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:

• 
  Q. (@qdr_tech) reported 5 minutes ago

  @zehf probably unpopular but neovim if I actually want to write code otherwise most of my agentic dev workflow starts from the issue tracker now. agents pick up issues, open PRs, and I mostly drive from inside github

• 
  Prince Soni (@princesoni_oo1) reported 6 minutes ago

  I got tired of building projects that ended as soon as I pushed the code to GitHub. So I decided to build something around a problem I actually experience. I go to the gym regularly, and one thing kept standing out: Progress matters. But before progress comes consistency. so....

• 
  avery (@codeshaunted) reported 11 minutes ago

  @mgaruccio @Mithra_Is_Fine github is down bro, i had to do it the old fashioned way

• 
  Harshit Sharma (@itsharxit) reported 12 minutes ago

  @cesp2099 Partially agree. Can’t speak for Linux, but on my end T3 Code feels way snappier than Codex Desktop. And “it only got famous because Theo is famous” hell no. His audience obviously helps, but that doesn’t magically get you that many stars. If you’re actually hitting these issues, a detailed GitHub issue with repro/context would probably be more useful. Gives them something concrete to fix, and everyone wins.

• 
  Mike Kamo (@mikekamo) reported 13 minutes ago

  Cursor launched its own GitHub competitor this week. The part everyone's talking about is the timing. It went live the same morning GitHub was down with error rates near 20 percent. TechCrunch called it capitalizing on the frustration. Whether they planned it that way or got lucky, the lesson is the same. The best growth moment your company will get isn't a campaign you scheduled. It's the hour a competitor stumbles. Their outage, their price hike, their bad launch, the thing that makes their customers suddenly open to a switch. Almost nobody is ready to move when that window opens. I've watched this play out at NP Digital more times than I can count. A competitor drops the ball on a client, and the agencies that win the account aren't the best ones. They're the ones who called that same day. Most were still putting together a proposal a week later, after the window had already closed. The teams that catch these moments aren't faster thinkers. They just decided ahead of time that shipping at 80 percent beats being ready at 100 percent next month. You can't schedule your competitor's worst day. You can be the one thing that's ready when it comes.

• 
  Robin Gattis (@ColdShalamov) reported 15 minutes ago

  @PaulSolt Yes I remember this, I had codex do a sort of continual learning on the skill when I was working with it. It has a natural inclination to chunk the work too small and you have to encourage it to give more sizable tasks. The issue is that too large and it times out in the Pro thread and won’t make the PR or the zip file and lose the work also. My average “$delegate-to-ChatGPT” task would be like “I’m making this game, here’s the spec, we need to make 20 high quality blender assets for the game, make 20 ChatGPT pro threads and log the thread urls, upload the blender installer to the first thread and tell the rest to pull it from the library, tell them to generate 5 reference images, pick the best one, use that as a basis to replicate in blender. Each thread must iterate/improve on the blender model at least 20 times converging on 1:1 parity with the reference before stopping” and the skill specifies every prompt to use the GitHub connector with a zip file fallback. Something like that One thread couldn’t handle something that big, and codex would run for 3 days. Codex has a tendency to get stuck in loops just building the verification infrastructure and I think Pro is more apt to actually do the job done unless it’s physically impossible within the timeout. Another issue is that you have to spend some time getting codex to script the browser interaction into basically an api call so it’s not eating screenshots and using too many tokens rendering the process moot. I actually have a permanent codex thread that runs like this, and it’ll power down and poll every 10 mins keeping 5-20 threads open at all times depending on how many tasks it has. I have ChatGPT pro make PRs with plans using my iOS app and the skill pulls and delegates them to threads and has subsequent threads review them before merging them. It took some tweaking but I use probably 1/5 the codex usage for the same amount of work and it feels more persistent than stop-and-go threads that get left behind I also had to up the codex context to 500k tokens to orchestrate that better

• 
  Tony Sibert (@AiRiskBytes) reported 17 minutes ago

  For years, the thing protecting you after a credential theft was how boring it was to test them by hand. That's over. New research on three unrelated threat actors using AI through the whole intrusion. The number that matters isn't the tooling. It's this: one operation collected 2,975 validated keys from 1,742 hosts in seven weeks. Validated. Stolen credentials used to arrive as a pile of strings someone had to test by hand. That triage was slow, boring, and (entirely by accident) one of your better defenses. It bought you the days between theft and use. It isn't slow anymore. And the haul included SSH private keys, AWS keys, and API credentials for Gemini, OpenAI, GitHub and Anthropic...a class of secret most organizations don't inventory, don't rotate, and can't report the spend on. If time-to-abuse is now hours, faster detection isn't the answer. Shorter credential lifetime is. Can you name every model-provider API key your organization holds, and say when each was last rotated?

• 
  Suni (@suni_code) reported 17 minutes ago

  Never start Vibe Coding before understanding: HTTP APIs *** & GitHub JavaScript Databases SQL CRUD Authentication MVC DNS Debugging Deployment JSON Servers Basic security Frontend Backend Environment variables Data structures System design Error handling

• 
  Ashutosh Bhandari (@withashutos) reported 18 minutes ago

  Every hiring post today ends the same way. "DM me your github." Then nobody opens it. Not because they don't care. A stranger's repo list tells you almost nothing about what they built versus cloned, what got reviewed, what they threw away. The signal was never the profile. It's per project, and nobody writes it down.

• 
  Crypto Banter (@crypto_banter) reported 19 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.

• 
  fantom (@fantom7z) reported 23 minutes ago

  @melvindvivas Chatgpt Pro, Claude Plus, Google One Pro, Chatgpt Pro in web make the plan and analyse my github, gemini execute. When gemini does it bad I ask Claude to fix the details

• 
  Issam Hakimi (@killix) reported 27 minutes ago

  @RituWithAI A co-author trailer covers the PR, it never maps to lines. GitHub points at 094038e from Aug 2025. Wiz points at the June 18 merge, same history both times. GitHub's own security bot reviewed the final revision with the issue-title injection in it and stayed quiet.

• 
  Larry Chiang, 650-283-8008 (@LarryChiang) reported 31 minutes ago

  Was asked how I think about the things I work on holistically. I break them into 3 buckets: 1. Random productivity hacks or experiments. These are usually an evening or weekend vibe-coding session. If something ends up useful or packaged enough, I’ll push it to GitHub. 2. More in-depth products with no intention of making money. I use these as playgrounds for learning and experimenting with AI. They can easily turn into 100s of hours of focused product + AI work, but they’re a great way to test workflows and hopefully build something useful for the world. Examples are GPT Food Cam and Speakrphone. 3. Things that can make money. These tend to be more deeply researched problems, often in a specific vertical, where there’s a clear user, buyer, and willingness to pay. A lot of what I build at work falls into this bucket.

• 
  Gnyani !! (@gnyanendranath) reported 35 minutes ago

  GitHub ate my Monday. Eight hours. Couldn’t merge. CI was dead. Copilot wouldn’t even sign in. Cursor’s agents went down with it. Claude’s been blinking out for weeks. At some point, outages stopped feeling like rare incidents.

• 
  Teknium 🪽 (@Teknium) reported 37 minutes ago

  @Cobalt_Peak Sorry about the delay - its in now. Github outage has been a PITA!