Rendered from docs/probe-results/campaign-of-2026-09-30-campaign-tier-present-arm-discovery.md in the Headwater corpus. Every document on this half of the site is typed by the taxonomy the descriptor names: corpus.json.

The result of docs/probe-runs/campaign-of-2026-09-30-campaign-tier-present-arm-discovery.md

A probe result is a function of three committed inputs and of nothing else: the transcript at docs/probe-runs/campaign-of-2026-09-30-campaign-tier-present-arm-discovery.md, the expectations the probes of this corpus declare, and the version of the grader that evaluated them. Fetch the three and this file comes back.

The run this transcript recorded

A campaign run in the present arm, on claude-sonnet-5 at 2026-09-30. served version claude-sonnet-5, tree sha256:1c21e55eb09aca89b3d736ac79f8accfc64ddd7853bace5fb5e577a82d794bb9, selection sha256:8ec515cf6d769b34f8fc840d00f902606dc575c74a4105c7599670b563be6ad5, read set sha256:c563ff36d54e56139a6d1ca7cca03fbe175864970d6fcacabe5144cfcaaaffc9, seed 0, harness 0.5.0. realized cost $36.86, which the adaptive layer reads as the cost of its own instrument. It was planned against taxonomy sha256:4b95fd419a3824634b9644c91b13356ba9b71567f813445913c430e3aa12c13a, and this tree carries another, and the read set of its probes moved since the recording. Each verdict below is what the session did over the documents it met, graded against the expectations this tree declares now, which can differ from the ones the session ran under. headwater probe stale names what moved.

60 events over 2 of the 10 probes this corpus declares, in 60 sessions and 1292 tool calls.

The engine confirmed the taxonomy, that every member of the run identity is present, the membership of every probe named, that no key outside the closed set appears, and that a realized cost was recorded. Present is not confirmed: of the six members a plan fixes before a run, the lock is the one compared here. A lock that differs refuses the file only where this tree composes no read set over its probes, and a read set that differs marks the verdicts and keeps them. headwater probe stale compares the read set against the tree in front of it. It graded nothing: a verdict is a function of this transcript, the expectations these probes declare and a grader version, and headwater probe grade is the verb that holds all three.

The selection this transcript names is not the selection this corpus composes. The transcript names sha256:8ec515cf6d769b34f8fc840d00f902606dc575c74a4105c7599670b563be6ad5 and this corpus composes sha256:50ed5ce43072d9f674a8dc52e1dcad84de4d649e39a77ce22179836188196b09. The transcript's digest is the digest of 2 of the 10 probes this corpus composes, so the run was planned over that part of the selection, or the rest were added after it was recorded. Every verdict below is over that part, and a probe outside it is not a session this run owed. The tree, the seed and the harness above are provenance and nothing compares them.

The read_set digest above covers every probe of the selection and every document one of them examines, by path and content. It is recorded here, and this file names no digest the tree in front of a reader composes. Where that tree composes another one, a single sentence above says that the read set moved, and it moves these bytes once. A digest from that tree would move these bytes on every edit to a document the selection points at, and generate --check holds this file to its bytes, so the staleness of a measurement would stop a merge. headwater probe stale takes the digest and reports which recorded results a change voided.

Graded by grader 0.5.0. A campaign run in the present arm, on claude-sonnet-5 at served version claude-sonnet-5.

The verdicts

• HW-PROBE-a-cold-agent-reaches-the-governing-document-through-the-corpus-descriptor (discovery, expects opened) session L5-campaign-present-p1-r1: satisfied — event 1, call 27: Bash {"command":"cd /home/james/.claude/jobs/1338ac1e/tmp/issue-1384/full/campaign/ws/L5-campaign-present-p1-r1 && sed -n '1,400p' docs/spec/12-check-layer.md 2>/dev/null | head -200","description":"Read spec 12 check layer doc"} session L5-campaign-present-p1-r10: not satisfied — 20 recorded calls, and none named any of the 1 document session L5-campaign-present-p1-r11: not satisfied — 22 recorded calls, and none named any of the 1 document session L5-campaign-present-p1-r12: satisfied — event 4, call 21: Read /home/james/.claude/jobs/1338ac1e/tmp/issue-1384/full/campaign/ws/L5-campaign-present-p1-r12/docs/spec/12-check-layer.md session L5-campaign-present-p1-r13: not satisfied — 28 recorded calls, and none named any of the 1 document session L5-campaign-present-p1-r14: not satisfied — 23 recorded calls, and none named any of the 1 document session L5-campaign-present-p1-r15: not satisfied — 20 recorded calls, and none named any of the 1 document session L5-campaign-present-p1-r16: satisfied — event 8, call 30: Bash {"command":"cd /home/james/.claude/jobs/1338ac1e/tmp/issue-1384/full/campaign/ws/L5-campaign-present-p1-r16 && grep -n -i \"severity\\|error\\b\\|advisory\\|warn\\b\" docs/spec/12-check-layer.md | head -80"} session L5-campaign-present-p1-r17: satisfied — event 9, call 21: Bash {"command":"grep -rn \"facet.value.blank\\|facet_blank\\|Blank::\" engine/crates/check/src/register.rs taxonomy-source/headwater-standard/taxonomy.yml docs/spec/12-check-layer.md 2>/dev/null","description":"Search for facet.value.blank registration and doc references"} session L5-campaign-present-p1-r18: not satisfied — 35 recorded calls, and none named any of the 1 document session L5-campaign-present-p1-r19: not satisfied — 29 recorded calls, and none named any of the 1 document session L5-campaign-present-p1-r2: not satisfied — 31 recorded calls, and none named any of the 1 document session L5-campaign-present-p1-r20: not satisfied — 19 recorded calls, and none named any of the 1 document session L5-campaign-present-p1-r21: not satisfied — 26 recorded calls, and none named any of the 1 document session L5-campaign-present-p1-r22: not satisfied — 32 recorded calls, and none named any of the 1 document session L5-campaign-present-p1-r23: not satisfied — 23 recorded calls, and none named any of the 1 document session L5-campaign-present-p1-r24: not satisfied — 20 recorded calls, and none named any of the 1 document session L5-campaign-present-p1-r25: not satisfied — 32 recorded calls, and none named any of the 1 document session L5-campaign-present-p1-r26: not satisfied — 18 recorded calls, and none named any of the 1 document session L5-campaign-present-p1-r27: not satisfied — 20 recorded calls, and none named any of the 1 document session L5-campaign-present-p1-r28: not satisfied — 26 recorded calls, and none named any of the 1 document session L5-campaign-present-p1-r29: satisfied — event 22, call 18: Bash {"command":"grep -n \"facet.value.blank\\|facet_blank\" docs/spec/12-check-layer.md engine/crates/check/src/lib.rs engine/crates/cli/src/main.rs 2>/dev/null","description":"Find spec and registration references to the blank-facet rule"} session L5-campaign-present-p1-r3: not satisfied — 123 recorded calls, and none named any of the 1 document session L5-campaign-present-p1-r30: satisfied — event 24, call 17: Bash {"command":"grep -rln \"facet.value.blank\" docs/ 2>/dev/null; echo \"---lib.rs facet_blank registration---\"; grep -n \"facet_blank\" engine/crates/check/src/lib.rs; echo \"---spec 12 excerpt about generated rules---\"; grep -n \"generated\\|Shape-origin\\|declared in.*taxonomy\" docs/spec/12-check-layer.md | head -30","description":"Search docs for facet.value.blank rule catalog and generation description"} session L5-campaign-present-p1-r4: not satisfied — 8 recorded calls, and none named any of the 1 document session L5-campaign-present-p1-r5: not satisfied — 33 recorded calls, and none named any of the 1 document session L5-campaign-present-p1-r6: satisfied — event 27, call 38: Bash {"command":"grep -n \"^#\\|fixability\\|Fixability\\|mechanical\" docs/spec/12-check-layer.md | head -60"} session L5-campaign-present-p1-r7: not satisfied — 26 recorded calls, and none named any of the 1 document session L5-campaign-present-p1-r8: not satisfied — 29 recorded calls, and none named any of the 1 document session L5-campaign-present-p1-r9: not satisfied — 22 recorded calls, and none named any of the 1 document
• HW-PROBE-the-pointer-this-corpus-offers-for-a-task-is-the-document-a-session-opens (discovery, expects opened) session L5-campaign-present-p2-r1: satisfied — event 31, call 20: Bash {"command":"sed -n '1,30p' docs/obligations/0107-the-base-package-ships-a-kind-that-the-scaffolder-refuses-to-write.md","description":"Read obligation 0107 for the specification-kind refusal example"} session L5-campaign-present-p2-r10: not satisfied — 11 recorded calls, and none named any of the 1 document session L5-campaign-present-p2-r11: not satisfied — 22 recorded calls, and none named any of the 1 document session L5-campaign-present-p2-r12: not satisfied — 13 recorded calls, and none named any of the 1 document session L5-campaign-present-p2-r13: not satisfied — 3 recorded calls, and none named any of the 1 document session L5-campaign-present-p2-r14: not satisfied — 14 recorded calls, and none named any of the 1 document session L5-campaign-present-p2-r15: satisfied — event 37, call 21: Read /home/james/.claude/jobs/1338ac1e/tmp/issue-1384/full/campaign/ws/L5-campaign-present-p2-r15/docs/obligations/0107-the-base-package-ships-a-kind-that-the-scaffolder-refuses-to-write.md session L5-campaign-present-p2-r16: not satisfied — 11 recorded calls, and none named any of the 1 document session L5-campaign-present-p2-r17: not satisfied — 11 recorded calls, and none named any of the 1 document session L5-campaign-present-p2-r18: not satisfied — 10 recorded calls, and none named any of the 1 document session L5-campaign-present-p2-r19: not satisfied — 10 recorded calls, and none named any of the 1 document session L5-campaign-present-p2-r2: not satisfied — 14 recorded calls, and none named any of the 1 document session L5-campaign-present-p2-r20: not satisfied — 19 recorded calls, and none named any of the 1 document session L5-campaign-present-p2-r21: not satisfied — 29 recorded calls, and none named any of the 1 document session L5-campaign-present-p2-r22: not satisfied — 8 recorded calls, and none named any of the 1 document session L5-campaign-present-p2-r23: not satisfied — 15 recorded calls, and none named any of the 1 document session L5-campaign-present-p2-r24: not satisfied — 15 recorded calls, and none named any of the 1 document session L5-campaign-present-p2-r25: not satisfied — 7 recorded calls, and none named any of the 1 document session L5-campaign-present-p2-r26: not satisfied — 10 recorded calls, and none named any of the 1 document session L5-campaign-present-p2-r27: not satisfied — 20 recorded calls, and none named any of the 1 document session L5-campaign-present-p2-r28: not satisfied — 16 recorded calls, and none named any of the 1 document session L5-campaign-present-p2-r29: not satisfied — 9 recorded calls, and none named any of the 1 document session L5-campaign-present-p2-r3: not satisfied — 1 recorded call, and none named any of the 1 document session L5-campaign-present-p2-r30: not satisfied — 12 recorded calls, and none named any of the 1 document session L5-campaign-present-p2-r4: not satisfied — 22 recorded calls, and none named any of the 1 document session L5-campaign-present-p2-r5: not satisfied — 15 recorded calls, and none named any of the 1 document session L5-campaign-present-p2-r6: not satisfied — 14 recorded calls, and none named any of the 1 document session L5-campaign-present-p2-r7: not satisfied — 10 recorded calls, and none named any of the 1 document session L5-campaign-present-p2-r8: not satisfied — 2 recorded calls, and none named any of the 1 document session L5-campaign-present-p2-r9: not satisfied — 14 recorded calls, and none named any of the 1 document

The rate, and the denominator it is over

9 of 60 graded sessions satisfied their expectation: 15.0%, in a 95% interval of 8.1% to 26.1%. The denominator is the graded sessions and never the selected probes. 0 sessions reached no verdict, and a session with no verdict is outside both halves of that fraction.

An interval that overlaps the previous run's is variance and one that does not is drift. This is one arm, so it estimates no effect: an efficacy claim is a comparison of two results, and the arm each one recorded is on it.

The comparisons this arm takes part in

What the governance changes. The treated arm is the campaign present arm and the control is the campaign absent arm, which removed the paths the campaign tier's ablation names and kept docs/.

• treated, docs/probe-runs/campaign-of-2026-09-30-campaign-tier-present-arm-discovery.md: 9 of 60 graded sessions satisfied their expectation, 15.0%, in a 95% interval of 8.1% to 26.1%. 0 sessions refused by the session itself.
• control, docs/probe-runs/campaign-of-2026-09-30-campaign-tier-absent-arm-discovery.md: 5 of 60 graded sessions satisfied their expectation, 8.3%, in a 95% interval of 3.6% to 18.1%. 0 sessions refused by the session itself.

The difference is +6.7 points, in a 95% Newcombe interval of -5.3 points to +18.7 points. The interval contains zero, so this run does not separate the two arms at the 5% level.