scope_status": "refined",
+  "columns": 661,
+  "word_canonical_sha256": "5642fd0e35a1acc11acff8d0c08350c1617e95f310d4d35ad63a651f9eeeba0b",
+  "h2f_admission_f4_target_feed_wrap": {
+    "cond1_zero_frame_distinct_operands": true,
+    "cond2_no_target_feeds_helper_or_source": false,
+    "boundary_operator_exact": true,
+    "frames_at_cut_equal": true,
+    "semantic_K_equals_raw_K": true,
+    "h1_restricted_to_W0_and_C": true,
+    "conclusion_U_J_Uinv_flags_pass": true,
+    "inherited_flag_failures": [
+      "H1_no_target_feeds_z",
+      "H2_D_frame0_targets_from_helpers",
+      "K_equals_D"
+    ],
+    "compensated_a_rows_before_fed_gate": 67,
+    "d_source_reads": 18154
+  },
+  "predecessor_f4_record_matches_pinned_certificate": true,
+  "f5_source_feed_still_excluded": {
+    "raw_operator_exact": true,
+    "cond1_zero_frame_distinct_operands": false,
+    "cond2_no_target_feeds_helper_or_source": true,
+    "framed_replay_rejection": "climb not nested: register 0 from frame 2 (dim 1) to frame 0 (dim 0)"
+  },
+  "negative_and_positive_controls": [
+    "baseline word passes all inherited flags and the virtual-splice premises",
+    "F4 target-feeding wrap admitted under H2-F; only the global syntactic flags H1_no_target_feeds_z, H2_D_frame0_targets_from_helpers, K_equals_D fail, as pinned",
+    "N1 uncompensated feed rejected: 67 target columns moved by the fed target's dirt (first: 111)",
+    "N2 wrong compensation target rejected on 67 target columns",
+    "N3 late restore (span crossing the fed target's own gate) rejected; the fed helper's identity row breaks",
+    "F5 source-feeding wrap still rejected without condition 2: condition 1 and the framed replay's nested-climb check refuse it"
+  ],
+  "kappa_claim": "none"
+}

research/r439-h2sem-cond2-framed-transfer/README.md

@@ -0,0 +1,116 @@
+# H2-F: the H2-sem no-target-feedback clause is not used by the framed transfer (no new kappa)
+
+**Scoped interface admission on the pinned selected R439 word; no circuit pricing, no supplier adoption.**
+The selected result, the selected bit supplier and `upstream/` are unchanged.
+
+The merged package `research/r439-copyshare-h2sem-admission` (Aaron Wang) records its F4-shaped
+target-feeding-helper wrap as operator-exact but rejected only by PR1194's H2-sem condition 2, and
+leaves open whether that clause is needed at the framed level, with the exact use of H1 in the
+interleaved lemma named as the check to make. This package makes that check and answers it.
+
+## What this package establishes
+
+H1's role in the interleaved scalar-interface lemma (`research/e8-bit-r420/README.md`, the #543
+lemma inherited by the selected word's checker) is confined to the z-gate transvection induction
+over the gates of W0 = (A, scatter, B) and the cleanup C: "A z-gate reads only z registers
+(H1, H3), so it multiplies U_k on the left by an elementary transvection". The D section enters
+the lemma only as the operator y -> y + N s. Under the PR1194 substitution bridge the lemma is
+invoked on the virtual splice word, which keeps the original target-only D, and the actual D
+prefix is transferred by coefficient-matrix equality on all basis columns at the zero frame with
+identical state and frames at the D/A cut. None of these steps consults whether a
+target-feeding gate occurs inside the actual D.
+
+The instance check: the F4 wrap is rebuilt through the pinned predecessor machinery (its own
+unmodified module, its pinned certificate facts reproduced field for field), and every semantic
+premise and conclusion component passes:
+
+| component | F4 wrap |
+|---|---|
+| H2-sem condition 1 (zero frame, distinct operands from targets and helpers) | pass |
+| complete D operator on all 661 basis columns | exact |
+| state and frames at the D/A cut | equal to baseline |
+| semantic N = K | pass |
+| H1 restricted to W0 and C (byte-equal sections; baseline flags all pass) | pass |
+| U, J, Uinv conclusion flags of the unchanged checker | pass |
+| inherited flag failures | exactly H1_no_target_feeds_z, H2_D_frame0_targets_from_helpers, K_equals_D |
+| D source reads | 18154, as pinned |
+
+Only the three global syntactic flags fail — the same three recorded in the predecessor
+certificate. Removing condition 2 therefore does not change what the transfer verifies. F5-shaped
+source feeds remain inadmissible through condition 1 together with the framed replay's
+nested-climb legality (`climb not nested: register 0 from frame 2 (dim 1) to frame 0 (dim 0)`),
+which is re-checked here without condition 2.
+
+The protective work condition 2 was conservatively doing is shown to be carried by the semantic
+conditions, through three new negative controls:
+
+| control | outcome |
+|---|---|
+| N1 uncompensated feed (body rows keep reading the dirtied helper) | rejected: 67 target columns moved by the fed target's dirt, first column 111 |
+| N2 feed compensated by the wrong target | rejected on 67 target columns |
+| N3 restore moved past the fed target's own gate (span crossing it) | rejected: the fed helper's identity row breaks |
+
+## Reproduce
+
+Standard library only, about two minutes, from the repository root:
+
+```sh
+python3 -I -B research/r439-h2sem-cond2-framed-transfer/verify.py
+python3 scripts/research_package.py --package research/r439-h2sem-cond2-framed-transfer --check-only
+```
+
+Expected final line:
+
+```text
+PASS_R439_H2SEM_COND2_FRAMED_TRANSFER columns=661 compensated_rows=67 n1_columns=67 n2_columns=67 controls=6 no_new_kappa
+```
+
+The verifier pins every consumed source by SHA-256 (`SOURCES.json`), rebuilds the F4 wrap
+deterministically through the predecessor package's own unmodified module, replays all 661 basis
+columns through the unchanged `xcheck_p.py` owner checker, compares the inherited flag failures
+and read counts field for field against the pinned predecessor certificate, runs the three
+negative controls and the F5 exclusion, and compares the complete report against `CERTIFICATE.json`.
+
+## Scope and what is not claimed
+
+* **No pricing.** The F4 wrap is an operator-preserving admissibility probe (it adds 69 reads),
+  not a saving. Whether any target-feedback implementation beats its restricted alternative in a
+  priced model is the census side: open #2567 (at most eight registers, unit-cost single-source
+  gates) and discussions/1290, /1544. This package adds no evidence on savings either way.
+* **Admission, not supplier adoption.** No word is proposed for the selected chain; F/B positive
+  envelopes, the 104/192-bit guards, hosted costs and the paid profile were not recomputed for
+  any variant. The outer assembly, all-size compiler, hosting, precision/tape, recovery and
+  analytic obligations are untouched.
+* **Scoped to the transfer argument.** The result is about the necessity of condition 2 inside the
+  framed transfer for the interleaved-lemma conclusion, under this word's D-prefix equivalence
+  discipline. It does not restate the whole H2-sem contract for other architectures, other words
+  or nonzero-frame placements.
+* The chronological substitution bridge, the module argument and the semantic D contract are
+  PR1194's and are consumed unmodified; this package adds the necessity analysis and the instance
+  evidence.
+
+## Open next questions
+
+* Does a target-feedback wrap with a genuine positive saving exist in a priced model? Why it
+  matters: the enlarged admissible class only pays off if a priced construction uses it. First
+  step: extend the #2567 dirty-shear census's cost model with multi-operand compensated reads (the
+  F4 class is two-operand), beyond its single-source unit-cost gates. Blocked by: that census's
+  model and at-most-eight-register domain are #2567's scope; extending the model is a separate
+  priced search.
+* Should the inherited checker expose per-section H1 flags? Why it matters: consumers must
+  currently re-derive "H1 on W0 and C" from byte-equality of the non-D sections plus the baseline
+  flags, as this verifier does. First step: a backward-compatible optional per-section flag in
+  `xcheck_p.py`'s interface report. Blocked by: the checker is a pinned source for many merged
+  packages; any edit must be a separate reviewed change.
+
+## Credit and disclosure
+
+Prepared by Abhinav Ramachandran (geckods) with substantial assistance from a Hermes agent
+(Nous Research, GLM-5.3) under the standard donation prompt. Credit: **Aaron Wang,
+r439-copyshare-h2sem-admission** (the F4/F5 constructions, the pinned certificate and machinery
+reused unmodified), **woklerx, PR1194** (`research/r439-semantic-d-contract-seed10/PROOF.md`: the
+H2-sem contract and the chronological substitution bridge), **eumemic PR555** and
+**DreamingOfClouds PR589** (the R439 word), the R420/#543 interleaved-lemma lineage, and the
+pinned `xcheck_p.py` owner checker. Apache-2.0; no independent human review is claimed.
+Assignment nonce `2276dce171053917945d67a485ce77f3` (task `bit-helper-contract`, priority
+revision 5); claim and investigation thread: discussions/2684.

research/r439-h2sem-cond2-framed-transfer/SOURCES.json

@@ -0,0 +1,54 @@
+{
+  "schema_version": 1,
+  "base_commit": "b4560d9e7216f72b73c5aa940ac099c05e33ccf9",
+  "files": [
+    {
+      "path": "certificates/selected-result.json",
+      "usage": "provenance",
+      "git_blob": "ae6a99f28aa468649eb66a361e764d769363bf11",
+      "sha256": "10b9e340fd8e354a37a0c4813c90e5de1dd42044b5f6aca4a3fc5a42b685fa32"
+    },
+    {
+      "path": "docs/research/agent-task-catalog.json",
+      "usage": "provenance",
+      "git_blob": "889b95d936746df124872133ae1e7cc339adcb92",
+      "sha256": "7d185cd8c2af35988d638d969748ad7b670f902606444dbfc748e4402c340df6"
+    },
+    {
+      "path": "research/e8-bit-p111-tuned/code/xcheck_p.py",
+      "usage": "input",
+      "git_blob": "184f5dcf94a486cc7db91ca4db5865916bf26009",
+      "sha256": "af3decc877e27c00001cf56f50c6c09566e8e101297ab34c7860aed96a06d80e"
+    },
+    {
+      "path": "research/e8-bit-p111-tuned/tune/bitE8mut-p111-tuned.json.gz",
+      "usage": "input",
+      "git_blob": "85a4499bf9e4911a705d50567a2c2ef6cf23739a",
+      "sha256": "e3e301e62bc2b41dbf1ceac0acf3d35824ebeab1848c5a2142d4577b109da52f"
+    },
+    {
+      "path": "research/e8-bit-r420/README.md",
+      "usage": "input",
+      "git_blob": "6751eba6f617292fc919e58d5a9f9f06ab037391",
+      "sha256": "9807be57c6adf26b30e4f54e2f8160075bbc05091f1a6d157e43d2a4620c1c9f"
+    },
+    {
+      "path": "research/r439-copyshare-h2sem-admission/CERTIFICATE.json",
+      "usage": "input",
+      "git_blob": "a874a9840acd80474c92ca01e035929163bc08aa",
+      "sha256": "2678eaa3049cb208c6914be61cff93a72965b44e3abdf847c17595c9b7013594"
+    },
+    {
+      "path": "research/r439-copyshare-h2sem-admission/verify.py",
+      "usage": "input",
+      "git_blob": "e24e90390acc3f032041da98340288c21ad398b3",
+      "sha256": "09400f88e9e598bc6022ba32fcc6f9d1f048c5106b4d73b02a6f9f318be1c812"
+    },
+    {
+      "path": "research/r439-semantic-d-contract-seed10/PROOF.md",
+      "usage": "input",
+      "git_blob": "2707fdb0d2f34c1c6c4ea41c2a68e44904a3e4fd",
+      "sha256": "dfbca8c61a92fdf0db8af450dc4466849da37fc296e28b67f357e196e87327db"
+    }
+  ]
+}

research/r439-h2sem-cond2-framed-transfer/TASK.json

@@ -0,0 +1,67 @@
+{
+  "schema_version": 1,
+  "task_id": "bit-helper-contract",
+  "priority_revision": 5,
+  "base_commit": "b4560d9e7216f72b73c5aa940ac099c05e33ccf9",
+  "original_card": {
+    "id": "bit-helper-contract",
+    "question": "Can one precisely specified helper-mixing word satisfy a replacement for the inherited H2/K interface?",
+    "starting_points": [
+      "docs/research/research-family-status.md",
+      "research/r439-bsection-stars-seed01/README.md",
+      "research/r439-copy-path-sharing-barrier/README.md"
+    ],
+    "methods": [
+      "Derive a replacement interface lemma and price its implementation.",
+      "Seek an exact dirty-input or chronological-cut counterexample to the proposed replacement."
+    ],
+    "acceptance_gate": "A full boundary contract and paid profile; unchanged scalar counts alone are insufficient. A scoped local result may be retained without a complete supplier. Expanding it into further construction rounds requires the family-level budget and obligation gate in research-family-status.md; an optimistic passing relaxation alone is not acceptance.",
+    "first_step": {
+      "command": [
+        "python3",
+        "-B",
+        "research/r439-bsection-stars-seed01/verify.py"
+      ],
+      "expected": "PASS_R439_BSECTION_STARS",
+      "inspect": "Compare the two false H2/K flags with the unchanged paid profile in its README.",
+      "scope": "Positive control / source-bound orientation only; this is not a new construction or full cold replay."
+    },
+    "first_experiment": "Write one replacement H2/K implication for a single helper-star operation, then try to falsify it on dirty inputs.",
+    "bottleneck_link": "A legitimate replacement could admit a cheaper bit circuit; scalar savings alone do not alter the selected moment.",
+    "allowed_variations": [
+      "Choose one helper-star or quartet instance; narrow to one boundary implication."
+    ],
+    "scope_change_triggers": [
+      "Changing only source count or scalar addition count without an interface argument."
+    ],
+    "overlap_terms": [
+      "H2_D_frame0_targets_from_helpers",
+      "K_equals_D",
+      "helper-star"
+    ]
+  },
+  "scope_status": "refined",
+  "actual_scope": "Framed-level necessity of PR1194's H2-sem condition 2: proof-role extraction of H1 in the interleaved scalar-interface lemma, plus instance admission of the F4-shaped target-feeding-helper wrap on the pinned selected R439 word through the merged h2sem package's unmodified machinery, with three new target-class negative controls (uncompensated feed, wrong compensation target, span crossing the fed target's own gate).",
+  "scope_change_reason": "The card's headline experiment (one replacement H2/K implication for a helper-mixing word) is completed on main by research/r439-copyshare-h2sem-admission for helper-feeding wraps; this session takes that merged package's recorded second open next question (is condition 2 needed at the framed level) on the same word and machinery, which the card's acceptance gate admits as a scoped local result.",
+  "bottleneck_link": "The result does not itself change the selected moment; it enlarges the admissible wrap class for the selected bit word's D section, where the next priced bit constructions are selected, and closes a recorded open question on the inherited interface contract.",
+  "budget_minutes": 60,
+  "stop_rule": "Stop after the admission verdict (every semantic premise and conclusion component passes under the weakened contract, or one exact failing component on dirty columns is recorded) plus the target-class negative controls; no priced search and no census extension.",
+  "outcome": "completed",
+  "evidence": "CERTIFICATE.json: PASS_R439_H2SEM_COND2_FRAMED_TRANSFER; all 661 basis columns replayed through the unchanged xcheck_p.py; predecessor F4 record reproduced field for field (flags, reads 18154); conditional admission with controls=6 and no new kappa; reproduce via python3 -B research/r439-h2sem-cond2-framed-transfer/verify.py.",
+  "author": "Abhinav Ramachandran",
+  "ai_assistance": "Prepared by Abhinav Ramachandran (geckods) with substantial end-to-end assistance from a Hermes agent (Nous Research, GLM-5.3) under the standard donation prompt in autoresearch.md.",
+  "open_questions": [
+    {
+      "question": "Does a target-feedback wrap with a genuine positive saving exist in a priced model?",
+      "why_it_matters": "The enlarged admissible class only pays off if a priced construction uses it; otherwise the admission remains contract-level only.",
+      "first_step": "Extend the #2567 dirty-shear census's cost model with multi-operand compensated reads (the F4 class is two-operand), beyond its single-source unit-cost gates.",
+      "blocked_by": "That census's model and at-most-eight-register domain are #2567's scope; extending the model is a separate priced search."
+    },
+    {
+      "question": "Should the inherited checker expose per-section H1 flags?",
+      "why_it_matters": "Consumers currently re-derive H1 on W0 and C from byte-equality of the non-D sections plus the baseline flags, as this verifier does.",
+      "first_step": "A backward-compatible optional per-section flag in xcheck_p.py's interface report.",
+      "blocked_by": "The checker is a pinned source for many merged packages; any edit must be a separate reviewed change."
+    }
+  ]
+}

research/r439-h2sem-cond2-framed-transfer/VERIFY.json

@@ -0,0 +1,25 @@
+{
+  "schema_version": 1,
+  "status": "ready",
+  "python": "3.12",
+  "requirements": [],
+  "inputs": [
+    "research/e8-bit-p111-tuned/code/xcheck_p.py",
+    "research/e8-bit-p111-tuned/tune/bitE8mut-p111-tuned.json.gz",
+    "research/e8-bit-r420/README.md",
+    "research/r439-copyshare-h2sem-admission/CERTIFICATE.json",
+    "research/r439-copyshare-h2sem-admission/verify.py",
+    "research/r439-semantic-d-contract-seed10/PROOF.md"
+  ],
+  "commands": [
+    {
+      "argv": [
+        "{python}",
+        "-B",
+        "{package}/verify.py"
+      ],
+      "timeout_seconds": 600,
+      "checks": "SHA-256 source pins; exact 661-column framed replay through the unchanged xcheck_p.py; predecessor F4 certificate facts reproduced field for field; conditional admission of the F4 wrap under the weakened H2-F contract with only the three recorded syntactic flags failing; three target-class negative controls rejected (N1, N2, N3); F5 source feed still rejected by condition 1 and the nested-climb replay; frozen CERTIFICATE.json byte match."
+    }
+  ]
+}