About

down.new answers one question — is it down? — and shows the evidence behind the answer. It watches hundreds of services and publishes a verdict every minute.

What we measure

Three independent signals, fused: reports from people, measured against what is normal for that service at that hour of the week; HTTP, DNS and TLS probes run every 60 seconds from up to 12 regions; and the vendor's own status page where one is published. No single signal can declare an outage on its own.

Why we hedge

Our wording is deliberately careful, and the difference between phrasings is the difference between a fact and a guess. "Users report X is down" means the crowd is reporting it and nothing has corroborated it. "We can't reach X from our probes" means only our probes see trouble — our fleet runs inside Cloudflare, and plenty of origins block datacenter addresses, which looks identical to an outage from where we stand.

We learned this the hard way. On our first day live we told the world that fifteen services were down when they were not, because every probe region failed at once against origins that block datacenter IPs. The rules that came out of that are in how we know, and probe-only evidence can no longer produce a confident "down".

What we do not do

No ads before the answer. No user-level data is sold or shared, ever. We store no raw IP addresses, and regional breakdowns are published only above a k-anonymity threshold — see privacy.

Using our data

The API and MCP server are free to read and rate-limited. Every status page has a markdown twin at /{service}.md for agents. If you quote a figure, link to the page it came from and keep the wording of the verdict — the hedging in it is load-bearing.

Related

speed.new is our sibling: it measures your connection and publishes which provider is actually fastest where you live.