W1 status

How this is measured

Every five minutes, on a schedule. The checks run from Cloudflare's network whether or not anyone has this page open — so an outage at three in the morning is recorded the same as one at midday. A status page that only measures itself when someone is watching is publishing office-hours uptime and calling it uptime.

Each check asserts what it should get back, not merely that something answered. Asking the turn endpoint for work without a token should be refused; a refusal is the passing result, and anything else is a failure. The update feed has to return a real version, not just a 200. Sign-in has to actually hand off to the identity provider.

Components are things that break separately. Sign-in and agent turns run on the same service and fail for entirely different reasons — one can be perfect while the other is unusable, which is exactly the distinction a single "API: operational" line destroys.

Degraded is a real state, not a softer green. A component is degraded when some of its checks fail while others pass, or when everything passes but too slowly to be usable. Degraded checks count against the uptime figure — a percentage that quietly forgives slowness is not a measurement.

What it does not tell you

These are reachability checks made from outside, so a green bar means W1's services answered correctly — not that your run went well. A model that replies slowly or badly to a real, signed-in request is not something an anonymous probe can see. Days before this page existed were never measured and are drawn as not measured rather than filled in green.

So if W1 is failing for you while everything here is green, that is genuinely worth telling us: it means the failure is somewhere this page cannot yet see, and that is the most useful bug report we get. Support has what to send.

Nothing here exposes provider names, internal errors, traces, or infrastructure detail, and it never will.