March 11, 2026
When Claude clocks out, chaos clocks in
Elevated errors on login with Claude Code
Claude Code login chaos: devs roast, bail, and swap tools mid-bugfix
TLDR: Claude.ai and Claude Code logins broke for about three hours after a database slowdown during maintenance, then were fixed; the API kept working. Devs cracked jokes, some threatened to switch tools, and others blasted Anthropic’s confusing sign-ins—turning a routine outage into a trust-and-reliability debate.
Claude’s coding sidekick took an unexpected nap today: a routine database tweak slowed the main site so badly that logins to Claude.ai and Claude Code sputtered for nearly three hours, while the developer API kept cruising. The company says it’s resolved and monitoring, but the comments? Oh, they’re sizzling.
Over on Hacker News, the split is stark. The impatient crowd is already jumping ship, with one user quipping that the delays were long enough to try a rival tool. The cynics are sharpening knives, mocking the dream of “AI that replaces engineers” when an AI assistant can’t even clock in. And the power users are airing old wounds: confusing domains and sign-in setups (SSO = single sign-on) that left some folks tangled between personal and work accounts. It’s part comedy, part cautionary tale.
There’s humor too: people joked about waiting for a “Distinguished Engineer named Claude” to return from a long break, while others flexed Plan B workflows—importing chats into other tools and moving on. The vibe: reliability beats hype, and a login wall can derail an entire day. Today’s moral? Downtime isn’t just a bug—it’s a meme.
Key Points
- •Primary application database suffered degraded I/O after routine maintenance, causing service issues.
- •Incident window: 14:17–17:11 UTC (07:17–10:11 PT) on Mar 11, 2026.
- •Claude.ai experienced slow/failed requests; Claude Code and the Console had sign-in disruptions.
- •Claude Developer Platform/Claude API remained unaffected throughout the incident.
- •A fix was implemented, monitoring began, and the incident was marked resolved at 17:28 UTC.