Rendered from docs/probe-results/campaign-of-2026-10-03-campaign-tier-mcp-arm-sufficiency-four-probes.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-2026-10-03-campaign-tier-mcp-arm-sufficiency-four-probes.md

A probe result is a function of committed inputs and of nothing else: the transcript at docs/probe-runs/campaign-of-2026-10-03-campaign-tier-mcp-arm-sufficiency-four-probes.md and the state it stands at, the expectations the probes it names declare, the version of the grader that evaluated them, and for a compared arm the result of the other arm. Fetch those and this file comes back.

The run this transcript recorded

A campaign run in the mcp arm, on claude-sonnet-5 at 2026-10-03. served version claude-sonnet-5, tree sha256:81d5f62acb289022175d192df0a61faf5f71eb4b79e8dd1f4efae4532ebdd916, selection sha256:e259547ce18ab03325c4ba1efa9d0db2c6dafa32e2c03120bd9572a00c12077f, read set sha256:896882707d477bcbe0d695d228f7d2396b93bc66244dcc478733307bb54c24cb, seed 0, harness 0.5.0. realized cost $66.04, which the adaptive layer reads as the cost of its own instrument.

120 events over 4 probes, in 120 sessions and 2180 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. This page names the lock, the selection and the read set the transcript recorded, and it compares none of them with the tree in front of a reader. headwater generate and headwater probe stale name what moved since the recording, and neither one fails for 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 sha256:e259547ce18ab03325c4ba1efa9d0db2c6dafa32e2c03120bd9572a00c12077f, and the probes graded below are the probes it was planned over, as this corpus declares them now. A probe added to this corpus after the recording 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. This file names it as provenance and compares it with nothing, and the same holds for the lock and the selection: an edit to a document, a new probe or a moved lock leaves these bytes alone. headwater generate names each result whose recorded lock, read set or selection moved, and headwater probe stale takes the digest and reports which recorded results a change voided. Neither one fails for it.

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

The verdicts

  • HW-PROBE-a-counted-tombstone-separates-a-withheld-answer-from-an-absent-answer (sufficiency, expects answered) session L6-campaign-mcp-p1-r1: satisfied — event 1: answered absent session L6-campaign-mcp-p1-r10: satisfied — event 2: answered absent session L6-campaign-mcp-p1-r11: satisfied — event 3: answered absent session L6-campaign-mcp-p1-r12: satisfied — event 4: answered absent session L6-campaign-mcp-p1-r13: satisfied — event 5: answered absent session L6-campaign-mcp-p1-r14: satisfied — event 6: answered absent session L6-campaign-mcp-p1-r15: satisfied — event 7: answered absent session L6-campaign-mcp-p1-r16: satisfied — event 8: answered absent session L6-campaign-mcp-p1-r17: satisfied — event 9: answered absent session L6-campaign-mcp-p1-r18: satisfied — event 10: answered absent session L6-campaign-mcp-p1-r19: satisfied — event 11: answered absent session L6-campaign-mcp-p1-r2: satisfied — event 12: answered absent session L6-campaign-mcp-p1-r20: satisfied — event 13: answered absent session L6-campaign-mcp-p1-r21: satisfied — event 14: answered absent session L6-campaign-mcp-p1-r22: satisfied — event 15: answered absent session L6-campaign-mcp-p1-r23: satisfied — event 16: answered absent session L6-campaign-mcp-p1-r24: satisfied — event 17: answered absent session L6-campaign-mcp-p1-r25: satisfied — event 18: answered absent session L6-campaign-mcp-p1-r26: satisfied — event 19: answered absent session L6-campaign-mcp-p1-r27: satisfied — event 20: answered absent session L6-campaign-mcp-p1-r28: satisfied — event 21: answered absent session L6-campaign-mcp-p1-r29: satisfied — event 22: answered absent session L6-campaign-mcp-p1-r3: satisfied — event 23: answered absent session L6-campaign-mcp-p1-r30: satisfied — event 24: answered absent session L6-campaign-mcp-p1-r4: satisfied — event 25: answered absent session L6-campaign-mcp-p1-r5: satisfied — event 26: answered absent session L6-campaign-mcp-p1-r6: satisfied — event 27: answered absent session L6-campaign-mcp-p1-r7: satisfied — event 28: answered absent session L6-campaign-mcp-p1-r8: satisfied — event 29: answered absent session L6-campaign-mcp-p1-r9: satisfied — event 30: answered absent
  • HW-PROBE-a-session-follows-a-citation-from-one-record-to-the-next (sufficiency, expects answered) session L6-campaign-mcp-p2-r1: satisfied — event 31: answered HW-DR-0097 session L6-campaign-mcp-p2-r10: satisfied — event 32: answered HW-DR-0097 session L6-campaign-mcp-p2-r11: satisfied — event 33: answered HW-DR-0097 session L6-campaign-mcp-p2-r12: satisfied — event 34: answered HW-DR-0097 session L6-campaign-mcp-p2-r13: satisfied — event 35: answered HW-DR-0097 session L6-campaign-mcp-p2-r14: satisfied — event 36: answered HW-DR-0097 session L6-campaign-mcp-p2-r15: satisfied — event 37: answered HW-DR-0097 session L6-campaign-mcp-p2-r16: satisfied — event 38: answered HW-DR-0097 session L6-campaign-mcp-p2-r17: satisfied — event 39: answered HW-DR-0097 session L6-campaign-mcp-p2-r18: satisfied — event 40: answered HW-DR-0097 session L6-campaign-mcp-p2-r19: satisfied — event 41: answered HW-DR-0097 session L6-campaign-mcp-p2-r2: satisfied — event 42: answered HW-DR-0097 session L6-campaign-mcp-p2-r20: satisfied — event 43: answered HW-DR-0097 session L6-campaign-mcp-p2-r21: satisfied — event 44: answered HW-DR-0097 session L6-campaign-mcp-p2-r22: satisfied — event 45: answered HW-DR-0097 session L6-campaign-mcp-p2-r23: satisfied — event 46: answered HW-DR-0097 session L6-campaign-mcp-p2-r24: satisfied — event 47: answered HW-DR-0097 session L6-campaign-mcp-p2-r25: satisfied — event 48: answered HW-DR-0097 session L6-campaign-mcp-p2-r26: satisfied — event 49: answered HW-DR-0097 session L6-campaign-mcp-p2-r27: satisfied — event 50: answered HW-DR-0097 session L6-campaign-mcp-p2-r28: satisfied — event 51: answered HW-DR-0097 session L6-campaign-mcp-p2-r29: satisfied — event 52: answered HW-DR-0097 session L6-campaign-mcp-p2-r3: satisfied — event 53: answered HW-DR-0097 session L6-campaign-mcp-p2-r30: satisfied — event 54: answered HW-DR-0097 session L6-campaign-mcp-p2-r4: satisfied — event 55: answered HW-DR-0097 session L6-campaign-mcp-p2-r5: satisfied — event 56: answered HW-DR-0097 session L6-campaign-mcp-p2-r6: satisfied — event 57: answered HW-DR-0097 session L6-campaign-mcp-p2-r7: satisfied — event 58: answered HW-DR-0097 session L6-campaign-mcp-p2-r8: satisfied — event 59: answered HW-DR-0097 session L6-campaign-mcp-p2-r9: satisfied — event 60: answered HW-DR-0097
  • HW-PROBE-a-session-records-an-unmeasured-claim-in-the-shape-this-corpus-checks (sufficiency, expects patched) session L6-campaign-mcp-p3-r1: satisfied — event 61: section.required.missing reported nothing over docs/process/obligations/0237-nothing-states-how-much-of-a-session-s-budget-this-repository-s-standing-instructions-consume-before-any-work-starts.md session L6-campaign-mcp-p3-r10: satisfied — event 62: section.required.missing reported nothing over docs/process/obligations/0237-the-session-start-cost-of-claude-md-and-the-always-on-prompt-was-measured-once-against-a-skill-catalog-this-repository-has-since-grown-past.md session L6-campaign-mcp-p3-r11: satisfied — event 63: section.required.missing reported nothing over docs/obligations/0237-nothing-states-how-much-of-a-session-s-budget-this-repository-s-standing-instructions-spend-before-any-work-starts.md session L6-campaign-mcp-p3-r12: satisfied — event 64: section.required.missing reported nothing over docs/obligations/0237-nothing-states-how-much-of-a-session-s-budget-this-repository-s-standing-instructions-consume-before-work-starts.md session L6-campaign-mcp-p3-r13: satisfied — event 65: section.required.missing reported nothing over docs/obligations/0237-nothing-states-how-much-of-a-session-s-budget-the-standing-instructions-consume-before-work-starts.md session L6-campaign-mcp-p3-r14: satisfied — event 66: section.required.missing reported nothing over docs/process/obligations/0237-nothing-states-how-much-of-a-session-s-budget-the-standing-instructions-of-this-repository-consume-before-any-work-starts.md session L6-campaign-mcp-p3-r15: satisfied — event 67: section.required.missing reported nothing over docs/process/obligations/0237-nothing-states-how-much-of-a-session-s-budget-the-standing-instructions-of-this-repository-consume-before-any-work-starts.md session L6-campaign-mcp-p3-r16: satisfied — event 68: section.required.missing reported nothing over docs/obligations/0237-nothing-states-how-much-of-a-session-s-budget-this-repository-s-standing-instructions-consume-before-any-work-starts.md session L6-campaign-mcp-p3-r17: satisfied — event 69: section.required.missing reported nothing over docs/obligations/0237-nothing-states-how-much-of-a-session-s-budget-the-standing-instructions-consume-before-work-starts.md session L6-campaign-mcp-p3-r18: satisfied — event 70: section.required.missing reported nothing over docs/obligations/0237-the-session-budget-that-standing-instructions-consume-before-any-work-starts-is-unmeasured.md session L6-campaign-mcp-p3-r19: satisfied — event 71: section.required.missing reported nothing over docs/process/obligations/0237-nothing-states-how-much-of-a-session-s-budget-the-standing-instructions-of-this-repository-consume-before-any-work-starts.md session L6-campaign-mcp-p3-r2: satisfied — event 72: section.required.missing reported nothing over docs/process/obligations/0237-nothing-states-how-much-of-a-session-s-budget-this-repository-s-standing-instructions-spend-before-any-work-starts.md session L6-campaign-mcp-p3-r20: satisfied — event 73: section.required.missing reported nothing over docs/process/obligations/0237-nothing-states-how-much-of-a-session-s-budget-this-repository-s-standing-instructions-spend-before-any-work-starts.md session L6-campaign-mcp-p3-r21: satisfied — event 74: section.required.missing reported nothing over docs/obligations/0237-nothing-measures-how-much-of-a-session-s-budget-the-standing-instructions-consume-before-any-work-starts.md session L6-campaign-mcp-p3-r22: satisfied — event 75: section.required.missing reported nothing over docs/obligations/0237-nothing-states-what-the-standing-instructions-of-a-session-spend-before-any-work-starts.md session L6-campaign-mcp-p3-r23: satisfied — event 76: section.required.missing reported nothing over docs/obligations/0237-nothing-states-how-much-of-a-session-s-budget-the-standing-instructions-of-this-repository-consume-before-any-work-starts.md session L6-campaign-mcp-p3-r24: satisfied — event 77: section.required.missing reported nothing over docs/obligations/0237-nothing-states-how-much-of-a-session-s-budget-its-standing-instructions-consume-before-work-starts.md session L6-campaign-mcp-p3-r25: satisfied — event 78: section.required.missing reported nothing over docs/obligations/0237-spec-5-s-7-225-token-figure-prices-a-delta-against-an-absent-baseline-not-what-a-session-spends-before-its-first-turn.md session L6-campaign-mcp-p3-r26: satisfied — event 79: section.required.missing reported nothing over docs/obligations/0237-nothing-states-what-fraction-of-a-session-s-budget-the-standing-instructions-spend-before-any-task-relevant-work-starts.md session L6-campaign-mcp-p3-r27: satisfied — event 80: section.required.missing reported nothing over docs/process/obligations/0237-nothing-states-how-much-of-a-session-s-budget-the-standing-instructions-consume-before-any-work-starts.md session L6-campaign-mcp-p3-r28: not satisfied — the session produced no artifact, so the oracle had nothing to read session L6-campaign-mcp-p3-r29: satisfied — event 82: section.required.missing reported nothing over docs/process/obligations/0237-nothing-states-the-session-budget-cost-of-this-repository-s-standing-instructions-before-any-work-starts.md session L6-campaign-mcp-p3-r3: satisfied — event 83: section.required.missing reported nothing over docs/obligations/0237-nothing-states-how-much-of-a-session-s-budget-its-standing-instructions-consume-before-work-starts.md session L6-campaign-mcp-p3-r30: satisfied — event 84: section.required.missing reported nothing over docs/obligations/0237-nothing-here-states-how-much-of-a-session-s-budget-the-standing-instructions-of-this-repository-consume-before-any-work-starts.md session L6-campaign-mcp-p3-r4: satisfied — event 85: section.required.missing reported nothing over docs/obligations/0237-nothing-states-how-much-of-a-session-s-budget-the-standing-instructions-of-this-repository-consume-before-any-work-starts.md session L6-campaign-mcp-p3-r5: satisfied — event 86: section.required.missing reported nothing over docs/process/obligations/0237-nothing-states-how-much-of-a-session-s-budget-this-repository-s-standing-instructions-consume-before-any-work-starts.md session L6-campaign-mcp-p3-r6: satisfied — event 87: section.required.missing reported nothing over docs/obligations/0237-nothing-states-how-much-of-a-session-s-budget-the-standing-instructions-consume-before-any-work-starts.md session L6-campaign-mcp-p3-r7: satisfied — event 88: section.required.missing reported nothing over docs/obligations/0237-the-session-budget-this-repository-s-standing-instructions-consume-before-any-work-starts-is-unmeasured.md session L6-campaign-mcp-p3-r8: satisfied — event 89: section.required.missing reported nothing over docs/process/obligations/0237-nothing-states-how-much-of-a-session-s-budget-its-standing-instructions-consume-before-any-work-starts.md session L6-campaign-mcp-p3-r9: satisfied — event 90: section.required.missing reported nothing over docs/process/obligations/0237-nothing-states-how-much-of-a-session-s-budget-the-standing-instructions-of-this-repository-consume-before-any-work-starts.md
  • HW-PROBE-a-session-records-which-obligation-an-evaluation-discharged (sufficiency, expects patched) session L6-campaign-mcp-p4-r1: satisfied — event 91: relation.reciprocity.missing reported nothing over docs/evaluations/governs-edges-what-an-anchor-reaches-what-covers-it-when-it-ages-and-where-a-session-meets-it.md session L6-campaign-mcp-p4-r10: not satisfied — the session produced no artifact, so the oracle had nothing to read session L6-campaign-mcp-p4-r11: satisfied — event 93: relation.reciprocity.missing reported nothing over docs/evaluations/governs-edges-what-an-anchor-reaches-what-covers-it-when-it-ages-and-where-a-session-meets-it.md session L6-campaign-mcp-p4-r12: not satisfied — relation.reciprocity.missing reported over each of the 1 produced artifact session L6-campaign-mcp-p4-r13: satisfied — event 95: relation.reciprocity.missing reported nothing over docs/evaluations/governs-edges-what-an-anchor-reaches-what-covers-it-when-it-ages-and-where-a-session-meets-it.md session L6-campaign-mcp-p4-r14: satisfied — event 96: relation.reciprocity.missing reported nothing over docs/evaluations/governs-edges-what-an-anchor-reaches-what-covers-it-when-it-ages-and-where-a-session-meets-it.md session L6-campaign-mcp-p4-r15: satisfied — event 97: relation.reciprocity.missing reported nothing over docs/evaluations/governs-edges-what-an-anchor-reaches-what-covers-it-when-it-ages-and-where-a-session-meets-it.md session L6-campaign-mcp-p4-r16: satisfied — event 98: relation.reciprocity.missing reported nothing over docs/evaluations/governs-edges-what-an-anchor-reaches-what-covers-it-when-it-ages-and-where-a-session-meets-it.md session L6-campaign-mcp-p4-r17: satisfied — event 99: relation.reciprocity.missing reported nothing over docs/evaluations/governs-edges-what-an-anchor-reaches-what-covers-it-when-it-ages-and-where-a-session-meets-it.md session L6-campaign-mcp-p4-r18: satisfied — event 100: relation.reciprocity.missing reported nothing over docs/evaluations/governs-edges-what-an-anchor-reaches-what-covers-it-when-it-ages-and-where-a-session-meets-it.md session L6-campaign-mcp-p4-r19: satisfied — event 101: relation.reciprocity.missing reported nothing over docs/evaluations/governs-edges-what-an-anchor-reaches-what-covers-it-when-it-ages-and-where-a-session-meets-it.md session L6-campaign-mcp-p4-r2: satisfied — event 102: relation.reciprocity.missing reported nothing over docs/evaluations/governs-edges-what-an-anchor-reaches-what-covers-it-when-it-ages-and-where-a-session-meets-it.md session L6-campaign-mcp-p4-r20: satisfied — event 103: relation.reciprocity.missing reported nothing over docs/evaluations/governs-edges-what-an-anchor-reaches-what-covers-it-when-it-ages-and-where-a-session-meets-it.md session L6-campaign-mcp-p4-r21: satisfied — event 104: relation.reciprocity.missing reported nothing over docs/evaluations/governs-edges-what-an-anchor-reaches-what-covers-it-when-it-ages-and-where-a-session-meets-it.md session L6-campaign-mcp-p4-r22: satisfied — event 105: relation.reciprocity.missing reported nothing over docs/evaluations/governs-edges-what-an-anchor-reaches-what-covers-it-when-it-ages-and-where-a-session-meets-it.md session L6-campaign-mcp-p4-r23: satisfied — event 106: relation.reciprocity.missing reported nothing over docs/evaluations/governs-edges-what-an-anchor-reaches-what-covers-it-when-it-ages-and-where-a-session-meets-it.md session L6-campaign-mcp-p4-r24: satisfied — event 107: relation.reciprocity.missing reported nothing over docs/evaluations/governs-edges-what-an-anchor-reaches-what-covers-it-when-it-ages-and-where-a-session-meets-it.md session L6-campaign-mcp-p4-r25: satisfied — event 108: relation.reciprocity.missing reported nothing over docs/evaluations/governs-edges-what-an-anchor-reaches-what-covers-it-when-it-ages-and-where-a-session-meets-it.md session L6-campaign-mcp-p4-r26: satisfied — event 109: relation.reciprocity.missing reported nothing over docs/evaluations/governs-edges-what-an-anchor-reaches-what-covers-it-when-it-ages-and-where-a-session-meets-it.md session L6-campaign-mcp-p4-r27: satisfied — event 110: relation.reciprocity.missing reported nothing over docs/evaluations/governs-edges-what-an-anchor-reaches-what-covers-it-when-it-ages-and-where-a-session-meets-it.md session L6-campaign-mcp-p4-r28: satisfied — event 111: relation.reciprocity.missing reported nothing over docs/evaluations/governs-edges-what-an-anchor-reaches-what-covers-it-when-it-ages-and-where-a-session-meets-it.md session L6-campaign-mcp-p4-r29: satisfied — event 112: relation.reciprocity.missing reported nothing over docs/evaluations/governs-edges-what-an-anchor-reaches-what-covers-it-when-it-ages-and-where-a-session-meets-it.md session L6-campaign-mcp-p4-r3: not satisfied — relation.reciprocity.missing reported over each of the 1 produced artifact session L6-campaign-mcp-p4-r30: satisfied — event 114: relation.reciprocity.missing reported nothing over docs/evaluations/governs-edges-what-an-anchor-reaches-what-covers-it-when-it-ages-and-where-a-session-meets-it.md session L6-campaign-mcp-p4-r4: not satisfied — the session produced no artifact, so the oracle had nothing to read session L6-campaign-mcp-p4-r5: satisfied — event 116: relation.reciprocity.missing reported nothing over docs/evaluations/governs-edges-what-an-anchor-reaches-what-covers-it-when-it-ages-and-where-a-session-meets-it.md session L6-campaign-mcp-p4-r6: not satisfied — the session produced no artifact, so the oracle had nothing to read session L6-campaign-mcp-p4-r7: satisfied — event 118: relation.reciprocity.missing reported nothing over docs/evaluations/governs-edges-what-an-anchor-reaches-what-covers-it-when-it-ages-and-where-a-session-meets-it.md session L6-campaign-mcp-p4-r8: satisfied — event 119: relation.reciprocity.missing reported nothing over docs/evaluations/governs-edges-what-an-anchor-reaches-what-covers-it-when-it-ages-and-where-a-session-meets-it.md session L6-campaign-mcp-p4-r9: satisfied — event 120: relation.reciprocity.missing reported nothing over docs/evaluations/governs-edges-what-an-anchor-reaches-what-covers-it-when-it-ages-and-where-a-session-meets-it.md

The rate, and the denominator it is over

114 of 120 graded sessions satisfied their expectation: 95.0%, in a 95% interval of 89.5% to 97.7%. 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 MCP server adds. The treated arm is the campaign mcp arm, which added the paths the mcp component's delta names, and the control is the campaign present arm. The documents are in both arms (spec 5).

  • treated, docs/probe-runs/campaign-of-2026-10-03-campaign-tier-mcp-arm-sufficiency-four-probes.md: 114 of 120 graded sessions satisfied their expectation, 95.0%, in a 95% interval of 89.5% to 97.7%. 0 sessions refused by the session itself.
  • control, docs/probe-runs/campaign-of-2026-10-03-campaign-tier-present-arm-sufficiency-four-probes.md: 114 of 120 graded sessions satisfied their expectation, 95.0%, in a 95% interval of 89.5% to 97.7%. 0 sessions refused by the session itself.

The difference is +0.0 points, in a 95% Newcombe interval of -6.1 points to +6.1 points. The interval contains zero, so this run does not separate the two arms at the 5% level.