release: 0.68.2

[noahgift]

Release bump for 0.68.2 (06x release schedule §4.2): bump-version.sh 0.68.2 across every workspace, and the CHANGELOG section below.

When this merges, rel-0682-autopilot/autopilot.sh runs the rest of the train, fail-closed (APR-RELEASE-001 §4): T-1 deep run, pre-publish dogfood, tag + release, clean-room.yml dispatched on the tag with the run id recorded (T-3), all release assets checked with scripts/check_release_assets.sh, publish preflight, the crates.io cascade (T-4, automated per operator 2026-09-13), install.sh --version receipts on intel and gx10 plus the CUDA asset receipts on gx10 and yoga, and the epic and milestone close.

0.68.2 is an interrupt release for Qwen3 (EPIC #3477): dense Qwen3 produced garbage on CUDA in 0.68.1 because two fast paths dropped the per-head QK-norm — the manual decode graph never recorded the kernel (#3413 A) and the batched prefill never applied it (#3413 B) — and the CPU-vs-GPU guard could not see either because it probed a path generation does not use (#3413 C). All three are fixed and the guard now judges the real prefill path; a fourth finding (FP8 batched prefill fails parity on QK-norm models, #3483) is mitigated by routing those models to the serial prefill with FP8 off. Qwen3.5 real-world files (unsloth UD-IQ2_XXS / IQ4_XS) load and run on the CPU (#3432, #3091), and apr qa certifies a CPU-only architecture instead of aborting. The model-capability ladder (contracts/model-capability-ladder-v1.yaml) is green on lambda (sm_89) and gx10 (sm_121) at the cut; Qwen3-8B's optional rung stays optional pending the golden thinking-budget harness fix (#3486). #3090 (Gated DeltaNet on the GPU) is a dated non-goal carried to 0.69.0.

Added

• feat(ont): ONT-6 — one Verdict lattice, Kani-proved; per-repo arming; exit-vocabulary mapping (PMAT-3451) (feat(ont): ONT-6 — one Verdict lattice, Kani-proved; per-repo arming; exit-vocabulary mapping (PMAT-3451) #3463)
• feat(ont): ONT-2b — Σ with entity_types and extractors, and a symbol-level check on formal: (PMAT-3471) (feat(ont): ONT-2b — Σ with entity_types and extractors, and a symbol-level check on formal: (PMAT-3471) #3472)
• feat(dogfood): model capability ladder — T-2 proves every (architecture, backend, silicon) triple per host; RED at 4a538dd on both GPUs (EPIC: release train 0.68.2 (interrupt) — Qwen3 CPU+GPU verification, P0 #3477) (feat(dogfood): model capability ladder — T-2 proves every (architecture, backend, silicon) triple per host; RED at 4a538ddef on both GPUs (#3477) #3479)

Fixed

• fix(publish): preflight R6 judges the cycle, not the shape (0.68.1 T-4 stop, publish preflight R6 refuses the SHAPE (any versioned sibling dev-dep), not the CYCLE — false positive stops 0.68.1 at T-4 #3468) (fix(publish): preflight R6 judges the cycle, not the shape (0.68.1 T-4 stop, #3468) #3469)
• fix(ladder): a rung that claims only the CPU is not asked whether the GPU can run it (EPIC: release train 0.68.2 (interrupt) — Qwen3 CPU+GPU verification, P0 #3477) (fix(ladder): a rung that claims only the CPU is not asked whether the GPU can run it (#3477) #3481)

Changed

• APEX-2b: extended-Wilkinson tick placement (Talbot, Lin & Hanrahan 2010) with a CRAN-golden set that catches the paper's own erratum — NEW crates/aprender-viz/src/breaks.rs (pub fn extended, extended_loose; Q_DEFAULT, W_DEFAULT=[0.25,0.2,0.5,0.05] per the reference code, not the prose), tests/breaks_golden.rs (44 tests: 36 goldens, 6 properties, anti-vacuity floor, W_DEFAULT-swap mutation), fixtures/breaks/{manifest.json, README.md (R transcription, verbatim), generate.py}; EDITS Cargo.toml (+libm), src/lib.rs (+pub mod breaks), Cargo.lock; no renderer change, no new rendering dep (APEX-2b: extended-Wilkinson tick placement (Talbot, Lin & Hanrahan 2010) with a CRAN-golden set that catches the paper's own erratum — NEW crates/aprender-viz/src/breaks.rs (pub fn extended, extended_loose; Q_DEFAULT, W_DEFAULT=[0.25,0.2,0.5,0.05] per the reference code, not the prose), tests/breaks_golden.rs (44 tests: 36 goldens, 6 properties, anti-vacuity floor, W_DEFAULT-swap mutation), fixtures/breaks/{manifest.json, README.md (R transcription, verbatim), generate.py}; EDITS Cargo.toml (+libm), src/lib.rs (+pub mod breaks), Cargo.lock; no renderer change, no new rendering dep #3259)
• APEX-2a: deterministic render — svg_identical from SVG bytes on X64+ARM64; NEW viz manifest.rs/text.rs/render_determinism.rs (EV-2b breaks as ticks), ci determinism+compare jobs, libm-ban-live.sh; EDITS .clippy.toml (libm bans), breaks.rs powi->sq, svg/lib/scale/plots (APEX-2a: deterministic render — svg_identical from SVG bytes on X64+ARM64; NEW viz manifest.rs/text.rs/render_determinism.rs (EV-2b breaks as ticks), ci determinism+compare jobs, libm-ban-live.sh; EDITS .clippy.toml (libm bans), breaks.rs powi->sq, svg/lib/scale/plots #3273)
• release: 0.68.1 (release: 0.68.1 #3450)
• PMAT-3445: no tag while the milestone being cut holds an open item — check_milestone_cut.sh at the freeze and before the tag (PMAT-3445: no tag while the milestone being cut holds an open item — check_milestone_cut.sh at the freeze and before the tag #3455)
• ci(b2-gpu): aprender-gpu lib tests on hardware at the tag (0.68.1 chain, rule 14) (ci(b2-gpu): aprender-gpu lib tests on hardware at the tag (0.68.1 chain, rule 14) #3467)
• docs(release): v0.68.1 cascade dry-run receipt (T-4, committed before the first upload) (docs(release): v0.68.1 cascade dry-run receipt (T-4, committed before the first upload) #3470)
• docs(release): v0.68.1 SHIPPED — cascade timestamps, install receipts, ledger record, audit interval (docs(release): v0.68.1 SHIPPED — cascade timestamps, install receipts, ledger record, audit interval #3473)
• docs(spec): APR-RELEASE-001 revision 2026-09-17 — fan-out by construction, B2 split, unattended cascade; 0.68.1 SHIPPED (docs(spec): APR-RELEASE-001 revision 2026-09-17 — fan-out by construction, B2 split, unattended cascade; 0.68.1 SHIPPED #3474)
• ledger: clean-room chain at -j8 — 3877 s → 1474 s on v0.68.1, same tests, peak RSS at the 48g ceiling (P0·Fan-out (APR-RELEASE-001 rule 14): release verification fans out by construction — 0.69 does not cut on a serial chain #3475 lever a) (ledger: clean-room chain at -j8 — 3877 s → 1474 s on v0.68.1, same tests, peak RSS at the 48g ceiling (#3475 lever a) #3476)
• PMAT-3477: 0.68.2 — Qwen3 CUDA QK-norm restored in the graph and the batched prefill, FP8 prefill gated for QK-norm models, F2 guard judges the real prefill path; Qwen3.5 real IQ files load and run; apr qa certifies CPU-only qwen35 (P0: dense Qwen3 (Qwen3ForCausalLM) has no current proof it runs — last verification is 7 months stale and CI never covers it #3413, PP-QUANT-001 M4: tensor_byte_size and ONE refusal site read TRAITS — closes the #3091-reopen / #3341 class for all 35 ids at once #3432, CPU: Qwen3.5 / Qwen3.8 hybrid GGUFs are refused entirely — no fallback exists #3091) (PMAT-3477: 0.68.2 — Qwen3 CUDA QK-norm restored in the graph and the batched prefill, FP8 prefill gated for QK-norm models, F2 guard judges the real prefill path; Qwen3.5 real IQ files load and run; apr qa certifies CPU-only qwen35 (#3413, #3432, #3091) #3484)

🤖 Generated with Claude Code

Issue refs (§6 R-2)

keep-open: #3477 — the release epic closes at the end of the train (T-5 reconcile), not at the bump.
keep-open: #3090 — Gated DeltaNet on the GPU is a dated non-goal for 0.68.2, carried to 0.69.0 (slipped_from comment on the issue).

R-23: aprender: cmd="pv lint contracts --gate shapes --format json" pv=b1c0f122a exit=0 verdict="Pass, 0 armed violations (98 unarmed warnings on bound-symbols-resolve, not armed); armed ladder-measured, ladder-green, ont-shapes-v1; pv extract contracts --check rc=0; check_model_ladder.sh: every required rung green on every required host"

[github-actions]

§13.11 rung 1 — quorum shadow verdict

S13-SHADOW pr=3498 head=b1c0f122a3f80908a4e6755228ff87d1f85ef757 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.

[alfredodeza]

Edited the body's no-close: lines to keep-open: — GitHub's closing-keyword parser reads close: inside no-close: (the hyphen is a word boundary to it), so those two lines would have closed #3477 and #3090 on merge despite the negating prefix. Confirmed against the fixed body with scripts/check_pr_closes_issue.sh: FAIL before the edit (both flagged), PASS after. Same mechanism that closed #3090/#3477 twice today (#3400, fixed in #3497, now on main) — this PR was opened before that fix landed. Also ran gh pr update-branch to pick it up so CI validates against current main rather than a stale merge-ref.

🤖 Generated with Claude Code

[alfredodeza]

Holding this release. Auto-merge disarmed — do not re-arm without an explicit operator decision.

Operator ruling 2026-09-18 (this thread, #3477, #3090): GPU (#3090) is required for the 0.68.2 interrupt, not an optional non-goal. This PR's own changelog still reads "#3090 (Gated DeltaNet on the GPU) is a dated non-goal carried to 0.69.0" — that disposition was explicitly withdrawn. 0.68.2 does not cut until GPU is addressed (landed, or in progress with a credible plan the operator has signed off on), not merely until CPU is done.

CPU is genuinely ready (#3413/#3091/#3432 closed). GPU is not: #3413's prerequisite fix is merged, but the GDN device kernel itself has no branch or code anywhere yet.

Also already fixed on this PR: the no-close: landmine text (#3400) and the stale branch (see prior comments). Those are done; the GPU scope decision is what's holding this now.

🤖 Generated with Claude Code