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:

• 
  Jack O'Brien (@thejackobrien) reported 6 minutes ago

  @Mike_Walsh_19 Maybe they'll run into Github-level problems with scale, but man the network effects are strong. Don't see Slack going anywhere anytime soon, but this move might force all sorts of crazy moves (OpenAI creates a slack clone? Startup like @saradu's Ando rises to the top fast?)

• 
  cemaxecuter (@cemaxecuter) reported 10 minutes ago

  @OSINT_OpSec GitHub was slow, try again ha!

• 
  SUDDENDESU (@suddendesu) reported 16 minutes ago

  My robots.txt is configured to specifically allow crawlers and scraping. I also maintain the whole site on github, so anyone can pull the whole thing at any time. So, no, I'm not particularly pissed if someone scrapes my site. But I understand my site isn't as big as TCRF, so if the issue is about *excessive bandwidth usage* then I agree.

• 
  Rouzbeh (@JustRouzbeh) reported 18 minutes ago

  @ClaudeDevs This only works if it really sits there until I send it. If it starts opening GitHub issues on its own, that is going to get messy

• 
  Kunal (@kunmath00) reported 20 minutes ago

  @pierceboggan I've had problems in making this work. The sessions are not starting and giving errors. Raised an issue on github

• 
  TEJ (@sartejt) reported 21 minutes ago

  @gregisenberg The concept of *** as a versioning protocol goes far beyond what normies understand GitHub isn’t remotely hard to understand once you understand *** Tldr; GitHub is not the problem

• 
  Robby Seventeen (@Robby_Seventeen) reported 21 minutes ago

  GitHub actions seems to have so many issues lately. Who is self hosting alternatives and what's specs do you run.

• 
  Mofe (@Mofecloud) reported 21 minutes ago

  github down?

• 
  Rituraj (@RituWithAI) reported 22 minutes ago

  🚨BREAKING: Researchers just proved that every AI agent controlling your Android phone collapses the moment something unexpected happens. A popup appears. The wrong screen loads. A button moves. An app crashes mid-task. All 16 leading Android GUI agents tested. Every single one degraded significantly under these conditions. It's called AnTrap. A benchmark that injects the kind of real-world chaos that exists on every actual Android device — and measures whether AI agents can handle it. Here's what was actually tested. Every Android GUI agent benchmark until now tested agents in perfect conditions. Clean screens. Predictable flows. No interruptions. The agent taps, the expected screen appears, the agent continues. Real phones don't work like that. AnTrap injected four categories of real-world disruption into agent execution: → State anomalies — unexpected popups, wrong screens loading, UI elements missing → Thinking anomalies — the agent's reasoning gets disrupted mid-task → Action anomalies — buttons move, taps misfire, actions produce wrong results → Round anomalies — tasks get stuck in loops, states become deadlocked Then they tested 16 leading models — every major GUI agent available — against all four. Universal vulnerability. Every model. Significant performance degradation. Across the board. Here's the finding that makes this structurally alarming. Some failure modes are fixable. Single-step traps — unexpected popups, misplaced buttons — can be largely resolved by training agents in adversarial environments. The agent learns to handle interruptions. But deep contextual traps — particularly state deadlocks, where the agent gets stuck in a loop it cannot recognize or escape — expose something that training alone cannot fix. An intrinsic reasoning limitation. The agent doesn't know it's stuck. It keeps trying the same approach. It never asks for help. State deadlocks aren't learnable. They're architectural. Here's why this matters for every AI agent controlling your phone right now. ClawGUI. Computer Use. Every phone control agent being deployed to millions of users operates in exactly the environment AnTrap simulates. Real phones. Real apps. Real popups at inopportune moments. Real UI changes between versions. Every agent was benchmarked in ideal conditions. None were tested in real ones. Until now. 11 upvotes on Hugging Face. From Zhejiang University, Yale, and collaborating institutions. Code available on GitHub.

• 
  bonduelle (@bonduelleioat) reported 26 minutes ago

  YOU DON’T NEED 10 MORE AI SUBSCRIPTIONS. YOU NEED YOUR OWN COMPUTER. When you’re burning $600 every month on AI subscriptions, only to get hit with a RATE LIMIT exactly when you’re racing against a deadline - you feel like a complete idiot. And the worst part? You’re not paying for the result. You’re paying for the right to temporarily use someone else’s hardware. But that model is starting to crack. Elon released a working agent openly on GitHub. Moonshot put a powerful Kimi model on Hugging Face under an MIT license. And a modern mini PC with 256 GB of memory no longer looks like something only corporations can afford. Install Ollama. Drop a few lines of configuration into ~/.grok/config.toml. Point the local base_url to: And suddenly, a huge part of your AI infrastructure no longer needs to live in someone else’s cloud. 80% of routine tasks → local Kimi-Linear-48B in Q4. No queues. No rate limits. No sending every request to someone else’s server. No more $49, $99, or $200 monthly bills for “premium access.” And the other 20%? Keep a targeted API key for the hard stuff. That’s where the math gets uncomfortable for the entire subscription industry. Instead of paying $600 every month, you run your own hardware and pay for the electricity. Even at around 140W, the electricity cost can be orders of magnitude lower than $600, depending on your tariff and how many hours you run the system. You’re no longer renting AI. YOU OWN IT. And perhaps the most dangerous question for Silicon Valley right now is incredibly simple: WHAT IF WE DON’T NEED ALL THESE SUBSCRIPTIONS AT ALL?

• 
  Alonso G (@CryptoDegenTopG) reported 26 minutes ago

  @XPeterKappes @bot @cursor_ai Same issue here. Mine is connected with my GitHub and I am getting the same error

• 
  József Schaffer (@angelday) reported 28 minutes ago

  @AndyHewco If you have the source on GitHub I could look into the Mac issues real quick.

• 
  WhiteCap Data (@WhiteCapData) reported 30 minutes ago

  @nicbstme @github That screenshot data could flood the PRs if not normalized with the rest of the issue metadata.

• 
  Ranjan Goel (@ranjangoel) reported 33 minutes ago

  Is @github down for folks? Codespaces is suddenly out for me. @githubstatus

• 
  Aurorastar (@G_ameman) reported 33 minutes ago

  02:37:39–02:40:08 02:37:39 You just tell it to keep going until you tell it to stop. 02:37:45 It ended up... 02:37:45 I think we ended up with a 46 time execution improvement over the original. 02:37:51 It's incredible. 02:37:52 Absolutely incredible. 02:37:52 It's just incredible. 02:37:54 Um, I do, you know, I just don't think there's anything that compares to Fable in terms of planning. 02:37:58 So I usually do Fable for planning and for reviewing, and then something else for implementation, like Opus 5. 02:38:08 And it's really nice. 02:38:08 It's a, it's a really nice setup that allows you to not run out of tokens too much. 02:38:14 What I found, it's really interesting because Fable is, in my opinion, the best model right now, but it also makes mistakes. 02:38:18 And the best way to get the best software, I would actually rather have two differently sourced sort of... 02:38:25 I mean, they're not mid-tier, they're all frontier. 02:38:29 But have, let's say Opus 5 and Codex and have one check the other's job. 02:38:35 This is my standard operating procedure now. 02:38:40 I'll have Opus or Fable do the work, and then I always end it, review with Codex xHigh. 02:38:47 And I've also started using Grok just to test it out, and it's also quite good. 02:38:47 And it keeps finding stuff. 02:38:51 And then that's my workflow when I'm having my agents on my own machine do it, and then I push to GitHub. 02:38:55 And then Copilot, I kid you not, has actually gotten good. 02:38:59 Copilot keeps finding stuff that's legitimately broken, which is also incredible acceleration because the first version of Copilot that started doing this was literally retarded. 02:39:09 It would just constantly flag things that were nonsense. 02:39:13 It would constantly flag the same problem over and over again as you would push. 02:39:13 It was really annoying, so I think a lot of people actually ended up turning that off. 02:39:17 And if they did, they should turn it back on because it's actually quite good. 02:39:24 Keeps finding things. 02:39:24 And if you then take that, and we shouldn't be surprised. 02:39:24 Why are we surprised? 02:39:27 Even if you're a good programmer, if you finish a job and you ask your also very good peer to review it, you're gonna end up with better code. 02:39:35 Of course you're gonna end up with better code. 02:39:35 So build that into your process. 02:39:39 Pick one of the agents to drive with. 02:39:39 I've mainly been driving with Claude, which is actually interesting because I have some other reservations about Anthropic. 02:39:47 But the reason I'm sticking with Claude is, in my opinion, they actually have the best harness. 02:39:53 And one of the reasons it's the best harness is it's multi-agent running. 02:39:57 So if you wanna run multiple agents at the same time, you can do arrow left when you're inside a session, then it goes back to Agent View. 02:40:05 And here in Agent View, you can pick up another agent. 02:40:08 So if you wanna do this thing where you have multiple threads going on the Claude Code is just the nicest setup.