Both frames render identically — that's the goal. The win is in the bytes and requests below, not the look. They're served live by mod_pagespeed 1.1 on demo-httpd-1.1.modpagespeed.com; the optimized frame applies only this filter. Right after a cache purge it may briefly match the original while the worker rewrites it — reload to see the result. This filter is beacon-driven: mod_pagespeed instruments the page, a real browser reports back, then it serves the optimized form and re-measures every few minutes — so the optimized frame can take a reload, or briefly revert to the original, before it reflects the change.

mod_pagespeed applies this only after a real browser loads the page and reports back (and re-measures periodically), so there's no server-side diff to capture here — the frames above show it live.