GitHub Outage Map

Community Discussion

Tips? Frustrations? Share them here. Useful comments include a description of the problem, city and postal code.

Beware of "support numbers" or "recovery" accounts that might be posted below. Make sure to report and downvote those comments. Avoid posting your personal information.

GitHub Issues Reports

Latest outage, problems and issue reports in social media:

• 
  Nuvorlane (@nuvorlane) reported 2 minutes ago

  Cloud agents now subscribe to the PRs they open. Cursor's Aug 19 changelog: an agent can watch a PR, a Slack thread, or a timer, then wake when something happens. Subscriptions are cloud-only for now. Agents auto-subscribe to PRs they create and keep going: fix CI, answer bot comments, drive the PR to done. The help page is narrower than the pitch. Auto-fix is GitHub Actions, on PRs the cloud agent opened, Teams plans. Checks already red on the base branch are skipped. Kill it on one PR with @cursor autofix off, or globally in Dashboard → Cloud Agents → My Settings. If you don't want an agent still writing after you left Slack, don't let it open the PR, or unsubscribe. Send this to whoever owns the cloud-agent policy.

• 
  Swarn (@SwarnCode) reported 3 minutes ago

  GitHub was down for almost 8 hours. and one of the things keeping it down was literally the thing meant to help it recover 😭 the failure chain was wild: one Istio sidecar hits its concurrency limit autoscaling doesn’t react because it’s watching the wrong thing requests start timing out retries kick in then a VS Code retry bug amplifies Copilot traffic by 10x now the recovery mechanism is creating even more load. this is what i find interesting about distributed systems. retries look harmless when you're learning backend. at scale, badly designed retries can basically DDoS your own service. sometimes the “fix” is what keeps the incident alive.

• 
  Kevin Morales (@KevinhoMorales) reported 5 minutes ago

  @cursor_ai "Hold a goal until it's met" is the part I care about. On our iOS repo the failure was never bad code, it was the agent quietly redefining the goal halfway through a long session. Kicking off cloud agents from a GitHub issue and reading the diff later is how I actually ship now.

• 
  🧞‍♂️Martin Donadieu - oss/acc (@martindonadieu) reported 6 minutes ago

  Fixed From today we push as usual our package via our multi file Github actions with a token bypass_2FA=false as staged. And a Github action that unstage them with 2FA with no human involved. No need broken OIDC, no need change all our process. just hacked my own account to make things work as they should

• 
  Diego Oppenheimer (@doppenhe) reported 8 minutes ago

  is anyone shocked that github was taken down by k8s? the gh incident report reads like 1/3 of my life running algorithmia...

• 
  Defileo🔮 (@defileo) reported 13 minutes ago

  🚨 55,000 STARS AND IT'S STILL COMPLETELY FREE. NO F*CKIN WAY THIS ISN'T A DREAM. Strix just hit 55,731 stars on GitHub (+10k stars in 1 week) > multi-agent team that hacks like a real attacker, not a scanner > every finding is proven with a real PoC, no false positives > solved 96% of XBEN web challenges for about $3 each > local live dashboard with agent graph, findings, and steering Authorized targets only, this thing actually breaks things to prove they're broken. Fewer words, more action, repo in the comments.

• 
  capONE 💎 (@caponeWeb3) reported 13 minutes ago

  THE ENGINEER WHO BUILT CLAUDE CODE, ON WHY IT CAN DIG THROUGH *** HISTORY WITHOUT BEING TOLD HOW: "It knows it because the model is awesome." At Anthropic, technical onboarding used to run two to three weeks. It's now two to three days - and Boris Cherny credits a single habit almost nobody starts with: don't touch the code first, interrogate the codebase first. *** blame. GitHub issues. Why some function ended up with fifteen arguments and who's responsible. What actually makes that stick isn't the questions - it's what gets remembered afterward, and for how long. Same layer the article quoted below goes deep on.

• 
  Lloyd 👨‍💻 (@lloydtheophilus) reported 16 minutes ago

  GitHub going down recently wasn’t just a developer inconvenience. It was a business continuity event. Think about what sits behind a modern software organization: - Source code. - CI/CD pipelines. - Infrastructure automation. - Package registries. - Deployments. - Code reviews. - Release management. - Documentation. - Incident response. When a critical developer platform becomes unavailable, engineering productivity can stop and when engineering stops, business value stops moving. This is the infrastructure lesson we need to take seriously: We have spent years eliminating servers we manage ourselves and moving critical workflows into SaaS platforms. That’s great for productivity. But every abstraction creates a dependency. The future of version control isn’t simply going to be about who builds the better *** interface. It will be about resilience, portability, redundancy and business continuity. Can a company continue deploying if its primary *** platform is unavailable? Can engineers recover repositories quickly? Can CI/CD operate independently? Can infrastructure teams provision environments without access to the platform? Can the organization switch providers without months of migration? These questions will increasingly become architecture and executive-level questions not just DevOps questions. *** isn’t going anywhere. But the companies that understand *** as critical infrastructure, rather than just a developer tool, will be better prepared for the next major disruption. The future of software isn’t just about building faster. It’s about building systems that can keep the business running when the platforms we depend on fail.

• 
  Lucas G (@lucas_gr_) reported 18 minutes ago

  It was a blast working on a ton of things without any rules for a few hours, just following my curiosity... now I promise I'll get back on track! I tried to push all the ideas I had in my head and wanted to execute for a while to version 1.0, tracking everything with my Second Brain so I can calmly decide on priorities and pick up where I left off Things I worked on: -Built a small digital product: researched a problem from scratch using Google data, mapped pain points via Reddit, built and launched the product, with a test Meta Ad already running -Created the Second Brain version for SMMs (again, mapping their pain points) -Set up a Telegram Bot to send my voice notes to Claude even when I'm out and about -Got Claude running on the VPS (and sent a "good morning" prompt at 5:30 AM so my session would reset at 10:30 AM) -Upgraded my personal website to v1.1 -Created a presale page for the Corporate Second Brain (for the Italian market) -Put together a portfolio PDF for my outreach system -Adapted a trading bot, pulling the best ideas from GitHub and best practices from Reddit (I'll leave it in testing until the end of the year and progressively put real money into it starting in 2027, little by little) -Worked on a vibecoded website idea I've been testing for two weeks -Built v1.0 of a last-minute idea that popped up last night -Finished the remote job search system for Italians (adapted from a GitHub repo I found) I swear it was only because I didn't want to waste the tokens...back to being focused now 😮‍💨

• 
  Mark (@Milkdromida) reported 20 minutes ago

  @Carreyoki Thank you for helping, but I've had no luck so far. Other users on GitHub are reporting the same, so it is a temporary app issue it seems.

• 
  Ed Andersen (@edandersen) reported 23 minutes ago

  @robmen @bertyJobbo The danger is potentially giving money to an individual without checking they are on sanctions lists Hangfire has historically had this problem due to the location of the author and it’s a “proper” pro licensed set of binaries with a paywall The fact the binaries are not going to be paywalled for Polly is just inviting license violations, basically by design. With a 20k barrier basically any employed developer is covered, all this does is create a massive list for the Polly author to send legal letters and threaten audits after - the same business model as VS Community with its 1m USD barrier Is Polly going to pay Microsoft a portion of their revenue to host their commercial binaries on nuget and GitHub? The going rate for this kind of thing is 30%. Yes OSS authors should be paid but not like this. Paywall and use license keys

• 
  RICO (@RICO_Lumiere) reported 25 minutes ago

  Sorry to bother you out of the blue. I’ve honestly run out of options, and I heard you might be able to help with GitHub issues I can’t access, signing in just redirects back to login page, and my ticket(#4678706) hasn’t received a response Any help would mean a lot🙏 @moraes_c_

• 
  Doipayan (@doipayann) reported 25 minutes ago

  Don't build a backend. This is the fastest way to eliminate latency in a tool: Stash runs on React 18, Vite, and TypeScript. All bookmarks live strictly inside browser localStorage (key: STASH_AF_BOOKMARKS_V2). The trade-offs: 1. Zero network latency — search and filter run client-side in under 5ms. 2. Complete privacy — your links, notes, and ratings never touch an external server. 3. Serverless distribution — hosted for free on GitHub Pages. For data mobility, users can export and import their entire vault as a clean .json file. Not every product needs a database and authentication to be useful.

• 
  Rat King Crimson (@Alphiloscorp) reported 25 minutes ago

  @burkeholland Or mandate a poor usage pattern. I am presently staring down a likely impossible deadline for the project team that would be vastly assisted by use of the design product's MCP. Problem? We have Copilot for Github, but API keys and 3rd party MCPs are both blocked.

• 
  Crypto Banter (@crypto_banter) reported 25 minutes ago

  🚨BITCOIN’S FIRST INSCRIPTION SERVICE IS SHUTTING DOWN! OrdinalsBot, the original Bitcoin inscription platform launched just weeks after Ordinals went live, is winding down after three years. The founders cite a sharp contraction in the Ordinals market and say sustaining the business is no longer viable. They’re putting the full asset stack, brand, IP, domains, socials, Discord, GitHub, and 90+ code repositories, up for sale through an open bidding process.