For context, this isn't 100% true.

This is from a third-party tracker, who took the downtime of every GitHub hosted service and added them together. So a bunch of stuff that was down 0.5% of time and/or some beta+new release services down for longer make it look way worse than the core service uptime actually is.