Separating AI provider status from local product issues before restarting

When a model completion hangs or a reply stalls, the common reaction is to restart machines and rewrite prompts, which conflates provider, product, and local network failures. The article recommends first noting exactly when the symptom began, then reproducing it in a private window or on a second device to see whether only one path is affected. It advises checking two independent status pages and identifying the specific feature involved, since a healthy homepage can coexist with a failing generate function. Latency rises with successful connections are described as a likely shared degradation rather than a reason to revise prompts repeatedly. The author concludes that a clear triage note should guide whether to wait, switch models, or continue working offline.