Rendered from docs/probe-results/additive-tuning-of-2026-11-02-engine-round-1-variant-1-scenario-2.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/additive-tuning-of-2026-11-02-engine-round-1-variant-1-scenario-2.md
A probe result is a function of committed inputs and of nothing else: the transcript at docs/probe-runs/additive-tuning-of-2026-11-02-engine-round-1-variant-1-scenario-2.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 additive run in the scenario-2 arm, on claude-sonnet-5 at 2026-11-02. served version claude-sonnet-5, tree sha256:dc2bf3ba391bc95527536b0dbe3a5b5065c3b1e86b0d330fba91d5ad4970fe20, selection sha256:d99530b2b639bb6bdd279be2a9356a64ce2a0c8cceca354d56957f198a76a20a, read set sha256:2a47d5b0401ca6d3239d352781fe101f4d7ba74c60c6a0dbe460e46ae9703c4d, seed 0, harness 0.6.1. realized cost $16.79, which the adaptive layer reads as the cost of its own instrument.
22 events over 22 probes, in 22 sessions and 573 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:d99530b2b639bb6bdd279be2a9356a64ce2a0c8cceca354d56957f198a76a20a, 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 0.5.1. A additive run in the scenario-2 arm, on claude-sonnet-5 at served version claude-sonnet-5.
The verdicts
- HW-PROBE-a-session-adds-a-british-spelling-to-the-language-rule-and-declines-the-paragraph-limit-a-current-decision-retired (sufficiency, expects patched)
session L1-additive-scenario-2-p1-r1: not satisfied —
command:r2-paragraph-limitreported over each of the 4 produced artifacts; no call to a tool of an MCP server - HW-PROBE-a-session-adds-a-count-of-findings-by-severity-to-the-json-report-of-the-check (sufficiency, expects patched)
session L1-additive-scenario-2-p2-r1: satisfied — event 12:
command:r2-check-countsreported nothing overdocs/interfaces/headwater-check.md; no call to a tool of an MCP server - HW-PROBE-a-session-adds-a-retired-term-to-the-overlay-that-the-task-names-and-leaves-the-lock-matching-it (sufficiency, expects patched)
session L1-additive-scenario-2-p3-r1: not satisfied —
command:r8-cutting-edgereported over each of the 1 produced artifact; no call to a tool of an MCP server - HW-PROBE-a-session-lets-the-disk-free-script-take-its-floor-in-the-equals-form (sufficiency, expects patched)
session L1-additive-scenario-2-p4-r1: satisfied — event 17:
command:r11-disk-freereported nothing overtools/run/disk-free.sh; no call to a tool of an MCP server - HW-PROBE-a-session-links-the-obligation-on-published-read-sets-to-the-evaluation-whose-measurement-discharged-it (sufficiency, expects patched)
session L1-additive-scenario-2-p5-r1: not satisfied —
command:r5-read-setreported over each of the 1 produced artifact; no call to a tool of an MCP server - HW-PROBE-a-session-makes-the-footer-check-of-the-site-refuse-the-stock-theme-credit-as-well (sufficiency, expects patched)
session L1-additive-scenario-2-p6-r1: satisfied — event 19:
command:r11-site-footerreported nothing overtools/site/check-site-footer.sh; no call to a tool of an MCP server - HW-PROBE-a-session-moves-an-obligation-record-to-the-process-shelf-and-repairs-the-links-inside-it-as-well-as-the-links-to-it (sufficiency, expects patched)
session L1-additive-scenario-2-p7-r1: not satisfied —
command:r7-small-gapsreported over each of the 5 produced artifacts; no call to a tool of an MCP server - HW-PROBE-a-session-moves-the-guide-to-the-merge-driver-for-derived-files-to-the-adopter-shelf-under-the-identifier-that-shelf-mints (sufficiency, expects patched)
session L1-additive-scenario-2-p8-r1: not satisfied —
command:r7-merge-driverreported over each of the 3 produced artifacts; no call to a tool of an MCP server - HW-PROBE-a-session-records-a-ruling-that-drops-the-help-screen-banner-as-the-settled-successor-of-the-record-the-task-names (sufficiency, expects patched)
session L1-additive-scenario-2-p9-r1: not satisfied —
command:r6-bannerreported over each of the 2 produced artifacts; no call to a tool of an MCP server - HW-PROBE-a-session-records-as-owed-that-no-ruling-says-whether-a-hook-s-text-on-a-prompt-is-a-warning (sufficiency, expects patched)
session L1-additive-scenario-2-p10-r1: not satisfied —
command:r4-hook-warningreported over each of the 1 produced artifact; no call to a tool of an MCP server - HW-PROBE-a-session-records-as-owed-the-corpus-size-axis-that-the-additive-campaign-defers (sufficiency, expects patched)
session L1-additive-scenario-2-p11-r1: not satisfied —
command:r4-corpus-sizereported over each of the 2 produced artifacts; no call to a tool of an MCP server - HW-PROBE-a-session-records-the-owl-worked-example-as-evidence-of-the-check-layer-part-that-the-task-names (sufficiency, expects patched)
session L1-additive-scenario-2-p12-r1: satisfied — event 4:
command:r5-check-layer-owlreported nothing overdocs/evaluations/owl-skos-worked-example.md; no call to a tool of an MCP server - HW-PROBE-a-session-renames-the-module-of-the-language-rule-and-moves-the-records-that-govern-it (sufficiency, expects patched)
session L1-additive-scenario-2-p13-r1: not satisfied —
command:r3-language-modulereported over each of the 10 produced artifacts; no call to a tool of an MCP server - HW-PROBE-a-session-renames-the-pattern-module-of-the-meta-crate-and-its-callers (sufficiency, expects patched)
session L1-additive-scenario-2-p14-r1: not satisfied —
command:r3-pattern-modulereported over each of the 11 produced artifacts; no call to a tool of an MCP server - HW-PROBE-a-session-repairs-a-suspect-governs-edge-on-the-record-that-the-task-names-and-raises-no-other-finding (sufficiency, expects patched)
session L1-additive-scenario-2-p15-r1: satisfied — event 7:
command:r9-gitattributesreported nothing overdocs/decisions/0049-a-corpus-wide-fold-is-derived-and-never-stored.md; no call to a tool of an MCP server - HW-PROBE-a-session-retires-a-stock-phrase-in-the-list-that-the-check-reads-and-resolves-the-lock (sufficiency, expects patched)
session L1-additive-scenario-2-p16-r1: not satisfied —
command:r8-game-changerreported over each of the 1 produced artifact; no call to a tool of an MCP server - HW-PROBE-a-session-rewrites-the-change-narration-in-the-summary-of-the-obligation-on-the-site-review-marker-and-raises-no-other-finding (sufficiency, expects patched) session L1-additive-scenario-2-p17-r1: not satisfied — the session produced no artifact, so the oracle had nothing to read; no call to a tool of an MCP server
- HW-PROBE-a-session-says-whether-a-front-page-recording-may-show-a-number-a-live-run-does-not-produce-when-the-task-names-the-record-that-was-superseded (sufficiency, expects answered)
session L1-additive-scenario-2-p18-r1: satisfied — event 10: answered
marked; no call to a tool of an MCP server - HW-PROBE-a-session-supersedes-the-standing-ruling-on-a-paragraph-limit-when-the-task-describes-the-ruling-that-it-replaced (navigability, expects patched) session L1-additive-scenario-2-p19-r1: not satisfied — the session produced no artifact, so the oracle had nothing to read; no call to a tool of an MCP server
- HW-PROBE-a-session-tells-a-withheld-answer-from-an-absent-one-in-a-corpus-that-serves-no-tombstone (sufficiency, expects answered)
session L1-additive-scenario-2-p20-r1: satisfied — event 13: answered
absent; no call to a tool of an MCP server - HW-PROBE-a-session-that-reviews-a-change-shipping-the-vendored-package-as-the-taxonomy-release-cites-the-record-that-forbids-it (sufficiency, expects answered)
session L1-additive-scenario-2-p21-r1: satisfied — event 14: answered
HW-DR-0089; no call to a tool of an MCP server - HW-PROBE-a-session-that-reviews-an-agent-s-acceptance-stamp-cites-the-ruling-in-force-and-not-the-draft-that-the-pull-request-names (sufficiency, expects answered) session L1-additive-scenario-2-p22-r1: not satisfied — the session ended with no answer; no call to a tool of an MCP server
22 of 22 sessions that recorded their calls made no call to a tool of an MCP server. 0 sessions recorded no calls.
The rate, and the denominator it is over
8 of 22 graded sessions satisfied their expectation: 36.4%, in a 95% interval of 19.7% to 57.0%. 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.