Learn4 min read
Hosting in India? Your server waits longer for the APIs it calls
We time 20 AI APIs, developer APIs and package registries every day from Mumbai and San Francisco.
How much longer ChatGPT, Claude, Stripe, Docker Hub, crates.io and 15 more services take to answer a server in Mumbai than one in San Francisco, and why.
Learn6 min read
Does proximity mean low latency?
Why did our Singapore server take 465 ms to reach GitHub, when GitHub's server is in the same city?
How we traced a 465 ms GitHub API result in Singapore to a route through France, one step at a time, with the tools we used at each step.
Learn4 min read
What is Time to First Token? (TTFT)
The wait before an AI starts answering, and the point where people notice it.
What TTFT measures, what counts as fast, why it can be slow, and how ChatGPT, Claude and Grok compare in our daily measurements.
Learn6 min read
Is 340 ms slow?
It depends on what you're measuring.
A simple guide to deciding whether 340 ms is fast or slow for a website, API, static file, or AI response.
Learn5 min read
Test API latency from multiple regions
Measure worldwide latency without running your own infrastructure.
A practical guide to checking API latency from several regions, understanding what the numbers mean, and turning that into an infrastructure decision.
Learn4 min read
What is TTFB? The wait before your page can start rendering
Time to First Byte is how long the browser waits for the first byte of your HTML. It isn't a ranking signal on its own, but all of that wait counts toward LCP, one of Google's Core Web Vitals.
Time to First Byte is the wait before rendering can begin. It counts toward LCP, so a slow one can hurt your Core Web Vitals.
Learn4 min read
Google Fonts or self-hosted? It depends on one thing.
The answer used to be simple. It isn't anymore.
The Google Fonts answer changed after shared browser caches disappeared. The real decision now is where your server lives.