GitHub has overhauled its status page with three concrete changes: a new three-tier severity system adding 'Degraded Performance' below Partial and Major Outage, per-service uptime percentages calculated over a rolling 90-day window, and a dedicated 'Copilot AI Model Providers' component split out from the broader Copilot service. The old system forced every incident into at least a Partial Outage classification, overstating impact when services remained functional.
The uptime math is worth understanding. Major Outages carry a 100% downtime weight, Partial Outages carry 30%, and Degraded Performance carries 0%. A 1-hour Partial Outage counts as only 18 minutes of effective downtime in the calculation. That weighting methodology, described as industry-standard, shapes every uptime number GitHub now publishes publicly.
The Copilot component split reflects a real architectural fact: Copilot Chat and the Copilot cloud agent support multiple models, so a single model provider going down does not equal a Copilot outage. Future model-availability incidents will be filed under 'Copilot AI Model Providers' with specifics on which models are affected. Read the full post for the complete severity weighting table and the auto model selection detail that makes the component separation operationally meaningful.
[READ ORIGINAL →]