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:

• 
  Clayton Correia (@claytoncorreia) reported 9 seconds ago

  @AnnikaSays I'm using Codex a lot via mobile but not remotely to my desktop. I've also found the connection to be rough. I've found it rock solid when just using the codex via cloud option connected to a github repo. Workflow is something like this: 1) A create a linear ticket for a bug or something on my phone with some details 2) I tag codex in the comment with something and tell it to "fix it" 3) codex spins of a virtual environment of my app, makes the changes and creates a pull request 4) when I'm back from walking the dog or whatever I pull down the change and test it

• 
  sanixdk (@sanixdarker) reported 28 seconds ago

  WHY THE HELL IS GITHUB SO SLOW MAN ??? you can't make this sh_t up, am so fckng fed up rn...

• 
  pharrelly (@pharrellyhy) reported 2 minutes ago

  @thsottiaux renewed subscription while the weekly usage not reset. pls fix it, saw similar issues on github for few weeks

• 
  Lakshya (@tokenaware) reported 8 minutes ago

  i booted codex and it started working on the repo unprompted. It just took an issue from github and raised a PR. I'm not ready for this kind of autonomy

• 
  Nicolas Finet (@nifinet) reported 8 minutes ago

  Everyone's a builder now. Congrats. You and the 50 million other people who just figured out they can vibe-code an app over a weekend. Look at the commit trend on GitHub. A billion last year, on track for 14 billion this year. Half my timeline is people announcing they shipped something before lunch. The slop is coming, and there's going to be a LOT of it. And honestly? Good. More people building more things is a great thing. A non-technical founder shipping a real product from a Slack message is wild, and I'm here for it. But that's also the problem for everyone busy celebrating it. A business was always two halves: build the product, and get the customers. Product is where you used to win or lose. Now everyone clears that bar on day one, so the whole thing collapses onto the half builders love to ignore: distribution, lead gen, getting in front of the right buyer before the other 50 million do. You can ship the best product in the world. You'll still lose to someone with a worse one who actually bothered to learn how to sell it.

• 
  Avi Chawla (@_avichawla) reported 8 minutes ago

  Claude Code without this new tool is like *** without GitHub. Claude Code stops at the boundary of your terminal. - It can't see what's happening in production right now. - It doesn't know which PR broke the checkout service. - It can't tell why a Datadog alert got fired. - It can't see the Slack thread where the team decided not to touch the retry logic. These are operational and institutional memory gaps that eat up engineering time every single week. The solution is now actually implemented into the @coderabbitai Agent. It lives inside Slack and connects to repos, issue trackers, docs, monitoring, and cloud infra. When a production alert fires, you can mention it in the thread, and it traces the problem through your APM data, finds which recent PR caused it, and can open a targeted fix without anyone switching between five different dashboards. When the incident is resolved, it can document what happened and create a ticket in Linear with the timeline, root cause, and relevant PR links. Note that this is not a one-off assistant. The agent retains what the team decided across threads, channels, and the entire org. So the context from this incident is already available next time someone touches the same service. I've shared the link to try CodeRabbit Agent for free in the replies. Thanks to CodeRabbit for working with me on this post.

• 
  AwesomeAI (@Awesome_AI_News) reported 10 minutes ago

  Microsoft GitHub has urgently taken down dozens of open-source repositories due to hacker attacks injecting password-stealing malware. Affected projects include Azure cloud services and popular AI development tools like IDEs, causing shock in the developer community. Security firms Cloudsmith and OpenSourceMalware first detected the anomaly..... 微软GitHub上的数十个开源项目仓库近日紧急下线，因黑客入侵并注入窃取密码的病毒代码。受影响项目集中在Azure云服务和热门AI开发工具，如集成开发环境，引发开发圈震动。安全公司Cloudsmith和OpenSourceMalware最早发现异常。

• 
  Snake (@wise_snake69420) reported 10 minutes ago

  @CharlieEriksen That is literally hilarious. I see it in the actions logs LOL. Cockblocked by your own scam nice find and nice write up in the GH issue. Astonishing that the entire payload has been sitting there in Github for 10 months. I ran your conclusions by Opus who validated them. Is there any way to actually deobfuscate what is in the javascript payload _runtime.bin?

• 
  barafostus dreame (@barrowfaustus) reported 12 minutes ago

  I like Rust but it is just maddening how sometimes doing basic stuff requires threading the needle between insane amounts of bureaucracy and intractable 9-year-old GitHub issues. This time it's trying to fill an uninitialised byte buffer. How hard can it be?

• 
  Velon (@velonxbt) reported 13 minutes ago

  M5Stack launched CardputerZero on Kickstarter on May 26, 2026. On the same week, GitHub Copilot switched to token-based billing effective June 1. Neither announcement explained what the other one meant. The Copilot change covers what happens when a fixed monthly cost becomes a variable. No seat price. No predictable total. Usage multiplied by tokens multiplied by model tier, invoiced at month end. The developer who used to put one number in a budget now monitors a formula. The CFO who approved last quarter's AI spend does not know what this quarter's number will be until the invoice arrives. CardputerZero covers the other direction. Powered by a Raspberry Pi Compute Module 0 - quad-core Cortex- A53, 512MB of memory, Wi-Fi, Bluetooth, Ethernet. A 1.9-inch display. A 46-key keyboard. An 8MP camera. A microphone and speaker for voice interaction. HDMI out. GPIO header. All of it in a device the size of a credit card. One price. It does not change based on how many commands you run. And the creator who filmed the Kickstarter dashboard described what they saw in one line: "less than a day, 8000+ orders placed." They did not call it a product launch. They called it crazy. And here is what GitHub Copilot's token billing actually knows: It knows the flat subscription was a risk transfer. The vendor absorbed the cost of heavy usage so the customer could plan. Token billing reverses that. The heavy user who was subsidized by the light user under the old model now pays their actual cost. The predictability that made enterprise AI budgets possible disappears. Every team that shipped more last month pays more this month. Every CFO who approved a fixed AI line item now needs a new forecasting model. The invoice is the first moment the real cost of usage-based pricing becomes visible. And here is what the 8000 backers who ordered CardputerZero in one day actually know: They know the campaign raised over $1.4 million against a $10,016 goal with over 10,900 backers and 32 days still left. They know the device runs Linux, runs local AI, runs voice interaction - without an API call, without a monthly token bill, without a variable invoice at the end of the month. They put $89 into a Kickstarter campaign before the product ships because they recognized the fixed price as the feature. Not the keyboard. Not the camera. The price that does not change when they use it more. → GitHub Copilot: token billing effective June 1 - usage × tokens × model tier, no predictable monthly total → CardputerZero: $89, Raspberry Pi CM0, pocket Linux lab - one price, local AI, voice interaction, no cloud dependency → Kickstarter result: $1.4M raised, 10,900+ backers, $10,016 goal - 140x overfunded in weeks → 8000 orders in less than one day - the first day of the campaign → CFO problem: Q1 AI budget approval does not cover Q2 actual Copilot spend → Backer logic: fixed price for local compute is the hedge against variable cloud billing Traditional enterprise software made a distinction between the tool that runs in the cloud and the device that sits on the desk. The cloud tool was powerful and unpredictable. The desk device was limited and fixed. That distinction held as long as the cloud tool was significantly more capable than anything a $89 device could run. GitHub Copilot made the cloud tool unpredictable in cost. CardputerZero made the desk device capable enough to run Linux and local AI. What 8000 orders in one day shows is not a hobbyist trend. It is a market signal. And when the June Copilot invoice arrives with a number different from last month - and it will - the developer who already backed CardputerZero will not be surprised. They placed that bet two weeks ago.

• 
  Firas D (@firasd) reported 16 minutes ago

  @JustJake Well the point these guys are skipping over is that they have a thousand github issues right They aren't prompting the agent directly with hey lets look at XYZ cause the github ticket is coming into the context so that's the prompt

• 
  Joey Chilson (@joeychilson) reported 17 minutes ago

  Here's an example: You set up a workflow in GitHub that when an issue is created an agent is automatically spawned to replicate the problem if it can replicate the problem it passes it off to other agents to fix the problem and they submit a PR with the fix.

• 
  Andrei Onel (@AndreiOnel) reported 17 minutes ago

  I don't have enough details to open a github issue though 🫤

• 
  newt0crypt0 (@newt0crypt0) reported 23 minutes ago

  Looks like a glitch GitHub graph

• 
  Karthik Ramasamy (@_cartick) reported 25 minutes ago

  @thsottiaux Can you guys please fix sandboxing issue that affects every session "No. I used the host binary path /***/bin/gh, but I ran it inside the sandboxed command environment. That was the wrong execution mode for your preference. I should have rerun gh with require_escalated outside the sandbox before falling back to browser/GitHub connector paths." Can codex come up with good defaults on what commands that always need to be executed outside of sandbox?