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:

• 
  :) (@Eimen_Nur) reported 2 minutes ago

  @ThePrimeagen Github has had an issue the past 6 months…. We need something better

• 
  spb krishnan (@spbalaonline) reported 5 minutes ago

  @abhijeet_dipke Cockroach Party website taken down? Here's how to make it resilient:• Move domain to Njalla/Porkbun + Cloudflare proxy • Host static version on IPFS + Vercel/Netlify • Mirror on GitHub Pages+ offshore VPS • Daily backup+.onion version Decentralize like real cockroaches.

• 
  Oliver Rolle (@OliverRolle) reported 7 minutes ago

  @CryptoSp33d Since Microsoft started using AI Windows has become worse. Massive GitHub outage with 6h of *** merge data lost. Microsoft research proved when context window exhausts AI starts making huge mistakes yet are difficult to detect because it looks right on a superficial level

• 
  38twelveDaily (@38twelveDaily) reported 11 minutes ago

  Problem: Microsoft engineers have favored Claude Code over Copilot CLI. There are gaps between the products that Microsoft now has to close. The GitHub team is shipping improvements based on feedback.

• 
  Joshua McKenty (@jmckenty) reported 13 minutes ago

  Hey @bcherny — would it be possible to enable the actions toolset on the github MCP server in Claude Code on the web? get_job_logs(failed_only=True) would unblock CI debugging; my agent had to bisect a bunch of commits today instead of just reading the failure.

• 
  Timur Yessenov (@Timur_Yessenov) reported 14 minutes ago

  @BHolmesDev @mattpocockuk GitHub Issues is a good handoff surface if the issue has a cleanup rule. I’d add two fields to every /handoff: why it was deferred, and what would make it safe to close without doing it.

• 
  luxus (@luxus) reported 17 minutes ago

  @Daniel_Farinax Let grok write github issues of your brain dump

• 
  CyrilXBT (@cyrilXBT) reported 18 minutes ago

  A guy named nbatman on Reddit accidentally built the most censorship-resistant website on the internet. Hollywood tried to kill it. Spotify tried to kill it. Adobe tried to kill it. Google delisted it. Reddit shadow-banned it. The Motion Picture Association flagged it as a top threat. The RIAA pressured every hosting provider it ever used. It is still online. Updated every single month. By six anonymous volunteers working in their spare time. Here is why nobody can shut it down and what it teaches every builder about the future of the internet. THE ORIGIN 2018. One Reddit moderator. One Google Doc. A single person decided to organize the internet's free resources into one place. No company. No funding. No team. Just a document that kept growing because people kept finding it useful. Google killed it with a DMCA takedown in 2023. What happened next is the part worth understanding. THE REBUILD The community did not petition Google. They did not hire lawyers. They did not start a campaign. They rebuilt it on their own domain, mirrored it to GitHub, deployed it to IPFS, and distributed it across 12 backup domains simultaneously. In doing so they accidentally built one of the most resilient information architectures on the internet. No central server. No single point of failure. No CEO to pressure. No hosting provider that matters. When you remove every central point of control from a system the only way to kill it is to kill the internet itself. Hollywood has not figured out how to do that yet. THIS IS THE FUTURE OF INFORMATION What nbatman built without intending to is a blueprint for how information survives in an era where platforms can disappear anything with a single policy decision. IPFS does not work like a normal website. A normal website lives on a server somewhere. Find the server. Pressure the host. Site goes down. IPFS stores content across thousands of nodes simultaneously. There is no server to find. There is no host to pressure. The content exists as long as at least one node in the network holds a copy. This is the same architecture behind every major blockchain. It is the reason Bitcoin cannot be shut down by any single government. Applied to information it means the same thing. No single entity can decide what survives and what disappears. WHAT THIS MEANS FOR BUILDERS Every platform you build on right now has a terms of service. Every terms of service has a clause that can remove you without notice. X. YouTube. Substack. Medium. All of them. The builders who understand decentralized infrastructure are not just building products. They are building on foundations that no platform can pull out from under them. IPFS. Nostr. Distributed storage. Peer-to-peer protocols. These are not niche technologies for crypto enthusiasts anymore. They are the infrastructure layer for anyone who wants to build something that lasts. THE LESSON FROM SIX ANONYMOUS VOLUNTEERS Six people. No salaries. No office. No investors. Maintaining something that the most powerful entertainment companies on earth cannot destroy. The lesson is not about the content they organized. The lesson is about what becomes possible when you remove every central point of control from a system and distribute it across a community that believes in what it is building. That architecture is available to every builder reading this right now. The question is whether you are building something that a single policy decision can erase or something that survives because no single decision can touch all of it at once. nbatman did not set out to answer that question. He just made a Google Doc in 2018. The answer found him anyway. Follow @cyrilXBT for the exact tools, protocols, and infrastructure decisions that matter for builders who want to build things that last.

• 
  Brian Anglin (@BriansAngles) reported 18 minutes ago

  Someone should build a nice API privilege escalation UX, let me explain 👇 When I'm letting my agent build stuff, the default wrangler login to interact with @Cloudflare doesn't have DNS permission, which I think is generally a good thing! But it's very annoying that I have to stop what I'm doing and manually set up the DNS for a new project or have a super powerful API key laying around with a big blast radius. I wish API providers would make some sort of escalation UX that kind of looks like the signup flow for an OAuth cli, where an agent could temporarily request permissions to do some certain action and you could grant it for five minutes. Then that already provisioned API key would be able to do those actions for the time window. Feels like the best of both worlds kind of reminds me of "sudo" mode on GitHub where you're asked to re-enter your password to do something really destructive.

• 
  Füat.lfg (@0xFAlpha) reported 19 minutes ago

  @iowelosu @RialoHQ Some react after problems. Others design systems that prevent them. The GitHub incident made that difference clear. @RialoHQ is clearly building for the latter.

• 
  Harsh Kapoor (@0x_Kapoor) reported 21 minutes ago

  Here’s your cheat sheet. Bookmark this. Set up once: /init: Auto-generate documentation as CLAUDE.md /memory: Set global preferences forever. /pr_comments: Load GitHub PR comments into context. Daily Use: /btw: Ask side questions without interrupting. /compact: Compress conversation, keep going. ! command: Run shell without leaving. /cost: Check your token usage. Power Moves: /fast: Toggle faster responses. /review: Systematic code review. /model: Switch models mid-session. When Things Go Wrong: /clear: Wipe conversation, keep setup. /doctor: Check your setup health. /terminal-setup: Fix terminal integration. /help: See all available commands.

• 
  Censored Hacker (@censoredHacker) reported 24 minutes ago

  @GithubProjects A @github clone that doesn't go down every week and that doesn't leak my projects using #GHCP

• 
  EccentricVictor (@EccentricViktor) reported 24 minutes ago

  @Dappitdotio I'm finding it hard to sign in with my GitHub, honestly

• 
  Silvery Fighters Jr./CG-TAN (@SilveryFJ) reported 25 minutes ago

  You can still download Linux Multimedia Studio from its Github page but same cannot go for the assets from the site as it remains down. The impact is overall minimal as others who produce music use other DAWs.

• 
  Armandas Kaleinykas (@Eykarm) reported 26 minutes ago

  Three Cloud Computing labs at Vilnius Tech — Wireshark packet analysis (ND1), a private Dark Age of Camelot server two ways (ND2), and a Hugo + GitHub Pages blog (ND3) — form a tidy OSI-layer climb from raw packets to one-command deploys.