Skip to content
Vox: The AI-Native Programming Language
Unverified: This page has not yet been anchored to automated code tests.

v1-release-criteria

Vox v1.0 Release Criteria (Hardened)

Section titled “Vox v1.0 Release Criteria (Hardened)”

To reach a stable v1.0, the Vox foundation must satisfy the criteria below. They are evaluated in tier order: the Foundation tier (CR-F) gates the Distribution (CR-K), Desktop GUI (CR-U), and Product (CR-P/A/E/D/L) tiers. Per [CR-F0], no downstream-tier criterion may report met: true while any Foundation criterion is red.

Why this revision (2026-06-05). The prior criteria gated only outcomes built on top of Vox (deploy loops, uptime, LLM authorship, bundle size) and never the compiler’s own completeness or correctness. A 6-dimension forensic audit (see v1-foundation-criteria-research-2026.md) showed core-fix activity accelerating, table-stakes semantics landing at HEAD, the golden corpus verified by parse/typecheck only, four hand-synchronized execution arms with no differential test, and a non-existent crate-publish story. The Foundation/Distribution/GUI tiers and the machine-readable format below close that gap.

Implementation status (2026-06-05): this is a TARGET spec, not current state. Most Foundation (CR-F), Distribution (CR-K), and GUI (CR-U) gate harnesses are unbuilt — their verify_cmds reference vox audit --gate targets that are not yet registered. audit --gate all --strict-block-ga therefore exits non-zero today, by design. (CR-F1’s harness landed 2026-06-05 — see its status line.) Each criterion’s if_failing field is the build pointer. Do not mark any new-tier gate met: true until its harness lands and writes a real artifact.

0. Machine-readable format (CR-META)

Section titled “0. Machine-readable format (CR-META)”

Every criterion carries three machine-readable fields so an LLM agent can drive it to completion without human interpretation:

  • verify_cmd — the exact command that exits 0 on pass, non-zero on fail.
  • artifact_path — the JSON breakdown the gate writes (per-item results).
  • if_failing — the pointer to the plan section / fixture dir to build next.

[CR-META] The criteria doc fails its own CI lint (crates/vox-arch-check/src/evidence_ledger_check.rs, extended) if any [CR-*] block is missing a verify_cmd, a resolvable artifact_path that maps to a registered gate, or a non-empty if_failing pointer.

  • verify_cmd: cargo run -p vox-arch-check -- --lint criteria-format
  • artifact_path: contracts/reports/arch/criteria-format/<UTC>.json
  • if_failing: add the missing field to the offending criterion block.

Marquee app set. “Marquee app” references resolve to the canonical fixture set at contracts/marquee/manifest.v1.yaml (exists). Criteria [CR-P1], [CR-P3], [CR-E2], [CR-L0], [CR-L7] depend on it.

Tier 0 — Foundation: the language itself is complete and correct

Section titled “Tier 0 — Foundation: the language itself is complete and correct”

Evaluated first. While any CR-F is red, all downstream tiers report blocked_by_foundation.

[CR-F0] Foundation-first ordering. vox audit --gate all evaluates every CR-F gate before any CR-K/CR-U/CR-P/A/E/D/L gate, and reports downstream gates as blocked_by_foundation (never met: true) whenever any CR-F.met == false.

  • verify_cmd: cargo run -p vox-cli -- audit --gate all --strict-block-ga
  • artifact_path: contracts/reports/_snapshot/<UTC>.json (tier:"foundation" rows first)
  • if_failing: implement gate ordering in crates/vox-audit/src/registry.rs; add a vox-arch-check rule asserting foundation gates sort first.

[CR-F1] Behavioral goldens. Every .vox under examples/golden/ that produces output must execute and match a committed // EXPECT: block (or an @test fn); zero parse/typecheck-only goldens remain. Coverage of top-level goldens by // EXPECT ∪ @test must be 1.0.

  • verify_cmd: cargo test -p vox-integration-tests --test golden_behavioral_gate
  • artifact_path: contracts/reports/behavioral-goldens/<UTC>.json
  • if_failing: extend the // EXPECT harness coverage toward 1.0 and add the ratchet assertion. Status: harness LANDED 2026-06-05 (crates/vox-integration-tests/tests/golden_behavioral_gate.rs, green) — 7 EXPECT goldens execute+match; the harness’s first catch was a real bug (vox run --mode interp Debug-printed main’s return value as Str("ok") — fixed in run.rs). Remaining: author EXPECT/@test for the ~41 behaviorally-uncovered goldens (printed by the census test) and flip coverage to a hard 1.0 ratchet.

CR-F2 — Arm correctness (split by domain, not “three identical renderings”). Vox emits along a domain boundary (maintainer steer 2026-06-07): logic → Rust; browser/GUI → TypeScript; never both. So CR-F2 is not “every golden runs identically under interp/script/ts” — codegen-ts is a web/GUI emitter that does not (and should not) render a stdout logic program. CR-F2 splits into three independent criteria (CR-F2a/b/c). See codegen-ts-domain-boundary-and-cr-f2-correction-2026.md.

[CR-F2a] Logic parity (interp ≡ codegen-rust). Every examples/golden/** program with fn main + // EXPECT: produces byte-identical stdout under --mode interp and --mode script (codegen-rust). The divergence allowlist is non-growing (ratchet to empty).

  • verify_cmd: cargo test -p vox-integration-tests --test golden_arm_parity_test -- --ignored
  • artifact_path: contracts/eval/arm-parity-allowlist-script.txt (committed ratchet; empty = met)
  • if_failing: fix the codegen-rust emit site for the diverging construct, then remove its line from the allowlist. Status: MEASURED 2026-06-07 — codegen-rust 3/10 (pass: decimal_math, mesh/noop, while_loop_algorithms); 7 in the allowlist with verified rustc errors.

[CR-F2b] Web emit correctness (codegen-ts). Every examples/golden-ts/** web fixture emits TypeScript that type-checks and behaves correctly in a browser/DOM.

  • verify_cmd: cargo test -p vox-integration-tests --test ts_emit_typecheck_test -- --ignored (+ a planned jsdom/Playwright behavioral pass)
  • artifact_path: contracts/reports/ts-web-emit/<UTC>.json
  • if_failing: fix the codegen-ts emitter for the failing fixture; add the behavioral assertion. Status: typecheck gate exists (tsc --noEmit); behavioral DOM gate NOT yet built.

[CR-F2c] Split discipline (the boundary itself). Each top-level construct lands in exactly one arm — logic in Rust only, browser/GUI in TS only — nothing silently dropped by both or emitted by both.

  • verify_cmd: cargo run -p vox-cli -- audit --gate emit-routing
  • artifact_path: contracts/reports/emit-routing/<UTC>.json
  • if_failing: add the construct to the routing classification table and route it to its single correct arm; fix any drop/duplication. Status: gate NOT yet built; known wart — codegen-ts emits empty app boilerplate for logic-only programs.

[CR-F3] Language-spec coverage. A machine-readable checklist (contracts/spec/language-surface-coverage.v1.yaml) enumerates every grammar production, decorator, and builtin and maps each to ≥ 1 passing behavioral fixture; zero uncovered rows and zero incomplete-arm rows. A vox-arch-check rule fails if a new grammar production lands without a row.

  • verify_cmd: cargo run -p vox-cli -- audit --gate spec-coverage
  • artifact_path: contracts/reports/spec-coverage/<UTC>.json
  • if_failing: add the missing fixture or finish the named arm for the uncovered construct. Status: checklist file does not yet exist.

[CR-F4] No incomplete arms. Zero reachable todo!() / unimplemented!() / “not yet” / runtime-UnsupportedOnPlatform markers for any construct marked supported in CR-F3. A construct an arm cannot lower must fail at codegen time (the codegen_rust WASI compile_error! pattern), never at runtime. db.* operations and intra-project import "./x.vox" are carved out of the supported set for the script/ts arms and MUST be rejected there with a stable codegen-time diagnostic (council steer 2026-06-05: reject, do not silently accept), while remaining fully executable under --mode interp.

  • verify_cmd: cargo run -p vox-cli -- audit --gate no-incomplete-arms
  • artifact_path: contracts/reports/no-incomplete-arms/<UTC>.json
  • if_failing: implement the lowering arm, or convert the runtime throw into a codegen-time diagnostic. For the carve-out, add vox/codegen/db-unsupported-here + vox/codegen/local-import-unsupported-here diagnostics in the script/ts arms. Status: codegen-ts currently runtime-throws for typecheck-passing db ops (hir_emit/mod.rs:1050).

[CR-F5] Convergence. Core-fix commits per rolling 2-week window decline for 3 consecutive windows AND the final window is ≤ 25% of the peak window; AND the release-tagged commit body contains zero first-time-semantics entries (no “implement … interpreter”, “parity”, “exhaustiveness”, “de-stub”, or “NNN green” test-count thrash) and touches no crates/vox-compiler/src/{eval,hir} or crates/vox-codegen/src file beyond version bumps.

  • verify_cmd: cargo run -p vox-cli -- audit --gate core-convergence
  • artifact_path: contracts/reports/core-convergence/<UTC>.json
  • if_failing: keep fixing the core until the curve bends; do not tag a release whose own commit lands new semantics. Status: today the curve is rising (core-fix share 5.2%→19.8% Mar–Jun); HEAD #137 landed first-time semantics.

[CR-F6] Regression budget. Zero // vox:skip, zero de-stub-pending mocks, zero stub/placeholder returns in crates/vox-compiler, crates/vox-codegen, and the golden corpus; the count is non-increasing between tagged releases.

  • verify_cmd: cargo run -p vox-cli -- audit --gate regression-budget
  • artifact_path: contracts/reports/regression-budget/<UTC>.json
  • if_failing: replace the stub with a real impl, or scope the construct out of the CR-F3 supported set.

Tier 1 — Distribution: Rust users can install Vox and its crates

Section titled “Tier 1 — Distribution: Rust users can install Vox and its crates”

[CR-K1] Public set publishes clean. Every crate in the declared public set passes cargo publish --dry-run (no missing-version, missing-license, path-only-dep, or publish=false-transitive errors). (Today: 1/103 — vox-crypto only.)

  • verify_cmd: cargo run -p vox-cli -- audit --gate crate-publish (iterates crates/_public.toml, runs cargo publish --dry-run -p <N>)
  • artifact_path: contracts/reports/crate-publish/<UTC>.json
  • if_failing: fix the exact dependency/field cargo names per crate; strip workspace-hack from the public set’s publish surface.

[CR-K2] Public-set manifest + metadata. A canonical crates/_public.toml lists the external-publication set; every listed crate has non-empty description, license, repository, readme.

  • verify_cmd: cargo run -p vox-cli -- audit --gate public-set-metadata
  • artifact_path: contracts/reports/public-set-metadata/<UTC>.json
  • if_failing: create crates/_public.toml; add license.workspace = true + metadata to each listed crate (~34 omit license today; vox-cli sets license = false).

[CR-K3] No publish-false deps; versioned intra-deps. No public crate depends on a publish=false crate (incl. workspace-hack) at publish time; intra-workspace deps in the public set carry a version requirement.

  • verify_cmd: cargo run -p vox-cli -- audit --gate publish-dep-hygiene
  • artifact_path: contracts/reports/publish-dep-hygiene/<UTC>.json
  • if_failing: declare { path, version } on workspace deps; add publish-time workspace-hack stripping (75/103 manifests carry it today).

[CR-K4] voxup installs a real toolchain. voxup install default produces a working vox binary (not the current proxy-wrapper stub) and the installed vox --version matches the released version.

  • verify_cmd: HOME=$tmp voxup install default && $tmp/.vox/bin/vox --version (CI; asserts not the Vox Proxy Wrapper stub)
  • artifact_path: contracts/reports/voxup-install/<UTC>.json
  • if_failing: implement artifact download + checksum verification + real exec in crates/voxup/src/install.rs.

[CR-K5] SemVer policy + enforcement. A documented SemVer/public-API policy exists and cargo-semver-checks runs in CI for the public set, blocking unannounced breaking changes.

  • verify_cmd: cargo semver-checks check-release over crates/_public.toml
  • artifact_path: contracts/reports/semver/<UTC>.json
  • if_failing: author docs/src/contributors/semver-policy.md (with frontmatter); add the CI gate.

[CR-K6] Publish workflow. A .github/workflows/publish-crates.yml publishes the public set in reverse-topological order on v* tags, gated on CR-K1.

  • verify_cmd: cargo run -p vox-cli -- audit --gate publish-workflow (asserts workflow exists, triggers on v*, runs the dry-run gate first)
  • artifact_path: contracts/reports/publish-workflow/<UTC>.json
  • if_failing: add the workflow; order jobs by dependency topology.

[CR-K7] Signed/verifiable binary artifacts. Release binaries ship a SHA-256 checksum/provenance manifest; vox-checksum-manifest and vox-release-artifacts are landed (not [planned]) and wired into the release workflows.

  • verify_cmd: cargo run -p vox-cli -- audit --gate release-provenance
  • artifact_path: contracts/reports/release-provenance/<UTC>.json
  • if_failing: promote the two [planned] crates in layers.toml; emit + verify a manifest per asset.

Recommended public set (dependency order): vox-crypto, vox-jsonschema-util, vox-telemetry, vox-journal, vox-git, vox-grammar-export, vox-db-types → vox-db. Compiler/orchestrator crates stay internal until their API surface is deliberately frozen.

Tier 2 — Desktop GUI: VOX GUI installs and every surface works

Section titled “Tier 2 — Desktop GUI: VOX GUI installs and every surface works”

[CR-U1] Surface-registry gate bites in CI. vox ci gui-surface-registry runs as a required CI gate and fails on any drift or wiring violation. (Today: runs in zero workflows.)

  • verify_cmd: required CI job runs vox ci gui-surface-registry; a mutation must turn it red
  • artifact_path: contracts/reports/gui-surface-registry/<UTC>.json
  • if_failing: add the job to a .github/workflows/*.yml listed in the ci-summary needs: array.

[CR-U2] Every surface renders. Every live_backend / curated_decorator surface renders a non-empty, non-error panel in a headless run of the actual vox-gui frontend (not the emitted codegen app); count(surfaces_tested) == count(non-none registry entries).

  • verify_cmd: pnpm --dir crates/vox-gui/ui test:e2e -- surface-reachability
  • artifact_path: contracts/reports/gui-surface-reachability/<UTC>.json
  • if_failing: map each registered surface to a real panel; fix renderView default: return null blanking.

[CR-U3] vox-gui’s own suites run in CI. vox-gui’s Playwright e2e and vitest suites run as required CI gates.

  • verify_cmd: pnpm --dir crates/vox-gui/ui test && pnpm --dir crates/vox-gui/ui test:e2e
  • artifact_path: contracts/reports/gui-suites/<UTC>.json
  • if_failing: add both to a required CI job (today neither runs in CI).

[CR-U4] Installers build. tauri.conf.json is configured to produce installers for all three platforms with a complete icon set (.ico/.icns + PNG sizes); a CI dry-run yields a non-empty bundle.

  • verify_cmd: full pipeline (cargo build -p vox-cli --release && pnpm --dir crates/vox-gui/ui build && tauri build) → assert ≥ 1 installer artifact
  • artifact_path: contracts/reports/gui-bundle/<UTC>.json
  • if_failing: add bundle.active/targets + the icon set (today: only icon.png, no targets).

[CR-U5] Signed + verified release. release-gui.yml builds the sidecar binaries before bundling and signs+verifies the installer using existing paths (no src-tauri/).

  • verify_cmd: cargo run -p vox-cli -- audit --gate gui-release (asserts a cargo build step precedes bundling, a signature-verify step exists, and no src-tauri path)
  • artifact_path: contracts/reports/gui-release/<UTC>.json
  • if_failing: add the sidecar build step; fix the Windows signing files-folder path; add signtool verify / codesign --verify.

[CR-U6] Launch + IPC smoke. A smoke test proves the packaged app starts and the real Tauri invoke handlers (not the mocked e2e bridge) answer for a representative read-only command.

  • verify_cmd: cargo run -p vox-cli -- audit --gate gui-smoke
  • artifact_path: contracts/reports/gui-smoke/<UTC>.json
  • if_failing: add a headless launch harness invoking get_build_info / execute_command through the real handlers.

Tier 3 — Product (evaluated only when Tiers 0–2 are green)

Section titled “Tier 3 — Product (evaluated only when Tiers 0–2 are green)”

These criteria are unchanged in intent from the original council-ratified set; they are re-tiered to sit behind the Foundation gate. Criteria flagged external_infra: true cannot be self-completed by an agent from a checkout and are excluded from the autonomous-completion loop.

3.1 Production Validation

Section titled “3.1 Production Validation”
  • [CR-P1] ≥ 3 Marquee applications deployed live on OCI-compliant infra with zero manual config. verify_cmd: vox audit --gate cr-p1 · artifact_path: contracts/reports/perf/cr-p1/<UTC>.json · if_failing: §5.2 of the honest-completion plan · external_infra: true.
  • [CR-P2] 99.9% uptime for the vox-ml-cli inference endpoint over a 7-day soak. verify_cmd: vox audit --gate cr-p2 · artifact_path: contracts/reports/perf/cr-p2/<UTC>-7day.json · if_failing: §5.3 · external_infra: true.
  • [CR-P3] vox new web → vox deploy under 120s end-to-end. verify_cmd: vox audit --gate deploy · artifact_path: contracts/reports/deploy/<UTC>.json · if_failing: §3.2 · external_infra: true (live deploy leg).

3.2 Architectural Integrity

Section titled “3.2 Architectural Integrity”
  • [CR-A1] K-Complexity Freeze: cyclomatic complexity < 15 on all primary lowering paths. verify_cmd: vox audit --gate cr-a1 · artifact_path: contracts/reports/arch/cr-a1/<UTC>.json · if_failing: refactor the 14 functions currently > budget (max 28).
  • [CR-A2] Non-Null Boundary: 100% of FFI/IPC interfaces use non-null VoxProto v1 schemas. verify_cmd: vox audit --gate cr-a2 · artifact_path: contracts/reports/arch/cr-a2/<UTC>.json · if_failing: annotate remaining boundaries (currently 186/186 = met).
  • [CR-A3] Crate Decoupling: zero circular deps across contracts/db/data-storage-policy.v1.yaml#frozen_core_crates. verify_cmd: cargo run -p vox-arch-check · artifact_path: contracts/reports/arch/cycles/<UTC>.json · if_failing: break the offending edge.
  • [CR-A4] Lifecycle Metadata Parity: orchestration contracts declare lifecycle + migration window with CI parity checks. verify_cmd: vox audit --gate cr-a4 · artifact_path: contracts/reports/arch/cr-a4/<UTC>.json · if_failing: add lifecycle metadata to the bare contract.

3.3 Performance & Efficiency

Section titled “3.3 Performance & Efficiency”
  • [CR-E1] Cold Start: vox run --interp initializes + executes Hello World < 50ms. verify_cmd: vox audit --gate cr-e1 · artifact_path: contracts/reports/perf/cr-e1/<UTC>.json · if_failing: profile interp init (currently p99 = 0.25ms = met).
  • [CR-E2] Bundle Size: Marquee bundle (React + TanStack) ≤ 800KB gzip. verify_cmd: vox audit --gate cr-e2 · artifact_path: contracts/reports/perf/cr-e2/<UTC>.json · if_failing: add the build-time bundle gate in vox build.
  • [CR-E3] Training Parity: vox-populi training reaches loss parity with reference PyTorch/LoRA on the vox-lang corpus. verify_cmd: vox audit --gate cr-e3 · artifact_path: contracts/reports/perf/cr-e3/<UTC>.json · if_failing: tune the native training loop · external_infra: true (GPU).

3.4 Agentic DX

Section titled “3.4 Agentic DX”
  • [CR-D1] Planning Mode Fidelity: ≥ 85% multi-step “Wave 2” plan success without intervention (harness = [CR-L4]). verify_cmd: vox audit --gate plan-fidelity · artifact_path: contracts/reports/plan-fidelity/<UTC>.json · if_failing: port the refinement loop; corpus §4.3 (currently 0.40 vs 0.85).
  • [CR-D2] Self-Healing: vox repair resolves 90% single-file / 70% project-scope (project variant = [CR-L3]). verify_cmd: vox audit --gate repair-corpus · artifact_path: contracts/reports/repair-corpus/<UTC>.json · if_failing: expand corpus §4.2.
  • [CR-D3] Documentation Coverage: 100% of vox-cli subcommands have machine-readable help + a .vox example. verify_cmd: vox audit --gate cr-d3 · artifact_path: contracts/reports/arch/cr-d3/<UTC>.json · if_failing: author the ~60 missing examples (currently 8/68 = 11.8%).

3.5 LLM-Target Fidelity

Section titled “3.5 LLM-Target Fidelity”

These operationalize the claim that Vox is shaped so AI agents can author code reliably. Full audit: vox-as-llm-target-audit-and-plan-2026.md. Implementation: v1-llm-target-implementation-plan-2026.md.

  • [CR-L0] End-to-End Agent Authorship Loop: autonomous agent loop over contracts/eval/spec-to-app/ produces passing apps (vox check clean, tests pass, vox deploy succeeds, vox doctor green) at ≥ 60% (sub-bar < 40%). verify_cmd: vox audit --gate spec-to-app · artifact_path: contracts/reports/spec-to-app/<UTC>.json · if_failing: §3.7 (currently 0.667 = met, generate→check loop only).
  • [CR-L1] HumanEval-Vox: 164-program suite reaches ≥ 80% behavioral compile+test-pass when prompted to MENS or a reference LLM. The gate now runs vox run --mode interp per tests.vox. Split: L1a (corpus integrity, reference solutions, met 164/164) vs L1b (LLM-generated pass@k ≥ 0.80 — BLOCKING, harness per_llm currently empty). verify_cmd: vox audit --gate humaneval --llm-panel · artifact_path: contracts/reports/humaneval/<UTC>.json · if_failing: wire the LLM-panel generation phase (humaneval.rs P2.4+).
  • [CR-L2] On-Distribution Rate: ≥ 95% of MENS-emitted Vox clears vox check --strict + the vox-code-audit + retirement-guard. verify_cmd: vox audit --gate mens-on-distribution · artifact_path: contracts/reports/mens-on-distribution/<UTC>.json · if_failing: §3.4 MENS sampling (NOT YET FILLED).
  • [CR-L3] Project-Scope Self-Healing: vox repair . ≥ 70% on the 50-project corpus. verify_cmd: vox audit --gate repair-corpus · artifact_path: contracts/reports/repair-corpus/<UTC>.json · if_failing: §4.2 (currently 0.80 on 5 projects).
  • [CR-L4] Plan-Mode Fidelity Measurement: Wave-2 fixtures at contracts/eval/plan-fidelity/ with an automated 85% harness. verify_cmd: vox audit --gate plan-fidelity · artifact_path: contracts/reports/plan-fidelity/<UTC>.json · if_failing: §3.6/§4.3.
  • [CR-L5] ACI Envelope Default-On: agentos_aci_envelope_enabled defaults true; guardrail kernel rejects unclassified mutations at v1.0. verify_cmd: vox audit --gate aci-default · artifact_path: contracts/reports/aci-default/<UTC>.json · if_failing: flip the default (met).
  • [CR-L6] Retirement-Guard Parity: every row in AGENTS.md §Retired Surfaces has a detector or arch-check rule. verify_cmd: vox ci retirement-audit · artifact_path: contracts/reports/retirement/<UTC>.json · if_failing: add the missing detector (met 16/16).
  • [CR-L7] Deploy CLI Completeness: vox new/vox deploy/vox doctor ship structured JSON + telemetry + a CI integration test inside the 120s budget. verify_cmd: vox audit --gate deploy · artifact_path: contracts/reports/deploy/<UTC>.json · if_failing: §3.2 (met for all three legs).
  • [CR-L8] Diagnostic→Repair→Corpus Feedback Loop: quarterly telemetry export into vox-corpus runs in CI. verify_cmd: vox audit --gate corpus-feedback · artifact_path: contracts/reports/corpus-feedback/<UTC>.json · if_failing: §3.1 (met; CI fails if artifact > 90 days).

Acceptance for v1.0 GA

Section titled “Acceptance for v1.0 GA”
Terminal window
cargo run -p vox-cli -- audit --gate all --strict-block-ga

Must exit 0. The roll-up at contracts/reports/_snapshot/<UTC>.json must show every Foundation (CR-F), Distribution (CR-K), and GUI (CR-U) gate at met: true, AND every block-GA Product gate (external_infra: false) at met: true and incomplete: false. Per [CR-F0], the snapshot must not report any downstream gate green while a Foundation gate is red.

Original criteria (CR-P/A/E/D/L) approved by Vox Foundation Council — April 2026; §3.5 (CR-L0..L8) + Marquee manifest ratified 2026-05-15. Foundation tier (CR-F), Distribution tier (CR-K), GUI tier (CR-U), CR-META format, and tier re-ordering added 2026-06-05 per v1-foundation-criteria-research-2026.md; arm-parity scope (all three arms) and the db.*/import reject-with-diagnostic disposition reflect the 2026-06-05 maintainer steer. Council ratification of the new tiers is pending. Per ratification D2: realistic-v1.0 bars hold. Per D16: mesh Phase 2 LAN demoted to v1.1.