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:

• 
  Hrishikesha (@HrishikeshaNFTs) reported 11 seconds ago

  @Itsbrutox Lol. It took 20+ retries and it still failed to fix a silly github wf build error. On top. It lazily kept saying hang tight doing nothing. Forget opus, a random HF open model do much better.

• 
  Gustavo Alessandri (@webgus) reported 2 minutes ago

  If you find an error, have an idea, or want to propose an improvement, just open an issue or fork it on Codeberg or GitHub. Contributions are welcome. That’s exactly the point.

• 
  Nathan Colosimo (@thenathancolo) reported 2 minutes ago

  @jdalton Yeah, 100% agree, actually mentioned that in the tweet, the bun port actually isn't a good example lol But got me thinking a lot about the measurement problem. Who is active? Who is good? Who cares about what they do? Jarred is all 3 of these to the power of 1 million, but you don't see that just from a github graph

• 
  zSkerWizrdz (@zxxkgkillerxxz) reported 4 minutes ago

  PC gamers who use DLSS Swapper have been given a security warning. The app’s creator says a user uploaded a fake DLSS file that contained malware. He warned: “DO NOT download these files, they are likely malware.” The problem is not with DLSS Swapper itself, but with files uploaded by other users through its GitHub repositories. The developer recommends only downloading DLSS files from trusted sources like NVIDIA, official game installs, or verified releases #NVIDIA

• 
  Ray 👨🏽‍💻 website & app developer (@Raynerdtech) reported 6 minutes ago

  Yesterday I posted: “**** programming. **** Java. **** databases. **** servers. **** networking.” A lot of people thought I was joking 😭 I wasn’t. The code worked. The problem was everything around it. Deployment. Servers. Databases. Infrastructure. The funny part? Most of that headache could have been avoided if I knew about Symplax earlier. Built by my guy @LazyCode3 The easiest way to describe it: Vercel for your own VPS. GitHub deploys. Databases. Metrics. Backups. You keep full control of your infrastructure without the usual VPS pain. Self-hosting shouldn’t be harder than building. Link in the comments. 👇🏾

• 
  Yuan John (@yuanjohn01) reported 7 minutes ago

  @zarazhangrui There is currently an issue on GitHub regarding CodeX's design process. Occasionally, it mistakenly inserts the design requirement keywords you provide directly into the frontend design placeholders, which is essentially a bug.Furthermore, CodeX's underlying design capabilities have inherent flaws. Even when using MCP or skills like 'Creative Production', these limitations cannot be fully overcome.

• 
  Daria (@dariacupareanu) reported 10 minutes ago

  A senior engineer (Matt Pocock) put his whole AI skill library on GitHub. It's for coders, but it fixes 4 problems every AI user has: it builds the wrong thing, gets vague, breaks things, quietly makes a mess.

• 
  Atul Mishra (@atulmishra1996) reported 11 minutes ago

  I don’t understand why people want another github ? What’s the problem in current ?

• 
  Yannick Comte (@cyannick) reported 14 minutes ago

  @NimaZeighami You can already do it but it’s pain in the a**. If you look at the issues on GitHub someone was able to do it and ran beat saber. My plan is to package that into something easy to run.

• 
  Mikhail Rogov (@i_mika_el) reported 17 minutes ago

  @AyushSarode07 then probably not a GitHub-wide issue. I would check the workflow/status checks on that exact commit first, warning sign usually means one check or integration failed.

• 
  Kevin Whinnery (@kevinwhinnery) reported 21 minutes ago

  @threepointone This was after a configuration error on our Stainless SDK repos. Some Stainless customers were temporarily added as outside collaborators in Anthropic's GitHub enterprise. All resolved now and no data was exposed, details were emailed to affected customers 🙏

• 
  Electronic Intelligence Agency (@EI3065) reported 22 minutes ago

  @github @LinkedIn prevents acess for selected nationalities with programers doing imposible security checks on login; on repeat level of app becomes low of low for conflict

• 
  Cointelegraph (@Cointelegraph) reported 23 minutes ago

  🚨 UPDATE: Base says the B20 Token Standard launch has been delayed due to a GitHub outage.

• 
  War Alerts Analysis (@War__Alerts) reported 27 minutes ago

  China’s next big cyber weapon may not be a hacker group. It may be an AI agent that never sleeps. RealClearDefense reports that Chinese AI lab DeepSeek is hiring to build an “AI agent” that scans code and finds vulnerabilities. That’s more than a smarter static scanner. It points to agentic AI in cyber operations: systems that can plan steps, call tools, run code, and iterate toward an objective. In a state where the barrier between commercial labs and security services is thin, that kind of agent is not just a developer convenience. It is potential infrastructure for scalable, semi-autonomous cyber power. Until now, serious cyber exploitation has been constrained by people: elite operators to hunt bugs, maintain toolchains, and weaponize zero-days. Agentic AI changes the equation. If a model can triage huge codebases, propose likely weaknesses, generate proof-of-concept exploits, and refine based on error messages or partial success, then cyber capability becomes less about headcount and more about compute, training data, and tool integration. That is the strategic shift DeepSeek hints at. China already invests heavily in cyber and chases AI self-reliance. A low-cost domestic model stack tuned for vulnerability discovery is exactly the kind of sanction-resistant engine Beijing would want to grow offensive and defensive capacity in parallel. The West is not standing still. Pentagon work on “AI for Cyber Operations” and allied projects on AI-enhanced red-teaming and GitHub-integrated scanners show the same direction of travel. Both blocs are converging on the idea that AI should do the grunt work of finding and probing weaknesses at scale. The shared policy trap is to treat AI-for-cyber as a secret edge to be maximized, rather than a capability that will leak, proliferate, and empower everyone from state operators to ransomware crews. Once a capable agentic model for exploitation exists, containing it is hard. We have already seen how quickly model weights, jailbreaks, and red-team tools bleed into the wild. A stolen checkpoint, a neutered copy on a gray-market cloud, or a contractor selling access could turn what was meant as a state asset into a commodity service for criminals and proxies. Key watch points now: does DeepSeek’s cyber agent stay framed as a passive analysis tool, or does it move into active exploit generation and tool orchestration? And do Chinese and Western governments publish any rules of the road, or do they run classified races with no guardrails and no minimum norms? If both sides push agentic cyber AI without restraint, the winner is not China or the United States. It is whoever gets their hands on the leaked tools next.

• 
  Solomon De Leon (@solomonjcdeleon) reported 28 minutes ago

  I grew up from a young age playing/learning the guitar seriously Dad was a musician Learned it the right way took lessons, theory, scales, reading, aural training, exams and all that But then... I found out about tabs years after (my dad frowned upon me using it) Tabs are basically a number system to play the songs you want No theory needed, don't have to read the notes, don't have to understand scales and stuff just press the fret of the number you see on the screen It's probably how 80% of hobby guitarist learn to play the songs they like And it's nothing wrong with it tbh the end result you want is to play the song So now, vibe coding/creating with AI is like playing the guitar through tabs You can play the song but you have no idea how it works So vibe coding/creating anything for that matter works It just depends on what your goals are And for most people, "using tabs" is the right call If the goal is to ship, to move fast and test ideas, burning months on theory first is the wrong trade The tab player is playing the song this weekend. The theory student is still on scales Then again, even tabs need a floor You can't just read numbers off a screen and expect music to come out, you still need to know how to hold the pick, where your fingers go, how to fret a note cleanly Vibe coding has the same floor. If you don't know what GitHub is, what an API does, or how a database talks to your app, you won't even get the vibe coding to work in the first place So learn the basics. Version control, how to read an error log, what the moving parts of an app actually are and how they connect. Enough to not be flying blind. Because the person who only knows tabs hits a ceiling fast They can play the songs but can't write their own, can't improvise, and the second something breaks they're stuck Vibe coding is the same You move fast, right up until something breaks And the real trap is not that AI can't fix it. AI is an executor, it'll happily keep trying It's that you don't actually know what's wrong So you fall into slot machine prompting, pulling the lever again and again hoping the next prompt is the one, with no idea what you're even looking for Tabs can take you far, but if you want to write your own (good) music, you need the theory, you need to know what's happening under the notes So who should vibe code? Honestly most people who want to build simple but powerful tools Who should vibe code seriously? product, biz, marketing guys who are willing to learn If you're on the business side and you want to ship, test, get something real in front of people, vibe code away, it's the right tool for that Just know where the ceiling is, and learn enough of the basics that you're never fully at the machine's mercy Afterall as a non technical, your alpha is finding out what the right thing to build is (what people want), and selling it (monetizing it) Let the engineers take your 0-1 to 1-100 And if you want to be great, to be the engineer who builds the hard things, go for it and learn it properly, AI will still write code for you but you'll knock it out the park That skill isn't getting replaced. If anything it's worth more now, because everyone else stopped at the tabs Move fast with the tabs. But if you ever want to write your own music, learn the theory.