Elevated error rates on Opus 4.7

Anthropic’s status page quietly announced elevated error rates for Claude’s Opus 4.6 and 4.7, saying both the Claude API and Claude Code were affected and that the team was still investigating. But the real action wasn’t on the official update page — it was in the community, where users immediately started doing that classic internet thing: comparing notes, sounding alarms, and serving sarcasm with receipts.

The biggest mood was basically, “Uh, it’s not just Opus.” Multiple commenters rushed in to say Sonnet was wobbling too, with reports of overloaded messages and requests failing. One user warned that automatic retry systems — the feature that keeps trying again after a failure — could actually make the pileup worse, which turned the thread from mild annoyance into full this-is-how-cascades-start energy. Another commenter dryly observed that the company was “having quite the day for devrel,” which is tech-speak for: imagine being the person whose job is to calmly explain all this in public.

And then came the spiciest take: a jab at executives who pushed teams to depend heavily on cloud-based tools without keeping local backup options. That turned a simple outage thread into a mini culture-war about modern software work: convenience versus resilience, hype versus planning. Even the shortest reply — just a link to the status page — had the vibe of someone wordlessly sliding the official paperwork across the table while the room burns. In other words: part bug report, part roast session, fully online.