probes(D-TFLP 6.5): Codec8 semantic admission + DT64/Quack/D-NAM handoff by AdaWorldAPI · Pull Request #1801 · AdaWorldAPI/lance-graph

AdaWorldAPI · 2026-10-26T20:41:08Z

Why

Amendment 6.4 F4: a Codec8 payload is a bare &[i8]. The producer's code space lives only in its Rust type, so an asymmetric −128 can be decoded as SymNaN's NaN by the wrong consumer. The fix is not in the decoder. It is admission at preparation time: compare a receipt with what the consumer expects before any value is interpreted. This PR lands that lab contract and a short note on where it would attach in DT64, Quack and D-NAM.

Scope (one commit)

.claude/probes/pal256-admission-v1/admit/: Receipt, Expect, bind with named Refusals, Bound::decode; 17 tests; the matrix example.
matrix65.json (6 × 6, diagonal-only binding) and disable65_p0.txt.
README.md and HANDOFF.md, the compact integration document.
One board entry, entries/2026-10-26-d-tflp-65-landing.md, with the generated index. SUPERSESSION-INDEX was regenerated and is unchanged.

Landing-PR changes:

the identity claim is narrowed to TESTED_COMPATIBLE_FOR_PINNED_CODEC_SET;
a new test checks that all 21 Codec8 instantiations admitted on ndarray master af4910a1 have pairwise-distinct signatures.

Evidence used

Research branch claude/bpe-tokenization-architecture-3xd4eh (Amendment 6.5 P0; not rebased). ndarray src/hpc/codec8.rs on master af4910a1, identical to the research-time version.

Tests run

cargo test --release against ndarray master: 17 passed.
matrix65.json regenerates byte-identical.
Disable run: collapsing the rounding probe to one half-tie fails the scope test (TiesEven/Floor collide).
The earlier P0 disable runs are recorded in disable65_p0.txt, each check failing its own test.

What this does not prove

No identity beyond the 21 pinned instantiations. A new codec impl must be added to the scope test.
No binding store, no generation-identity issuer, no contract-type change.
No source permissions or writes, and no semantic identity from a digest.

Production impact

None. Lab crate, not a workspace member. It path-depends on ndarray five levels up. Codec8 public semantics and the −128 decoding policy are unchanged. No second binding store is added: D-NFR S5 stays MISSING.

Dependencies / rollback

No dependency on #1800; the handoff doc only references it. Rollback: revert the single commit.

Links

Companion: probes(D-TFLP 6.5): softmax fixture replay + certified preparation gate #1800 (softmax replay + certified preparation gate).
DT64: docs(DT64-0): versioned DTO predicate adapter capstone, staged PR board and EXO-50 gates #1799 (DT64-QB), stages docs(DT64-1): 64 typed predicate mappings for IAM Odoo SAP #1794–docs(DT64-5): native DTO behavior and ONNX/time-series lowering gates #1798.
D-DR256-WHERE: capstone, DR64³ inventory, P2 Quack-path lab, evidence reconciliation, D-NFR-1 ownership correction #1792, board: post-merge records for #1788–#1792 #1793 (merged).
D-NAM capstone §2 / PZL-4; D-NFR S5 (d-native-repr-contract-v1.md).
EXO-152, EXO-50: BOARD_SYNC_PENDING, because Linear is not available in this session.

Generated by Claude Code

A lab binding contract between an 8-bit artifact and the consumer that interprets it: a Receipt (codec signature derived from Codec8's sealed traits, scale/zp, operator, generation, source, payload digest), an Expect, and bind() that refuses with a named Refusal before any value is interpreted. 17 tests; 6x6 matrix binds on the diagonal only. Imported from claude/bpe-tokenization-architecture-3xd4eh (6.5 P0). Changes made while landing: - the identity claim is narrowed to TESTED_COMPATIBLE_FOR_PINNED_CODEC_SET and backed by a new test: all 21 Codec8 instantiations admitted on ndarray master af4910a1 have pairwise-distinct signatures (disable run: collapsing the rounding probe to one half-tie fails it); - README and HANDOFF.md: where the receipt would attach (D-NAM recipe receipt / PZL-4, D-NFR S5 binding store - MISSING, DT64-QB #1799) and what stays open; - compact board entry for the 6.5 landing (#1800 and this PR) with the generated index. No crate, ABI, kernel or decoder change. matrix65.json regenerates byte-identical. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01ArVbbq3DsToBM7r79zGeEf

coderabbitai · 2026-10-26T20:41:16Z

Important

🔍 Trigger review

This repository does not receive automatic reviews because it has fewer than 10 stars.

Run configuration

Source mode: Auto
Repository: Unavailable
Branch: Unavailable
File: No configuration file from this repository applied
Commit: Unavailable
Layers: Organization UI
Review profile: CHILL
Plan: Advanced
Run ID: c1f5f863-4db5-4354-bd73-90962191cd3d

Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

Warning

Billing warning: we have not been able to collect payment for this subscription for more than 72 hours. Please update the payment method or pay any pending invoices in Billing to avoid service interruption.

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

_{Comment @coderabbitai help to get the list of available commands.}

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01ArVbbq3DsToBM7r79zGeEf

chatgpt-codex-connector · 2026-10-26T20:43:21Z

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-10-26T20:44:03.739519Z	`01ef507`	Draft marked ready

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

PR_ARC_INVENTORY: the #1800 Confidence line said only "MEASURED (lab)". It now records the certification: every fixture row confirmed by the real TFLM int8 SOFTMAX kernel (5f64e4ad), the TFLM-free Rust replay reproducing those bytes on all 529 + 27 + 7 rows, the exact-reciprocal negative control failing all 556 witness rows, and the scope stated in the probe's own README. A dated correction line notes that the replay crate landed with #1800, not #1801; #1801's Confidence line now names its admit/ contract and points to #1800 for the certification. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01VncKoyG6JkpsoqhqTGTv3f

Brings the research branch up to main eeaaafb (136 commits) without rewriting its history: the audit and ledger cite this branch's hashes. Conflicts were all files that landed on main in hardened form through #1800/#1801/#1803 (table-identity check, compact data, pinned-codec-set claim); main's versions were taken. The research-only run record softcert65_run.json stays here as evidence. The audit entry heading now reads D-TFLP-AUDIT so its index id no longer collides with main's D-TFLP 6.5 landing entry; the entries index is regenerated. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01ArVbbq3DsToBM7r79zGeEf

AdaWorldAPI mentioned this pull request Oct 26, 2026

probes(D-TFLP 6.5): softmax fixture replay + certified preparation gate #1800

Merged

AdaWorldAPI pushed a commit that referenced this pull request Oct 26, 2026

board: D-TFLP 6.5 landing note (#1800, #1801)

b1853dd

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01ArVbbq3DsToBM7r79zGeEf

AdaWorldAPI marked this pull request as ready for review October 26, 2026 20:43

AdaWorldAPI merged commit f1ff647 into main Oct 26, 2026
4 checks passed

AdaWorldAPI added a commit that referenced this pull request Oct 26, 2026

docs(onnx): account for merged #1800/#1801 evidence and receipt tests

f3fd106

AdaWorldAPI added a commit that referenced this pull request Oct 26, 2026

docs(board): note merged #1800 and #1801 as ONNX frontend prerequisites

e496249

This was referenced Oct 26, 2026

board: #1800 is certified bit-exact against the original TFLite kernel #1807

Merged

docs(D-TFLP): preserve certified TANH evidence, learning/gamma gates and Tract harvest #1810

Merged

Navigation Menu

Repository navigation

probes(D-TFLP 6.5): Codec8 semantic admission + DT64/Quack/D-NAM handoff - #1801

probes(D-TFLP 6.5): Codec8 semantic admission + DT64/Quack/D-NAM handoff#1801
AdaWorldAPI merged 1 commit into
mainfrom
claude/d-tflp-codec8-admission

AdaWorldAPI commented Oct 26, 2026

Uh oh!

coderabbitai Bot commented Oct 26, 2026 •

edited

Loading

Uh oh!

chatgpt-codex-connector Bot commented Oct 26, 2026 •

edited

Loading

Uh oh!

Uh oh!

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Conversation

AdaWorldAPI commented Oct 26, 2026

Why

Scope (one commit)

Evidence used

Tests run

What this does not prove

Production impact

Dependencies / rollback

Links

Uh oh!

coderabbitai Bot commented Oct 26, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

chatgpt-codex-connector Bot commented Oct 26, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codex Review Summary

Uh oh!

Uh oh!

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

coderabbitai Bot commented Oct 26, 2026 •

edited

Loading

chatgpt-codex-connector Bot commented Oct 26, 2026 •

edited

Loading