[voltcrash]

Provide the platform-independent Rust engine and generated browser Wasm bindings for the existing typing behavior. Stage 1 of three stacked PRs; the website still uses its TypeScript engine until Stage 2, followed by the offline Ratatui app in Stage 3.

Rust owns input/progression/timing, generation/transforms, shared configuration policies, scores/results, replay, pace and learning. Hosts supply UTF-16 text, clocks, content and observations. The facade provides owner-scoped initialization, disposal, structured recovery, generation-tagged asynchronous sources and bounded incremental updates. Original reference snapshots remain unchanged; actual-web/headless discrepancies are explicitly classified.

Verification:

• Native suite, Clippy and formatting pass. Parity covers 19 command traces, all 61 original generator snapshots, section/fractional/live rules, quote queue RNG, eight historical results and 1,387 live prefixes.
• Built Cloudflare proof: 24 browser cases pass, including schema compatibility, eight unchanged transport hashes/backend outcomes, Unicode history/replay, lifecycle and download/version failures.
• Malformed/bounded DTOs and 128-seed / 12,800-command lifecycle stress pass; backend suite passes 787 tests.
• Optimized Wasm: 755,466 raw / 301,279 gzip6 bytes, within frozen budgets. Final timing rerun passes: initialization p95 45.1 ms, boundary p95 0.01050 ms/unit; 13,540-byte deployed facade. Baseline reset retains 9,240,576 Wasm bytes.
• Isolated e7b7ad8 checkout with deleted target/generated/dist passes the actual Cloudflare build and all workspace typechecks. Earlier first-test/typecheck bootstrap also passes. Full CI passes at final Stage 1 head dc84395: engine/native/Wasm, frontend/backend/packages, all browser shards, SSR, delivery, updates, assets, quality and ci-ok. Marked ready for review; website integration continues in stacked PR feat(web): migrate website typing runtime to Rust and Wasm #185.

No production deployment or release publication. Full website host behavior/delivery/performance cutover is Stage 2; native platform packaging/manual terminal matrix is Stage 3. Wasm retains its high-water linear allocation: the 2,000-unit reset budget passes; the separately reported 20,000-unit/full-log soak does not claim memory shrinks to baseline.

API: packages/typing-engine/README.md. Acceptance evidence, immutable provenance, compatibility decisions and performance conditions: docs/rust-engine/.