probe: DAV × EWA × revision active-observation cycle

[AdaWorldAPI]

What

One diagnostic jc example that wires three existing pieces into a closed, replayable loop:

1. lance-graph-contract::sigma_propagation::ewa_sandwich, the certified EWA kernel, defines local reach.
2. jc::quorum::pairwise_agreement_u8, the quorum dispersion 1 − σ/σ_max(k), gives the route disagreement as its complement.
3. GadamerRevision is the only way an observation changes what the probe believes.

There is no new pillar, primitive, 0..63 ordinal, production policy or write path.

The loop

epistemic hole -> EWA frontier -> route disagreement -> top-1
  -> reveal (only entry of hidden truth) -> GadamerRevision -> replay from delta.resulting -> collapse

The belief state is the revision horizon.

• Bit i of independent_roots means site i was observed; bit i of projected_claims is what was observed there.
• The routes read that horizon and nothing else.
• A revealed observation is presented as a new independent root, and the result is adopted only if the verdict is EvidentialEffect::IncreaseEligible. Replay then reads delta.resulting.
• The counterfactual docket (is_acceptable) is not walked: this updates the probe's working horizon, not actual-world state.

Falsifiers (runtime asserts in main; CI now runs the example)

• Reach rule, at every aperture: the target is in the frontier iff a^10 >= floor, computed independently with powi. When it is reachable it ranks first. The floor 0.0025 puts the boundary between 0.50 and 0.55, so this demonstrates the rule, not that 0.55 is special.
• Enumeration order cannot change the ranking. The check includes a real score tie, at a=0.65 between sites 6 and 13, with an anti-vacuity guard.
• Before observation, the selected site has disagreement 255 (maximally split).
• No laundering: DAV's own route consensus, presented as an inherited root, is refused by revision, the horizon is unchanged, and the hole stays open.
• After observe, revise and replay: disagreement at the selected site is 0, and residual disagreement over all withheld sites goes from 255 to 0.
• Ordinal baseline: k=1 picks site 6, which already had disagreement 0, and leaves the residual at 255.
• Determinism: running the cycle twice gives an identical result, horizon included.

Disable runs, each red at its own assertion:

disable	fails at
observe skips adopting the revision result	replay did not collapse the selected hole
court-of-appeal gate removed	route consensus changed the belief state
idx tie-break removed	enumeration order changed the ranking (a=0.65)
disagreement forced to 0	reachable target must rank first
reveal sent as inherited instead of independent	must enter as a new independent root

Changed from the first version

• The revision call was decorative. Its delta was asserted on and discarded, and replay wrote the observation directly. Deleting the call left the probe green.
• Disagreement re-implemented the quorum formula locally. It now calls the jc function.
• The 0.45 misses / 0.55 hits asserts held only because of the chosen floor. They are replaced by the reach rule above.
• Fixed rustfmt and a clippy map_or error, both CI failures.
• CI: jc-proof.yml now runs the example and triggers on revision.rs and sigma_propagation.rs.

Known limits

• The EWA field is isotropic (M = √a·I), so the kernel computes a^h. It is the real kernel, but its anisotropy, which is the part that would matter, is not exercised.
• Below reach, top-1 is a zero-disagreement site. A scheduler should probably skip those; that is not changed here.

Follow-up question: can Option<bool> become an existing LG carrier?

Yes, without changing the control contract, because the contract already reads exactly two planes: observed and value. That is the NULL representation ruled in #1334: an ordinary value lane plus its resident validity plane (lance-graph-quack: Mask, Filter::is_null / is_not_null, Filter::sql_where).

The smallest follow-up slice:

• Hold the sites as a quack population with one value lane and its validity Mask.
• Select candidates as Filter::and([Filter::is_null(valid), frontier]). With an isotropic EWA field the frontier is a contiguous hop range, so Cmp::Range works.
• Write back only from delta.resulting, by setting the validity bit and the value. On the revision side, EvidenceMask for [u64; N] already covers more than 64 sites.
• The routes stay probe-local scans over the lanes.

One trap to avoid: do not list the frontier's column in sql_where's nullable. Cmp::Range is gated on its lane's validity there, so it would drop exactly the NULL rows that are the candidates.

This is not trivial, because it needs a mask-risc population fixture, so it is not in this PR.

Run

cargo run --manifest-path crates/jc/Cargo.toml --example dav_active_observation_probe

🤖 Generated with Claude Code

https://claude.ai/code/session_01DgxrCafsJuAahpBs7oC14R

Summary by CodeRabbit

• Tests
  • Added an observation-probing example that evaluates how candidate observations are ranked and how new evidence affects revision results.
  • The automated pull-request checks now run this example and include additional revision and propagation changes in their trigger paths.

[coderabbitai]

Important

Draft PR not reviewed

Draft PRs are not automatically reviewed by default.

• Pull request is closed - (🔄 Check again to try again)

To automatically review draft PRs, update your CodeRabbit configuration:

reviews:
  auto_review:
    drafts: true

• 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.}

[cursor]

Bugbot couldn't run - usage limit reached

Bugbot is counted against Cursor usage for this user or team, and this run hit a usage or spend limit.

A user or team admin can review and increase usage limits in the Cursor dashboard.

(requestId: b2dff3cd-a090-4ee9-9c53-dd45ec6638c4)

[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 2026-10-06T08:26:12.610829Z	943b430	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.

[cursor]

Bugbot couldn't run - usage limit reached

Bugbot is counted against Cursor usage for this user or team, and this run hit a usage or spend limit.

A user or team admin can review and increase usage limits in the Cursor dashboard.

(requestId: 40612ab4-f463-4c26-b8f9-201a9cab2f8f)