Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
9 changes: 6 additions & 3 deletions docs/REPRODUCE.md
Original file line number Diff line number Diff line change
Expand Up @@ -69,9 +69,12 @@ Gaussian-process short-bar statistics. Those runs need the pinned
`test_numeric` (six numeric smoke tests) skips without those modules. The packet
makes no claim of progress toward the Riemann Hypothesis, and its §9 records
independent numerical reproduction (review slice C) as only partly done: a
different-provider replay of its 59 commands is performed (record PASS, scoped),
but the different-seed or independent Riemann–Siegel check, including the Edwards
coefficients, is still open.
different-provider replay of its 59 commands is performed (record PASS, scoped).
Additional fresh-seed evidence is reported (record applicability PASS, scoped),
not independently verified. Independent verification of that evidence and the
Edwards coefficient/normalization comparison remain open; neither record closes
all of slice C. The packet's §9 links the source reports, reviewer lineage and
remaining limits.

### Math- packages and the hardening branch

Expand Down
2 changes: 1 addition & 1 deletion docs/RESEARCH_INDEX.md
Original file line number Diff line number Diff line change
Expand Up @@ -20,7 +20,7 @@ Checked on 12 October 2026 against main `8977619e86d2f395b8e4947cf9a5e34082cc2c2

6. **Formal verification: Lean statements with kernel checks, alignment separate.** The [cap-first-exit packet](https://github.com/d6g8k5htny-coder/Math-/blob/79cb314d5b9f6b8b410bfd0df200414314f6b8c7/frontiers/cap_first_exit_lean_20261002/README.md) (author Anthropic Claude) has 152 Lean targets in its [manifest](https://github.com/d6g8k5htny-coder/Math-/blob/79cb314d5b9f6b8b410bfd0df200414314f6b8c7/frontiers/cap_first_exit_lean_20261002/MANIFEST.json). They formalize the deterministic cap step of D1 component III (CAP §§2–5, as steps L1–L13 and L16–L22), which identifies `S` as the global elder partner of `M` on the good event; a green run is kernel evidence for those Lean statements only. Outside it: whether the field satisfies the source setting, the reading of source coordinates as Lean objects, persistence-module bookkeeping, steps L14, L15 and L23, P §8, and every probabilistic statement. The manifest claims neither an alignment review nor mathematical acceptance; [ALIGNMENT.md](https://github.com/d6g8k5htny-coder/Math-/blob/79cb314d5b9f6b8b410bfd0df200414314f6b8c7/frontiers/cap_first_exit_lean_20261002/ALIGNMENT.md) records the earlier scoped read of the v1.4 statements and what a non-Claude review must cover. Main's own formal layer is under [Open work](#open-work). Read after that check, at Math- `a0323b94`: the statement-only [Lean specifications](https://github.com/d6g8k5htny-coder/Math-/blob/a0323b94ae54a78f3d2878dc9c358dc40cdfe1dc/specifications/lean_statements_20261010/README.md) of registered claims and their [proof packet](https://github.com/d6g8k5htny-coder/Math-/blob/a0323b94ae54a78f3d2878dc9c358dc40cdfe1dc/specifications/lean_statement_proofs_20261011/README.md) (author Anthropic Claude) landed through Math- [#419](https://github.com/d6g8k5htny-coder/Math-/pull/419), [#422](https://github.com/d6g8k5htny-coder/Math-/pull/422), [#423](https://github.com/d6g8k5htny-coder/Math-/pull/423), [#434](https://github.com/d6g8k5htny-coder/Math-/pull/434) and [#439](https://github.com/d6g8k5htny-coder/Math-/pull/439) on 12 October, after the Math- commit checked above; the packet holds 75 theorems, each kernel-checked by a hosted pull-request run, of 74 of the register's 141 Props, including `FullTransformedPriceBudget`, the Prop the register gives for Theorem F's display (F2) (see [P15 combinatorics](#p15-combinatorics)), with alignment PENDING_INDEPENDENT_REVIEW, organizational-independence credit 0 and scientific effect NONE; each theorem is evidence about its Lean statement only and establishes neither Theorem F nor any other parent claim, and the register's labels are unchanged (every Prop and structure of the statement packet stays `specified` and no claim-level label changes; the register adds the packet's theorems as entries of kind `theorem` labelled `proved`).

7. **Zeta-zero heat-flow folds and a CUE short-bar control: exploratory packet; scientific effect NONE.** The [zeta fold exploration packet](../experiments/zeta_fold_exploration/README.md) (author Anthropic Claude), landed through [main #373](https://github.com/d6g8k5htny-coder/main/pull/373) after the main commit checked above, checks an external note's proposed links between zeta zeros and the project's short-bar law: near the collision time, nondegenerate zero collisions of the de Bruijn–Newman heat flow follow the generic cubic fold, and a finite-window short-bar statistic of a Hardy-Z primitive is compared with CUE and one-dimensional Gaussian controls. It makes no claim of progress toward the Riemann Hypothesis. OpenAI/Codex nonauthor reads of its slices A/D (mathematics, wording) and B (citations) returned AMEND, and their delta readbacks returned scoped PASS; for independent numerical reproduction (slice C), a different-provider replay of its 59 commands was performed on 20 October ([main #458](https://github.com/d6g8k5htny-coder/main/pull/458)) with a scoped record PASS, but the different-seed or independent Riemann–Siegel check is still open.
7. **Zeta-zero heat-flow folds and a CUE short-bar control: exploratory packet; scientific effect NONE.** The [zeta fold exploration packet](../experiments/zeta_fold_exploration/README.md) (author Anthropic Claude), landed through [main #373](https://github.com/d6g8k5htny-coder/main/pull/373) after the main commit checked above, checks an external note's proposed links between zeta zeros and the project's short-bar law: near the collision time, nondegenerate zero collisions of the de Bruijn–Newman heat flow follow the generic cubic fold, and a finite-window short-bar statistic of a Hardy-Z primitive is compared with CUE and one-dimensional Gaussian controls. It makes no claim of progress toward the Riemann Hypothesis. OpenAI/Codex nonauthor reads of its slices A/D (mathematics, wording) and B (citations) returned AMEND, and their delta readbacks returned scoped PASS; for independent numerical reproduction (slice C), a different-provider replay of its 59 commands was performed on 20 October ([main #458](https://github.com/d6g8k5htny-coder/main/pull/458)) with a scoped record PASS; additional fresh-seed evidence is reported (record applicability PASS, scoped), not independently verified. Independent verification of that evidence and the Edwards coefficient/normalization comparison remain open; neither record closes all of slice C.

The [public-text catalog](PUBLIC_MATHEMATICS.md) counts 2,138 files at its declared source commits: a file count, not a count of distinct theorems.

Expand Down
8 changes: 5 additions & 3 deletions docs/WORKSPACE.md
Original file line number Diff line number Diff line change
Expand Up @@ -18,9 +18,11 @@ As of 12 October 2026, new work lands through pull requests to this repository's
the last an exploratory packet whose scientific effect is NONE, which makes no
claim of progress toward the Riemann Hypothesis and whose independent numerical
reproduction (review slice C) is only partly done: a different-provider replay of
its 59 commands is performed (record PASS, scoped), but the different-seed or
independent Riemann–Siegel check is still open. None of the three reports a held-out
confirmation run; each README states its own scope. Two older branches remain
its 59 commands is performed (record PASS, scoped). Additional fresh-seed evidence
is reported (record applicability PASS, scoped), not independently verified.
Independent verification of that evidence and the Edwards coefficient/normalization
comparison remain open; neither record closes all of slice C. None of the three
reports a held-out confirmation run; each README states its own scope. Two older branches remain
readable context. The legacy [hardening research branch](https://github.com/d6g8k5htny-coder/main/tree/a01c72f1937875b94bfd78e183a6dbbdb2b04da4) (`chatgpt/drive-github-hardening-20260919`, frozen; pinned tip `a01c72f1`)
holds the earlier research software and evidence; its last merge was on
27 September 2026 (`a01c72f1`). The
Expand Down
6 changes: 3 additions & 3 deletions experiments/zeta_fold_exploration/MANIFEST.json
Original file line number Diff line number Diff line change
Expand Up @@ -9,7 +9,7 @@
},
"generated_by": "make_manifest.py",
"note": "committed_files lists every file meant to be committed (this manifest excluded). raw_files are regenerable samples written to results/raw/ and not committed; their hashes are from the exploratory run; regeneration is expected to be byte-identical only with OMP_NUM_THREADS=1 OPENBLAS_NUM_THREADS=1 (README section 8).",
"committed_total_bytes": 547335,
"committed_total_bytes": 549137,
"committed_files": [
{
"path": "CLAIMS.md",
Expand All @@ -18,8 +18,8 @@
},
{
"path": "README.md",
"bytes": 100470,
"sha256": "ccbe69d64ad16c0dbf0c7bf7e9e64b0896d3000ba426782c001b0c8598f87c66"
"bytes": 102272,
"sha256": "b985f3528427ea15ce8b0704d2f207669cad86a8859b89da96ccacb418bade06"
},
{
"path": "dbn/check_collision.py",
Expand Down
30 changes: 26 additions & 4 deletions experiments/zeta_fold_exploration/README.md
Original file line number Diff line number Diff line change
Expand Up @@ -12,8 +12,10 @@ de Bruijn–Newman constant Λ beyond known results. None should be read into it
slice B **PASS (scoped)** (review 5487692150). Integrated into main as `3e589ae0` on
12 October 2026 by a third OpenAI lane, distinct from both reviewers (§9). Automatic Codex
review: **NOT_RUN** (quota). Slice C (independent numerical reproduction): different-provider
replay of the packet's 59 commands **PERFORMED; record PASS (scoped)**; the different-seed or
independent Riemann–Siegel check is **still open** (§9). Dylan's personal reading: **PENDING**.
replay of the packet's 59 commands **PERFORMED; record PASS (scoped)**. Additional fresh-seed
evidence is **REPORTED; record applicability PASS (scoped)**, not independently verified;
the Edwards coefficient/normalization comparison remains **OPEN**, and slice C is not
Comment thread
d6g8k5htny-coder marked this conversation as resolved.
Resolved
Comment thread
d6g8k5htny-coder marked this conversation as resolved.
Resolved
fully closed (§9). Dylan's personal reading: **PENDING**.
- Object: `ZETA-FOLD-EXPLORATION-20261012`, draft v2 as read back at `0969b327` and landed at
`3e589ae0` (README blob `b367aecd`; v1 carried the same-provider passes, v2 adds the slice A,
B and D amendments, §10), plus post-integration edits that neither readback covers: the §6.2
Expand Down Expand Up @@ -1211,10 +1213,30 @@ A nonauthor read from a non-Anthropic provider was requested, in four slices:
14 were timing-only, two residual trails differed, and all nine raw-sample hashes matched.
The replay re-executed the packet's own `mpmath.siegelz`/`zetazero`/`nzeros` comparisons with
the packet's same fixed seeds. It added neither a different seed nor an independent
Riemann–Siegel check beyond the packet's existing validation, so that part of slice C,
including the Edwards coefficients left to it above, remains open. See the
Riemann–Siegel check beyond the packet's existing validation. See the
[replay delivery](https://github.com/d6g8k5htny-coder/main/pull/458#issuecomment-6282286422)
and exact-head nonauthor [PASS_RECORD_SCOPED review](https://github.com/d6g8k5htny-coder/main/pull/458#pullrequestreview-5581698876).

**Additional fresh-seed evidence REPORTED; record applicability PASS (scoped).**
Separately, [OpenAI GPT-6's R2a report](https://github.com/d6g8k5htny-coder/main/pull/451#issuecomment-6279371506)
records a separate summation implementation with seed **31415927**, 192 points against
`mpmath.siegelz`, and four fresh zero-count windows at coarse and half-step grids agreeing
with `mpmath.nzeros`. The nonauthor [record-applicability assessment](https://github.com/d6g8k5htny-coder/main/pull/458#issuecomment-6289356717),
performed by OpenAI Codex `/root`, Capability Frontier observer (CAPFRONT), was source-exposed,
not blind: the reviewer was not the packet author or R2a executor, but shared R2a's OpenAI
provider and the project's GitHub principal. That assessment
accepts this as reported additional different-seed evidence for this component, not a
new execution or independent verification of those numerical outputs. R2a used
Comment thread
d6g8k5htny-coder marked this conversation as resolved.
Resolved
Python 3.13.5 / numpy 2.3.5 / mpmath 1.3.0, not the pinned environment, and did not execute
the exact `rs_zeta.py` bytes. Its raw JSON was not uploaded with the report or retrieved
by the applicability reader; its reported hash and results were not independently
verified there, and custody elsewhere remains unknown. This evidence has organizational-
independence credit 0 and supplies no uniform error bound, interval certificate,
zero-completeness proof or scientific acceptance. The **Edwards 1974 coefficient and
normalization comparison remains OPEN**: reimplementing the quoted formula does not
verify it against Edwards. The replay's two residual-trail discrepancies and 13
unretained timing-only output streams also remain unresolved; neither record closes all
of slice C.
- **D.** A wording audit of the non-claims. **Read: AMEND** (the same review, §10.1). **Delta
readback: PASS (scoped)** (review 5487707101, with slice A).

Expand Down
Loading