speedlify2

Recently measured 60 sites

Newest first, across the whole corpus. The dots are the four Lighthouse categories, then axe and Core Web Vitals, in the colors that run carried.

1. frankfuchs.net 6h
2. ivoputzer.blog 6h
3. gaiety.life 6h
4. blog.fontawesome.com 6h
5. vite.dev 6h
6. sony.com 10h
7. react.dev 10h
8. preactjs.com 10h
9. jspm.org 10h
10. jquery.com 10h
11. emberjs.com 10h
12. jekyllrb.com 10h
13. sap.com 10h
14. firebase.google.com/products/hosting 10h
15. nvidia.com 10h
16. tesla.com 10h
17. jsr.io 10h
18. eslint.org 10h
19. samsung.com 10h
20. hex.pm 10h
21. docusaurus.io 10h
22. intel.com 10h
23. docs.gitlab.com/user/project/pages 10h
24. broadcom.com 10h
25. nejsconf.com 10h
26. google.com 10h
27. nordcraft.com 15h
28. vuepress.vuejs.org 15h
29. vitepress.dev 15h
30. svelte.dev 15h
31. render.com 15h
32. railway.com 15h
33. fossgaming.codeberg.page 15h
34. germanfrelo.dev 15h
35. fuzzylogic.me 15h
36. fullstax.de 15h
37. frontenso-11ty-starter.netlify.app 15h
38. frontendnorth.com 15h
39. freetransitedmonton.ca 15h
40. francisrub.io 15h
41. fotis.photos 15h
42. tsuwave.studio 15h
43. surge.sh 23h
44. gustavlindqvist.se 23h
45. gsap.com 23h
46. salesforce.com 23h
47. surajwate.com 23h
48. flag.is 23h
49. pypi.org 23h
50. kiri-vadivelu.ca 23h
51. fifty.snook.ca 23h
52. tailwind.buildawesome.one 23h
53. zapparov.dev 23h
54. fberriman.com/blog 23h
55. starter.buildawesome.one 23h
56. tampamonitor.com 23h
57. pico-blades.buildawesome.one 23h
58. exyon.ee 23h
59. paulmakeswebsites.com 23h
60. buildawesome.one 23h

The measuring machines Lighthouse benchmark index

Recorded once per measurement, from the run whose numbers were kept. Higher is faster.

Median

2,486

±35% Median

1,616–3,356

Mean

2,653

Range

838–4,588

From 12,747 measurements. The mean sits above the median because a few unusually fast machines pull it there; most runs land close together. A site is told “the measuring machine was faster than most in this project” when its run lands outside that band. That is a distance rather than a percentile because the notice claims the CPU-bound numbers moved, and how far off the machine was decides that; how rare it was does not. These machines vary by running fast rather than slow, so in practice the notice is a fast-machine one.

Reading the logs

Each run writes a full NDJSON log to logs/<runId>.ndjson, one JSON object per line, plus a summary line appended to logs/runs.ndjson. Both are greppable:

# Every warning and error across all runs
grep -h '"level":"warn"\|"level":"error"' logs/*.ndjson | jq .

# What happened to one site in one run
jq 'select(.data.url == "https://example.com/")' logs/2026-01-01*.ndjson

# Run history at a glance
npx speedlify runs