type: index

Distribution measurements

Receipts and consumer locks are preserved as recorded; temporary archive URLs and receipt output paths are historical identifiers rather than files to open.

• 2026-10-11-oxc* and 2026-10-11-terser*: source 5aac8654, identical installed package/dependency identities, measured minifier experiment. The canonical minifier remains Oxc because Terser increased gzip and build time.
• 2026-10-11-runtime-tarball.json: packed ownership/size audit at that source.
• 2026-10-11-entrypoint-comparison.json: preserved producer artifacts at 5aac8654 and 8b5b0a05, identical canonical esbuild options. The focused PROJJSON graph improves; the generic hero graph is unchanged.
• 2026-10-11-ci-installed-quickstart-budget.json and matching .consumer-lock.json: exact-head 8b5b0a05 CI calibration, run 38130548203, artifact 11698283813.
• 2026-10-11-ci-merge-installed-quickstart-budget.json and matching .consumer-lock.json: separate GitHub test-merge b304bef7 calibration, run 38130548207, artifact 11698779861.

The execution configuration uses the exact-head calibration, with the separate merge run showing variation across provisioned runner CPUs. These receipts are packed development evidence, not published registry or live candidate certification.

• 2026-10-11-final-ci-installed-quickstart-budget.json and matching .consumer-lock.json: source 7c6a67b57, GitHub run 38137168675, artifact 11702062323. Live bundle: 1,870,600 raw / 497,269 gzip bytes. Closed-fixture CPU medians: 82.624 ms script / 45.524 ms parse-compile, passing the declared Ubuntu 24 x64 / Chromium 145 / cold 4x profile. The runner CPU model is recorded rather than required. Fixture execution and live bundle measurements stay separate.
• 2026-10-11-final-runtime-tarball.json: the same clean source and admitted producer artifact 11701912054, after canonical browser/inventory preparation; 4,071,732 compressed / 19,627,609 unpacked bytes, passing both reduced ceilings.

Final receipts retain their original sourceSha and sourceDirty: false; later report commits do not become their measurement source. The local installed comparison independently matches the hosted live byte totals with all 121 non-SDK dependency version/integrity identities fixed to the prior consumer lock. Moving the consumer outside the repository removes undeclared optional-peer resolution; that correction is not claimed as an SDK graph optimization. The focused PROJJSON validator reduction remains the measured source improvement.