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:

• 
  Gokul Rajaram (@gokulr) reported 1 minute ago

  EVIDENCE LOOP FOR PRODUCTSPEC A Product Spec should not stop at launch. The common failure mode with product docs is that they describe intent before the work, then disappear once the work starts. A PR ships, an eval runs, a dashboard moves, a customer complains. BUT the product doc stays frozen. Then 3 weeks later, nobody knows which acceptance criterion the PR satisfied, which eval run proved the model behavior, or which dashboard showed whether the product bet worked. To fix this, we just added evidence support to ProductSpec. The core idea is simple: ProductSpec defines intent. Evidence shows what happened. Decision Trace records what changed. Related Artifacts now let teams attach evidence directly to ProductSpec IDs: • AC-1 can link to the PR, test, release, or code that implemented it • EVAL-1 can link to the eval run or human review record that checked model behavior • SM-1 can link to the dashboard, analytics snapshot, or experiment that measured the post-launch outcome This matters more as agents write more code. An agent can claim it implemented something. A PR can look complete. A test suite can pass. But the useful question is: which piece of product intent did this evidence satisfy? That is where structured specs start to matter. If AC-2 says the user can export a dashboard with visible filters preserved, the implementation PR should point back to AC-2. If EVAL-1 checks whether an AI support triage model correctly identifies account-risk tickets, the eval run should point back to EVAL-1. If SM-1 measures median time to first human response, the dashboard or analytics snapshot should point back to SM-1. This turns a Product Spec from a planning document into a record of intent plus proof. A few important boundaries: ProductSpec does not run evals. ProductSpec does not collect production traces. ProductSpec does not replace Braintrust, Langfuse, Datadog, GitHub, Linear, or your analytics stack. ProductSpec gives all of those artifacts a stable place to attach. The latest validator now catches stale evidence links. If a Related Artifact points to AC-99 and no AC-99 exists, that is invalid. It also warns when the evidence type looks mismatched, like an eval run attached to a success metric instead of an eval. This is the direction I’m most excited about: Software intent that survives implementation. Evidence that connects back to intent. Decision traces that explain what changed when reality pushed back. Founders and builders: if your team is using AI agents to build software, start asking for evidence against the spec, not just code against the ticket.

• 
  Mudit Raj (@haxonit_) reported 4 minutes ago

  @Preeti_ly Claude is over hyped af I have used both the models for weeks, fable and gpt 5.6, I will always go for 5.6. Reason: Fable is **** at cyber. I just asked for fixing an GitHub issue related to cybersecurity, it totally denied to fullfill the request.

• 
  Gregor (@bygregorr) reported 10 minutes ago

  @github The fields aren't the bottleneck. Half my issues have no priority set even with labels and milestones already there.

• 
  Md Irfan Hasan Fahim (@mihf05) reported 11 minutes ago

  @GitHubCommunity @GithubProjects @github My account got restricted during Edu reverification for "missing 2FA", but I've ALWAYS had Google Authenticator enabled! It's a system glitch. Please check Ticket #4557612, my dev work is completely blocked. 🙏

• 
  zencoderai (@zencoderai) reported 12 minutes ago

  Zenflow reads straight from a Jira, Linear, or GitHub issue, pulls the relevant parts of your repo, and writes a plan before it touches code. The ticket is the prompt. Half the "prompt engineering" problem disappears when the agent reads the same ticket your team already wrote.

• 
  Rohit Kashyap | AI + Full-Stack (@rohit_jsfreaky) reported 12 minutes ago

  @Harry_The_Nerd sorting the articles sequentially in a github repo is the fix for x random ordering, smart

• 
  Rahul Verma (@RahulVerma989) reported 13 minutes ago

  Day 60 - Building Quillly in public 🚀 Two months. 60 straight days. Today's ship came straight out of my own frustration 👇 I've been shipping so fast that Quillly's view of my own site kept going stale between deploys - I'd push a fix, but the dashboard wouldn't notice until the next daily sync. So I built a deploy hook. one token-protected URL. drop it in your CI/CD, and the moment you deploy, Quillly re-fetches your sitemap and re-checks every tracked page. no waiting for the daily run. → curl one-liner or a GitHub Actions step → pass a delay so it waits for your build to actually go live → auto re-checks at +5 and +15 min if nothing changed yet built it for me. shipping it for you. which is pretty much the whole theme of 60 days. 🙌

• 
  Jay Carlson (@JayAtHomeOnX) reported 16 minutes ago

  @Elon -- maybe you should host a GitHub on your server farms...your next billion dollar...idea...

• 
  Atef Ataya (@atef_ataya) reported 17 minutes ago

  The scale is real. BlueRock Security analyzed 7,000+ MCP servers. 36.7% vulnerable to some form of SSRF. Their proof of concept: Microsoft's MarkItDown MCP server. 85,000+ GitHub stars. Real AWS access keys pulled from an EC2 instance.

• 
  Santiago (@svpino) reported 17 minutes ago

  This model can generate coherent 1+ hour videos across multiple scenes without skipping a beat. I read their paper so you don't have to. Here is how it works: First, this is an open-weight model. You can find the GitHub repo and link to the model weights below. Most video models generate their videos one frame at a time. They look at previous frames to decide what comes next. Unfortunately, if there's a mistake in one frame, every subsequent frame will compound the error. It gets ugly really quickly. @robbyant_brain's new world model, LingBot-World-Infinity, works differently. Here are the main highlights: • The model practices using its own mistakes. Since training only shows the model clean footage, it learns to generate videos that look perfect but never learns to recover when errors show up. Instead, this model runs on its own output for long stretches, letting it drift into the same flawed states it hits in real use. • Then, the model gets shown the way back. From each of those flawed states, a slower and more careful version of the model demonstrates what a clean continuation looks like. The fast model adjusts its frames to match. It practices the recovery over and over, always starting from its own mess. • The result is a model that steers itself back to center. Wherever it drifts, it knows the way back, because that recovery is exactly what it trained on. This is what lets it hold together for an hour instead of melting after a minute. Alongside the 14B primary model, the paper describes a 1.3B variant designed to run on a single consumer GPU. See links below:

• 
  Gold (@G_E_AGLE) reported 17 minutes ago

  @ShitpostRock While I know that githubs purpose is just to hold code, i understand that the average user hates github when the only solution to their very specific problem is on github and they dont know how to execute code

• 
  Zach Vorhies / Google Whistleblower (@Perpetualmaniac) reported 20 minutes ago

  @Rohansguliani @thdxr it is, you need to break it into investigation and then execution stages. The memory system needs to be a github issue. Just using github at all makes the agent look for history in github to see what the current state and context is

• 
  Jake Halloran (@jakehalloran1) reported 27 minutes ago

  i speak fluent claudese (aka claude took it upon himself to write replies to github review comments with my account when asked to fix the issues lol)

• 
  Manol T. (@manol_ai) reported 28 minutes ago

  I make ios apps and I don't have an IPhone Neither mac. This is how I do it: - made the app over @expo with claude code - set up my github actions (ci cd) to create android (adb) and ios builds (ipa) - they go directly to the internal testing track for android dev console and testfight in apple store connect - I rent a @MacinCloud for $30 to run ios emulator and make sure the app works for emulator - invited 2 friends to testflight - they report me bugs and I fix it I am stubborn enough not to buy an iPhone but spend money on ads. What are your app building challenges? #buildinpublic #mobileapps #testflight

• 
  Anthony (@anthohad) reported 28 minutes ago

  For almost seven years, I tried to get closer to software in the margins: evenings after work, weekends with courses and lectures, unfinished side projects, and the occasional GitHub streak that made me feel like I was getting somewhere. But for most of that time, I was still watching from the side. I became a product manager, learned how software gets built with other people, and loved the work. Still, there were days when I watched engineers ship and wished I could contribute more directly. Then AI made building software accessible in a way I still find hard to fully process. That can sound like terrible timing after years spent learning, but I have come to believe the opposite: those unfinished years became the foundation that lets me understand the tools, catch what is wrong, and push them further. Today, contributing to code is simply part of my week. I can take an idea, build the first version, test it, and bring something real back to the team. In a software startup, that changes what a small team can attempt. The feeling is exhilarating, but "everything feels possible" cuts both ways. Building is becoming more accessible, while judgment is becoming more scarce. If you have worked close to software while wishing you could build more of it yourself, this may feel familiar. I wrote this from the middle of that shift: what it changes, why technical understanding still matters, and why the most exciting part may not be how much faster we can build, but what we can build that was not possible before. Software can now understand intent, filter enormous amounts of information, and surface what matters to the person using it. Crypto is where that potential feels most obvious to me, and it is a big part of what we are exploring at @elitra_xyz. We are only beginning to discover what that unlocks.