Release 0.0.2

[aksOps]

Cuts 0.0.2: version bump, ## [Unreleased] closed as ## [0.0.2] - 2026-07-29, compare links updated.

Release preconditions, checked rather than assumed

• CI green on all three platforms at 8bf0c90 — CI, Corpus gates and Push on main all success.
• All 29 corpus gates and all 15 target pins pass.
• Resource ceiling met. The coverage wave took cold RSS to 158.8% of the hard 512 MiB limit; Stop holding the whole tree's references: cold RSS 830,612 kB to 286,780 kB #55 brought it to 286,780 kB — 54.7% on 2 vCPU reference hardware over a 5,353,211-line tree. This was measured, not assumed, and the release was held until it passed.
• cargo package --list inspected by hand: 126 files — 119 under src/, plus Cargo.toml, Cargo.lock, Cargo.toml.orig, README.md, CHANGELOG.md, .cargo_vcs_info.json, and the repository's own MIT LICENSE. No corpus path, no vendored licence, nothing outside the anchored include block. 0.0.1 shipped 21 licence files out of a private corpus repository through an unanchored pattern, so this list is read rather than trusted.
• cargo publish --dry-run rc 0.

What is in 0.0.2

Coverage, honesty and one capability the gate could not previously see:

• Go emits type uses, non-call selector reads and composite-literal keys — references per tree up 2.82×.
• One locals policy across every tier-1 language, every baseline deliberately re-based.
• Ambient globals reclassified out of NoMatchingDefinition; tsconfig customConditions implemented — zod 27.2% → 46.3%.
• A probe corpus per tier-1 language pinning method-call outcomes by name.
• arthron pin: a committed target pin per corpus, so a reference that resolves to the wrong target fails CI. The rate gate compares four integers and could never see it.
• The call-graph claim retracted, including the crates.io keyword; the rate's denominator share now published beside every rate.

Known misses are recorded in the README rather than left to be inferred: TypeScript cold throughput is outside the 60 s / 1M line target at 70.1 s and 82.5 s, and warm re-index is 12.37 s against a 1 s target. Resource ceilings are hard; timing is a target.

🤖 Generated with Claude Code

https://claude.ai/code/session_01SF9KvRKVzA2vEYqtFq23zm