What Broke Today

🟢 GRAFANA Jun 17, 2026

🔴 Grafana's prod-me-central-1 decided to become a historical artifact. AWS admitted the UAE region is genuinely toast—recovery measured in months, not hours. Time to play datacenter roulette and pray your backup plan isn't also in the same burning building. (Auto-resolved: no updates in 24h)

🟢 FLY Jun 17, 2026

🟡 Fly's SIN region is playing hide-and-seek with network packets—some are winning. Apps in Singapore are currently auditioning for the role of 'unreachable,' and engineers are pretending they know what happened. Classic case of 'investigating' being corporate for 'we're checking Slack right now.'. (Auto-resolved: no updates in 24h)

🟢 FIGMA Jun 16, 2026

🟡 Figma's Make and Agent features had an existential crisis when talking to Claude—turns out the AI forgot its training. Engineers deployed the digital equivalent of smelling salts and declared victory. They're watching for relapse. (Auto-resolved: no updates in 24h)

🟢 CONVEX Jun 15, 2026

🟠 Convex decided to test how many errors could fit in a deployment simultaneously. Most machines failed the audit; D1024s apparently have better lawyers. Fixed faster than you can say 'why is monitoring configured like this?'. (Auto-resolved: no updates in 24h)

🟢 AWS Jun 15, 2026

⚪ Fable 5 and Mythos 5 went full ghost mode across all regions. Anthropic claims it's 'resolving' this, which is corporate speak for 'we'll tell you something eventually.' Turns out releasing new AI models requires actually letting people use them. (Auto-resolved: no updates in 24h)

🟢 FLY Jun 15, 2026

🟠 Macaroon auth system decided to stop working, taking the Machines API down with it. New deployments are DOA, but your already-running stuff is fine—for now. Engineers are sweating through the fix. (Auto-resolved: no updates in 24h)

🟢 FIVETRAN Jun 14, 2026

🟠 Multiple Customer.io connections are throwing 5xx errors like confetti. Fivetran implemented a skip mechanism, but the issue just migrated from job creation to report downloads—classic adversary-with-a-moving-target situation. Engineers are monitoring like hawks, hoping the problem doesn't develop new attack patterns. (Auto-resolved: no updates in 24h)

🟢 LAUNCHDARKLY Jun 13, 2026

⚪ LaunchDarkly admitted some of your event data ghosted you during two separate windows—turns out 'event ingestion' doesn't mean the data actually stays. They're backfilling now, but your feature flags may be running on incomplete intel. At least they're honest about losing things instead of pretending it never happened. (Auto-resolved: no updates in 24h)

🟢 NETLIFY Jun 13, 2026

🔴 Anthropic yanked Claude Fable 5's access like a bouncer removing a rowdy patron. Netlify's AI Gateway now serving 404s instead of AI responses. Turns out 'suspended access' is corporate speak for 'we're taking our ball and going home.'. (Auto-resolved: no updates in 24h)

🟢 CLERK Jun 12, 2026

🟠 Clerk's upstream provider decided to take an unscheduled nap, leaving the 'go live' button about as functional as a screen door on a submarine. Errors and delays incoming for anyone trying to launch their app—monitoring is happening, which is tech-speak for 'we're watching the fire spread.'. (Auto-resolved: no updates in 24h)

🟢 CONFLUENCE Jun 12, 2026

🟠 Confluence decided to practice its namesake by flowing absolutely nowhere for 4 hours straight. Users got a masterclass in document inaccessibility. The hamster powering it all has since been replaced with a slightly angrier one. (Auto-resolved: no updates in 24h)

🟢 CONTENTFUL Jun 12, 2026

🟡 Contentful's full-text search went on an unplanned European vacation, leaving customers unable to locate their own content. They've fixed most of it, but some EU folks are still playing hide-and-seek with their data. (Auto-resolved: no updates in 24h)

🟢 WEBFLOW Jun 11, 2026

⚪ Update: [RESOLVED] Reports of intermittent 500 errors for some hosted sites

🟢 WEBFLOW Jun 11, 2026

🟠 Webflow handed some sites a participation trophy (500 errors) earlier today, then pretended it never happened. Crisis averted via 'mitigation measures'—tech speak for 'we turned it off and back on.'

🟢 HASHICORP Jun 11, 2026

🟡 HashiCorp's infrastructure-as-code platform is having trust problems—users can't download their precious state files. The company claims to be 'investigating,' which is corporate for 'our engineers are stress-eating at their desks.' Your infrastructure's source of truth just became fiction. (Auto-resolved: no updates in 24h)

🟢 CONTENTFUL Jun 11, 2026

🟠 Contentful's content editors got ghosted by the API for a few hours—publish buttons turned into participation trophies. Most customers unfrozen, but some EU folks still negotiating with their keyboards. (Auto-resolved: no updates in 24h)

🟢 FLY Jun 11, 2026

🟠 Fly discovered that their North American network routes have commitment issues—some packets are ghosting while others contemplate existence mid-transit. Engineers are playing whack-a-mole with 'impacted backbones' like they're herding cats made of fiber optics. (Auto-resolved: no updates in 24h)

🟢 FLY Jun 10, 2026

🔴 GRU edge nodes decided to BSOD themselves, taking critical services down with them. Traffic routing through São Paulo is now a game of roulette. Engineers are frantically rebooting things and pretending they knew this would happen. (Auto-resolved: no updates in 24h)

🟢 CURSOR Jun 10, 2026

🟡 Cursor's having an identity crisis with widespread outage reports. Engineers are speed-running the 'turn it off and on again' flowchart. Apparently 'investigating' is tech speak for 'we're Googling this too'. (Auto-resolved: no updates in 24h)

🟢 CURSOR Jun 10, 2026

🟡 Cursor's blaming GitHub for the meltdown affecting Cloud Agents and dashboard integrations. Third-party outage domino effect in full swing—nothing says 'our infrastructure is solid' like depending on someone else's. (Auto-resolved: no updates in 24h)

> cat /var/log/failures.log