FAQ

No idea. This is an honest attempt, measured in your browser on this load. If you know a faster page, or spot a cheat, tell me. Feedback is welcome.

Yes. Slow sites cost attention, conversions, and patience. The point is that the web can be much faster than most of what gets shipped, and it should be, for both business and basic courtesy.

There is no central origin server in the path that must wake up before you get the page. HTML is pre-built and served from the edge network. Cache misses and dynamic proof still run at a nearby PoP, not on a single home machine. Longer write-up on the About page.

No. Requests land on one of many Cloudflare points of presence, often the one closest to you on the network. The PoP cell on the live board is which one answered this visit. Someone else can see a different code, RTT, and cache state. That spread is the point of edge delivery, not a bug.

Yes, lots of them. Originless means no single distant origin in the critical path, not “no computers.” Builds and deploys still go through Cloudflare’s platform. You just do not wait on one rack in one region for every first byte.

Core Web Vitals and FCP are measured in your browser for this visit. Edge and payload cells use the nearby server that answered you, plus Navigation and Resource Timing for this load. Lab scores are static marks I update after deploys, not live third-party widgets. Field averages (when configured) appear only in the note under This visit; they do not replace the cells.

Hints that say “vs avg” or “typical web” are ballpark baselines, not a live crawl of the whole internet and not a promise that every site hits those numbers. Paint targets (LCP ~2.5s, FCP ~1.8s, CLS 0.1, INP 200ms) follow common Core Web Vitals / “good” thresholds. Payload and edge ballparks are rounded last-mile figures I keep in the proof scripts: HTML ~100 KB, full page ~2.2 MB, ~70 requests, TTFB ~800 ms, RTT ~80 ms. Tile colors use my own good / ok cutoffs for this demo (for example TTFB under 200 ms looks strong here). If a baseline is wrong for your lab, treat the raw value as the source of truth and send a better citation.

A lab comparison of well-known homepages under one PageSpeed Insights mobile profile — LCP, transfer size, and request count. See Rank. It is not this visit’s live numbers and not Chrome’s field report (CrUX). The fairness contract is on Rules.

Interaction to Next Paint only counts a real click, tap, or key — not mouse move, scroll, or a scripted fake event. The tile says “Click to calculate” until a value exists; the tip opens only after that. If the browser doesn’t expose Event Timing, the tile stays blank. On Chromium, if the page is this fast you may see ≤16ms — below what the API reports.

Weight is the product. Extra chrome, webfonts, and badge networks would argue with the claim. The metrics are meant to be read, not buried under decoration.

Sometimes. A warm cache, a close PoP, or a quiet network helps. Reload, try another network, or run your own lab test if you want a second opinion. A healthy tile means healthy for this visit, not a lifetime trophy. Comparison percents inherit the same baseline caveats as above.

Reach out through mikkelscheike.com.