Dyno
Address
aadhar.shGarageDyno

Dyno

A dyno tells you what an engine actually puts down. The spec sheet is a claim; the rollers are a measurement. This one straps the site down every night: a GitHub Action builds main, weighs what would go over the wire, and records a pull.

0140280420561 2026-08-09 · 470.6 KiB · 36d3e552026-08-10 · 470.7 KiB · 5df3d2a2026-08-11 · 476.8 KiB · d98bcc42026-08-12 · 484.9 KiB · 6e22d6c2026-08-13 · 494.2 KiB · 9eb993f2026-08-14 · 494.2 KiB · 17a44fe2026-08-15 · 497.0 KiB · 8ce2bbe2026-08-16 · 500.5 KiB · 610c2012026-08-17 · 500.4 KiB · da7f0472026-06-01 · 86.0 KiB · pre-consolidation — homepage Worker alone, before cal/ and serendipity/ were consolidated into it2026-06-30 · 129.2 KiB · consolidation — the one-Worker consolidation. the old three-Worker aggregate measured 130.94 KiB, so this was smaller than what it replaced2026-08-04 · 204.2 KiB · rebaseline — re-baselined after the 129.23 line had been in continuous breach for months. lens.js and serendipity.js are 35% of the bundle between them2026-08-08 · 258.3 KiB · 295ee97 — measured while building the wire-size diff, four days after the re-baseline and already over its 25% alert2026-08-09 · 266.6 KiB · 36d3e552026-08-10 · 266.7 KiB · 5df3d2a2026-08-11 · 269.2 KiB · d98bcc42026-08-12 · 276.2 KiB · 6e22d6c2026-08-13 · 276.6 KiB · 9eb993f2026-08-14 · 278.6 KiB · 17a44fe2026-08-15 · 272.0 KiB · 8ce2bbe2026-08-16 · 274.8 KiB · 610c2012026-08-17 · 228.5 KiB · da7f0472026-08-09 · 58.0 KiB · 36d3e552026-08-10 · 58.0 KiB · 5df3d2a2026-08-11 · 61.0 KiB · d98bcc42026-08-12 · 67.5 KiB · 6e22d6c2026-08-13 · 68.1 KiB · 9eb993f2026-08-14 · 68.1 KiB · 17a44fe2026-08-15 · 74.0 KiB · 8ce2bbe2026-08-16 · 74.4 KiB · 610c2012026-08-17 · 75.0 KiB · da7f047 2026-06-012026-08-17 KiB
The worker bundle is the pull worth watching. It went from 86 to 228 KiB gzip between 2026-06-01 and 2026-08-17, up 166%. Every one of those numbers got found by somebody tripping over a stale threshold rather than by anyone watching the slope, which is the specific failure a per-change check cannot catch and a series can.

Recent pulls

datecommitworker gzippages brassets brsource
2026-08-17 da7f047 228.5 500.4 75.0 measured
2026-08-16 610c201 274.8 500.5 74.4 measured
2026-08-15 8ce2bbe 272.0 497.0 74.0 measured
2026-08-14 17a44fe 278.6 494.2 68.1 measured
2026-08-13 9eb993f 276.6 494.2 68.1 measured
2026-08-12 6e22d6c 276.2 484.9 67.5 measured
2026-08-11 d98bcc4 269.2 476.8 61.0 measured
2026-08-10 5df3d2a 266.7 470.7 58.0 measured
2026-08-09 36d3e55 266.6 470.6 58.0 measured
2026-08-08 295ee97 258.3 — — by hand
2026-08-04 rebaseline 204.2 — — by hand
2026-06-30 consolidation 129.2 — — by hand
2026-06-01 pre-consolidation 86.0 — — by hand

Every number here is deterministic: identical source bytes weigh the same every time, so a flat line means nothing changed rather than that nothing was measured. Sampled figures are deliberately absent, because wrangler check startup read 9.6, 7.6, 6.4 and 16.4 ms across bytes nobody touched, and charting that would draw weather. A dyno with a wandering needle is a decoration. Raw series at /garage/dyno.json.