Skip to content

D-BAG-FOLD v4: attested pointer lens and candidate-gated execution (lab) - #1939

Merged
AdaWorldAPI merged 1 commit into
mainfrom
ccr-85acaff0-oszq40
Nov 3, 2026
Merged

AdaWorldAPI merged 1 commit into
mainfrom
ccr-85acaff0-oszq40

Conversation

@AdaWorldAPI

Copy link
Copy Markdown
Owner

This PR adds D-BAG-FOLD v4. #1938 merged before this commit was pushed, so #1938 contains only v3; v4 arrives here as its own change on top of current main.

Everything lives in the isolated crate .claude/probes/2026-11-03-d-bag-fold/probe-v4/, which has its own [workspace].

  • No production code changes. mask-risc's IR is used as is, Quack is unchanged, and MaskOp::Gather runs only inside the lab crate.
  • v1–v3 are unchanged.
  • Arm L is not promoted, EXO-140 and EXO-46 are unchanged, and no DT64 route is added.

Results (21/21 checks)

A. Attested pointer lens. Two witnesses are minted at construction:

  • Presence witness, minted from the source edge facts with the production ordered_lane::digest_of over child keys. It refuses a plane misaligned at construction, a permuted plane, and a deletion that bypasses the generation and frame.
  • Pointer lens, minted by witnessed seek_from and origin-checked at mint. It refuses duplicate keys, out-of-bounds pointers, value-chosen pointers and cross-ClassId pointers.

Production FrameStore and OrderedLaneWitness refuse foreign or stale frames and versions. Remaining gap: a forged table with a re-sealed digest still passes verification, so origin can only be attested at mint.

B. Candidate-gated execution. One mask-risc program gates both Gathers by an in-program register (Range ∧ T_in).

  • It matches Quack's Arm L and the oracle, and disabling either Gather turns it red.
  • It scales with the candidate count, not the lane size: 1.2× faster than Quack's Arm L on the smallest lane, 4.7–46.6× on larger ones.
  • Quack cannot lower it, because gate_walk only gates by one resident plane.
  • Presence verification costs more than the gated execution, so it must run once per generation, not per query.

Verdict: Arm L is still NO-GO for admission. It is ready to be written up as a narrowly scoped proposal. The decision memo is REPORT-V4.md.

Board

New entry D-BAG-FOLD-V4; the entries index is regenerated.

Validation

Run cargo +1.99.0 run --release -q in probe-v4/; it exits non-zero if any check fails. The sibling ../ndarray checkout is required.


Generated by Claude Code

New isolated probe crate .claude/probes/2026-11-03-d-bag-fold/probe-v4/ (own
[workspace]); v1-v3 unchanged. 21/21 checks.

A. Construction-time witnesses on production types: a presence witness
minted from the source edge facts (ordered_lane::digest_of over child keys)
refuses a plane misaligned at construction, a permuted plane and a deletion
that bypasses generation and frame; a pointer lens minted by witnessed
seek_from and origin-checked at mint refuses duplicate keys, out-of-bounds,
value-chosen and cross-ClassId pointers. FrameStore and OrderedLaneWitness
refuse foreign/stale frames and versions. A re-sealed forged table still
passes verification: origin is attestable only at mint.

B. One mask-risc program (IR unchanged) gates both Gathers by an in-program
register (Range and T_in). Equal to Quack's Arm L and the oracle, both
Gather disables red; it scales with candidates rather than lane size
(1.2x on the smallest lane, 4.7-46.6x on larger ones). Quack's gate_walk
cannot lower it.

Arm L stays NO-GO for admission; decision memo in REPORT-V4.md.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01XAqKMNZqKXSaeb9g2ypywF
@coderabbitai

coderabbitai Bot commented Nov 3, 2026 •

Copy link
Copy Markdown

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: f722d55b-d31b-4e28-9240-3f5b947c8cec
  • Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

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

CodeRabbit Verify

  • Run CodeRabbit Verify to test the changes in this pull request.

@AdaWorldAPI
AdaWorldAPI marked this pull request as ready for review November 3, 2026 15:29
@AdaWorldAPI
AdaWorldAPI merged commit 119787f into main Nov 3, 2026
4 checks passed
@chatgpt-codex-connector

chatgpt-codex-connector Bot commented Nov 3, 2026 •

Copy link
Copy Markdown

Codex Review Summary

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

Review Status Commit Review trigger
📝 Code Review ✅ Completed 2026-11-03T15:29:55.805586Z 6891d58 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.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants