D-CE64-SEM-0: prove semantic upper-half transition loop

[AdaWorldAPI]

Claim under test

Exercise the proposed CausalEdge64 upper-half reading on the real D-PEARL-IO-0 fixture without changing the wire layout or production enum names.

Probe-local candidate reading:

• 40..42 Pearl question (existing CausalMask)
• 43..45 orientation / syntax (existing direction triad)
• 46..49 signed activation from measured reaction, not opcode
• 50..52 novelty / epistemic entropy proxy from surviving Epi5 population width
• 53..58 belief-state update breadcrumb
• 59..63 canonical EpistemicState5

Falsifiers

1. Hydration and PO advance W only when EpistemicState5 actually changes; SPO reaction alone must not steal the breadcrumb.
2. The same SPO operator must be able to produce opposite activation signs, proving i4 can represent reaction rather than operation echo.
3. Novelty is derived from the width of the still-legal EpistemicState5 population: 24 states -> 5, four states -> 2, one state -> 0.
4. All eight direction triads must produce the same PO measurement and survive revision unchanged, so orientation cannot secretly select the Pearl operator.

Scope

• Probe only.
• No layout change.
• No production accessor rename.
• No new storage object.
• Existing D-PEARL-IO-0 behavior and its 12 falsifiers remain intact.

If this survives CI, the next question is whether these alternate readings deserve contract names or should remain experimental.

[coderabbitai]

Warning

Review limit reached

Your organization has reached its usage spending cap. Adjust your spending cap in the billing tab.

Next included review available in 19 minutes.

Check out review usage here.

View limit details

Limit details: You’ve used the included review currently available. Your 60 included PR review attempts over the past 7 days set your current allowance at 1 review per hour.

Learn how review limits work.

Review configuration:

⚙️ Run configuration

• Configuration used: Organization UI
• Review profile: CHILL
• Plan: Essentials
• Run ID: 5362092f-54d8-4eb0-aa2b-c231612c3836

📥 Commits

Reviewing files that changed from the base of the PR and between cc15ec0 and 6c98015.

📒 Files selected for processing (2)

• .claude/board/entries/README.md
• crates/cognitive-shader-driver/examples/pearl_ladder_probe.rs

• Autopilot · 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.

---

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

[chatgpt-codex-connector]

Codex Review Summary

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

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-10-07T21:31:04.648628Z	97789b2	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.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 97789b2ed7

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".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[AdaWorldAPI]

Merged main into this branch in 0f2682c, a normal merge commit with no rewrite of the branch. #1391 moved the Pearl ladder out of this probe into lance_graph_planner::pearl, so the four tests are ported onto the production API:

• hydrate(.., Reading) and revise(&Measured, Reading). revise no longer takes an edge, because a measurement is now bound to the edge it measured.
• The orientation test now compares only what the operator computed. A Measured carries its input edge, so results for two directions can no longer be equal as whole values.
• The activation quantizer uses div_ceil (clippy manual_div_ceil).

All 4 tests pass. Clippy -D warnings is clean, and the probe's printed output is byte-identical to main's.

Findings on the proposed readings

Each of the three new readings reuses bits that already mean something in code:

1. Bits 46..49 are the inference type. CausalEdge64::inference() decodes them through InferenceType::from_mantissa, so writing an activation there changes what the edge says about how it was inferred. The test's -3 decodes as Abduction. The quantizer also produces ±6 for any frequency change of 161 to 192 (for example 20 → 200 gives +6), and ±6 are exactly the Intervention and Counterfactual codes. assert_ne!(activation(down), Counterfactual.to_mantissa()) checks only one fixture, where the value happens to be -3. This also conflicts with the rule that the ±6 tag stays separate from the measured result.
2. Bits 50..52 are plasticity. causal-edge's network.rs filters on plasticity() == ALL_HOT, so a novelty value written there changes which edges that filter keeps.
3. Bits 53..58 are the witness handle. w_slot is documented as the witness corpus root handle, and since D-SPOG-W-0: the Witness slot as a SPOG sub-context under classid-G #1371 it is the SPO-G sub-context. Using it as a belief-update breadcrumb overwrites the anchor.
4. CI does not run these tests. They live in an example with required-features = ["with-planner"], and CI builds the shader driver with default features only. So "if this survives CI" holds whether or not the tests pass. To gate them, they would have to move into a crate CI tests.

None of this affects production while the readings stay test-only and the layout is unchanged. Before any of them gets a contract name, each needs either its own bits or an explicit decision to take these bits over.

---

Generated by Claude Code