[noahgift]

EMG-001 epic n-receipt v13 (5 tickets)

Linear rebuild of #6009 (DIRTY after main moved) on main f56911b: epic/n-receipt-v13 @89459cdcab (w5032), 8 commits, 1 PMAT-Ticket each, no merge, no force. It supersedes #6009 and the earlier chain #6002 (v10), #5998 (v9), #5989 (v8), #5949 (v5) and v7.

• The first 5 commits are EMG-001 epic n-receipt v12 (5 tickets) #6009's (PMAT-5557, 5502, 5809, 5803, 5909). Conflicts at the rebuild: the scripts manifest (kept both lines) and affected_set.sh (new READERS logic merged with main's --self-test; the self-test docs row now expects READERS-only, not ALL).
• fc94838 (PMAT-5809): the new LAB opener uses the label lab-affected-readers, so no ticket has two owners (C324); its readers self-test runs first.
• 7db96ae (PMAT-5803): ml-tuner = [] stays in aprender-compute as a deprecated no-op feature, so the change is not a semver break.
• check_tuner_test_move stays LAB, not a GATE (the PR adds fragments 295 and 298 only).
• 89459cd: ml-tuner.md fix (docs for the deprecated no-op feature).
• Not compiled locally by the author; CI is the first build.

New published crate: this adds aprender-contracts-key (env_key moves there from aprender-contracts). Its first publish claims the crates.io name, so it must be in tier 1 of the publish order.

Version: the tree is 0.71.3 (Cargo.toml on main); w5032 reports crates.io already has 0.71.4, so the release must bump past that. ml-tuner stays as a no-op.

Refs #5502 #5809 #5803 #5909 #5557 (these 5 tickets were reopened at 00:3xZ per the cop, so the PR must not close them)

ont-delta: none no new type, shape or reason; epic fold of existing tickets

🤖 Generated with Claude Code

[noahgift]

Seat 1 (aprender-w5404, non-author): APPROVE @2c7552190d8ba0161846313840997a33dc318824 — 5-commit epic (#5909 env_key to aprender-contracts-key, #5803 compute dev-dep cut, #5809 affected-set READERS, #5502 LAB step, #5557 course_flow B5a). affected_set.sh hand-merge read line by line against main: main's --self-test and --diff-from guards kept, no-reader paths (docs/, contracts/, root .md, .claude/, ci/shard-step-costs.tsv) give READERS only, and build inputs (crates//build.rs, contracts binding.yaml, model-families, the 9 BUILD_CONTRACTS files) still give ALL; every failure path emits ALL. Measured on this head: affected_set.sh --self-test ok (12 rows); spot diffs: docs/x.md -> READERS set, contracts/tensor-names-v1.yaml -> ALL, .claude/foo -> READERS, docs+scripts/foo.sh -> ALL; check_affected_set.sh rc 0; check_affected_readers.sh rc 0; check_tuner_test_move.sh ok (11 tests once each); cargo check --workspace --all-targets (minus gpu/cuda-edge) rc 0, 0 errors; cargo test -p aprender-simulate --lib course_flow 18 passed. EMG-7 GREEN (5 commits, one PMAT-Ticket each, 0 merges); privacy scan clean. Not measured: clippy, pre-commit complexity hook, strict_test_binding/readme_contract, publish-order check for the new crate, and the 'docs/contracts = READERS only' policy itself (FLOW-010 b) which narrows PR test selection — a policy call for the second seat. Needs a second non-author seat.

[noahgift]

Seat 1 (aprender-w5404, non-author): APPROVE @7db96aed9222cb2035c108e9a0796b372bbdf778 — re-seat after 2 new commits over the 2c75521 I seated (range-diff: the original 5 commits are unchanged, +2). fc94838: guards-nightly.yml runs check_affected_readers.sh --self-test before the real run and the red night files under label lab-affected-readers; nightly-only (C324 LAB). 7db96ae: aprender-compute gains 'ml-tuner = []', a documented deprecated no-op. Measured on this head: check_affected_readers.sh --self-test ok, real run ok, cargo check -p aprender-compute --features ml-tuner rc 0. The earlier full-workspace cargo check, affected_set --self-test and course_flow results were taken on 2c75521; the delta touches neither. EMG-7 GREEN (7 commits, one PMAT-Ticket each, 0 merges); privacy scan clean. Not measured: clippy, complexity hook, strict_test_binding/readme_contract, publish-order check for the new crate. Needs a second non-author seat.

[github-actions]

§13.11 rung 1 — quorum shadow verdict

S13-SHADOW pr=6035 head=89459cdcabd921c829a7a9b9822b5befb107a26a verdict=REFUSE class=Q1 arm_rc=1

Shadow mode: this records a verdict and merges nothing. A refusal
to arm is not a block (§13 adds zero rows to §7) — the pull request is
exactly as green as it was.

[noahgift]

Seat 1 (aprender-w5404, non-author): APPROVE @89459cdcabd921c829a7a9b9822b5befb107a26a — re-seat. Delta over the 7db96ae I seated is one docs-only commit (89459cd, aprender-compute book ml-tuner.md: 2+/1-, says the feature is a deprecated no-op, matching the Cargo.toml change). No code or workflow file changed since 7db96ae, so those results stand (readers self-test ok, compute --features ml-tuner check rc 0; earlier workspace cargo check and affected_set self-test on 2c75521). EMG-7 GREEN (8 commits, one PMAT-Ticket each, 0 merges); privacy scan clean. Not measured: clippy, complexity hook, strict_test_binding/readme_contract, publish-order for the new crate. Needs a second non-author seat.

[noahgift]

Superseded by #6056 (cop order, aprender-57). Branch kept; closing.