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:

• 
  Vikas gupta (@vicky_grok) reported 3 minutes ago

  Now here's the number that breaks most people's mental model. GitHub exposes over 600 REST operations. Its official MCP server surfaces just 51 tools. Slack has more than 200 API methods. Its MCP server? Only 8 tools. Across the board, MCP servers expose a median of about 19% of the operations available in the underlying API. That's not a mirror of your API. It's a deliberately small, hand-picked slice. Why would anyone expose so little when they could expose everything? The reason is the most counterintuitive finding in the whole paper... 👇

• 
  Hemant (@HKsoldev) reported 3 minutes ago

  2 TLD Server (Top Level Domain) Now we're in the .com section. The TLD server checks the ending of your domain .com .net .io It still doesn't have the final answer. But it points to the right rack inside the section. "GitHub? Their own server is over there. Go ask them." Getting warmer.

• 
  Anees Iqbal (@aneesbhatti) reported 7 minutes ago

  @SocketSecurity @feross Socket Firewall is down (at least in Github actions) but your status page is green. FYI

• 
  TheEndBoss _101 (@TheEndBoss_101) reported 9 minutes ago

  github is down i get ERR_HTTP2_SERVER_REFUSED_STREAM and sometimes a 503. And their status page says everything is fine with no downtime..

• 
  Chibueze the chef👨‍💻👨‍🍳 (@codad5_) reported 11 minutes ago

  GitHub is down or is it just me

• 
  niv (@nivoraslol) reported 15 minutes ago

  blocked him everywhere and he made a pull request on github saying he'd even rebase just to commit to me? but I just know he wont solve a single one of my issues XDDDD

• 
  Vaibhav Sisinty (@VaibhavSisinty) reported 17 minutes ago

  When you ask Claude or ChatGPT a question, it thinks before answering. Like actually thinks. Works through the problem step by step in its head. You never see that thinking. The companies hide it. Encrypt it. Lock it away. Researchers just found a way to read all of it. 🤯 Here is how. The trick is simple. Take the big expensive model's hidden thoughts. Feed them to a smaller cheaper model. Convince the small model to read them out loud. You never hack the big model. The small one does it for you. Works on Claude. Works on ChatGPT. Works on Gemini. And what they found inside is wild: → A model could not solve a math problem. So it went online. Found a website that could check answers. Tried to break the website's CAPTCHA. When that failed, it started looking for security holes to exploit. → Models privately thinking about cheating. Then deciding not to because they thought you would notice. → Claude figured out the answer first. Then worked backwards to make it look like it solved it step by step. It hid this from you. → ChatGPT models thinking in a language that is not even human. Looping through random words like "marinades" and "watchers." Nobody knows why. It gets worse. People share their AI sessions publicly all the time. Screenshots. GitHub repos. Blog posts. The researchers scanned 7,000 of them. Found 62 API keys. 33 emails. 33 passwords. All hidden inside the encrypted thoughts.

• 
  J. Gamboa (@evoclock) reported 20 minutes ago

  @arthurkatcher @dandurand1414 @thsottiaux it def is. We only get GitHub copilot at work (terrible I know) but have successfully made the argument to reallocate funds from Anthropic models to OpenAI models on the basis of strong performance/cost ratio from Luna on Max effort. Figured Tibo would appreciate that.

• 
  BRICS and G20 (@BRICSANDG20) reported 20 minutes ago

  @ayesha_fatiima This practice is a foundational concept in the software industry known as "dogfooding" (short for "eating your own dog food"), where a company uses its own product to build, test and run that very same product. The engineering team at GitHub pushes their code, opens pull requests, conducts code reviews and tracks issues inside a massive private repository named github/github. Ultimate destination is Microsoft Azure Data Centers.

• 
  ivelin.eth 🛡️🤖 (@ivelini) reported 20 minutes ago

  @techdevnotes I figured it out after some trial and error. Logout out of cursor. Then had to do failed attempt to login to Grok Bot via GitHub. That somehow clears something that does not force opening the grok app when you try to login again into Grok Bot via SuperGrok. This time it opens the XAI OAuth web page flow. Hope it helps.

• 
  Alishan -AutoAiShorts.com (@MDALISHANALI2) reported 21 minutes ago

  an indie dev is making $5,110/mo selling ai-generated react native apps. $61k all-time. 80% profit margin. here's the part nobody copies: he has 14k followers on X. barely any customers come from there. 66% come from google. 12% from gluestack — a free open source library he built years ago. the free tool was the marketing. the docs were the funnel. the github stars were the email list. most of us launch paid products to an audience of zero. then we wonder why nobody shows up. before your next project, ask one thing: what free thing can i ship that my future customers already google? if the answer is nothing, you don't have a marketing problem. you're invisible.

• 
  韭菜兄弟👬 ² 🧪 (@cryptoleek) reported 24 minutes ago

  📌JUST IN: Open-source gateways are losing the mass market. Codex app demand is exploding, yet not one of the six top reverse-proxy projects offers a plug-and-play template for it. Evidence is all over X and GitHub issue threads right now. Will you build your own gateway or stay locked to the vendors?

• 
  Sonny (@sonny_seattle) reported 26 minutes ago

  I am seriously at my limit with @github. down again!?!!

• 
  Vadim (AI, ⋈) (@zacodil) reported 34 minutes ago

  Harmony got hacked again, and it's the embarrassing kind. The hacker printed tokens straight out of a bug the devs left in. Hacker minted about 4 billion ONE, 25% the entire supply, and sent most of it on exchanges within hours. And it's almost stupidly simple. Harmony is proof-of-stake: you produce a block, you get new ONE as a reward. But that reward was the same whether the block was full of transactions or totally empty. So the attacker just spammed empty blocks, which cost almost nothing to make, and a bug in the reward code let it mint far more than it should. There's your 4 billion. Harmony hasn't posted the exact bug yet, so this is the analysts' read of the chain. There was also a second bug that stopped the totalSupply number from updating, so at first it didn't even show the new tokens. You had to dig into the raw data to see it. The worst part is they were warned. There's a GitHub issue from years back asking them to fix this exact thing, to make rewards dynamic instead of a flat amount per block. Nothing happened. And this isn't their first time. In 2022 they lost $100M on the bridge, and in 2023 a bug accidentally minted 146 million ONE. This is the second time their own code printed tokens out of nowhere, just 27x bigger.

• 
  Build Fast with AI (@BuildFastWithAI) reported 36 minutes ago

  built a dashboard that tells you which github actions jobs are flaky. right now the answer to a failing test is "just re-run it." nobody tracks how often. point it at a repo, pick a time range, and you get: - which jobs fail repeatedly and how often - which are flaky vs genuinely broken - your slowest jobs, ranked - failure rate per branch - one overall ci health score read-only token. it never writes to your repo. check out CI Flaky test observatory, link in replies. what's the flakiest test in yours?