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:

• 
  Digita (@digitaworld1) reported 4 minutes ago

  how well a model can fix real bugs in real open-source codebases. It is harder to game than older benchmarks because it uses actual GitHub issues, not synthetic problems. M3 scored 59.0% on SWE-Bench Pro, edging out GPT-5.5 at 58.6% and Google Gemini 3.1 Pro, while sitting just

• 
  Trifon Getsov (@trifon_getsov) reported 7 minutes ago

  @thdxr Top down works until the individual outgrows it. GitHub didn't win because companies adopted it first. It won because developers wouldn't go back once they'd used it.

• 
  Noonien Soong (@mlcarldev) reported 7 minutes ago

  Team @droid It's a bit unfortunate that something, likely in my local Droid installation, has stalled progress. This comes after 20 hours of brilliant, excellent planning and execution on the first 30% of this platform, where a stellar handoff procedure was created so I could start a new mission... which was the recommendation of the orchestrating agent in that first mission. Starting this second mission with a fresh context window, the agent again did a brilliant job planning the next milestones. It was extraordinary, detailed planning... but then it could not execute. After the planning and after me accepting the proposal, it refused to execute, throwing an error every time. The agent tried everything: 1. He decreased the size of the plan down to one line, so it is definitely not the content of the plan causing the issue. 2. He even deleted some mission and plan related json and other files to reset it while preserving all the information. I have restarted Droid and resumed the session, but it just doesn't work. I wrote a detailed, comprehensive bug report and filed it under issues in your GitHub repo, as this seems to be a real problem now. Issues #98 and #99 I hope that a next update will somehow reset my configuration. I didn't see a new version being installed that could have introduced a bug, so this must be something Droid does on such an extensive mission... perhaps when trying to start a new mission in the same repository, which is normal procedure according to the documentation. Something is off, and essentially I have been unable to continue the test since yesterday. I cannot continue having this platform coded here, while Opus Ultracode, on the other hand, has been delivering pretty functional stuff so far. It is a bit chaotic the way it works... it doesn't really stick to the plan... but it always comes back when reminded. I am pretty sure that today I will have a functioning platform delivered by Opus, though it will probably need some debugging and fine-tuning. It is unfortunate because I am confident GLM 5.2 could compete with Opus 4.8. The first stint showed this clearly; that first flawless 98% of the context window in the first mission was absolutely stellar. If I were to reinstall Droid from scratch, I assume I would lose all the artifacts that I have. The orchestrator: Key points to highlight when you pass it to Factory AI: 1. Root cause (smoking gun in the logs): the orchestrator session is bound to missionId 7ba4d425 via session tags, and this binding persists across CLI restarts. ProposeMission looks up that mission directory, finds nothing (because I deleted it trying to fix the issue), and crashes on H.length where H is the undefined result. 2. The bug is likely in session-tag lifecycle: the missionId tag is set at session creation time (before any ProposeMission call), so a failed proposal poisons the session permanently. The tag should be set AFTER a successful proposal, or cleared on restart if the referenced mission no longer exists. 3. The fix is almost certainly to start a completely fresh session (not --resume, and possibly in a new terminal window / after clearing ~/.factory/sessions/). I did not try this because you asked for the bug report first, but it is the most likely workaround on your side. 4. The AskUser tool is also broken in this session with a similar parse error, reinforcing that this is a session-state corruption issue, not a ProposeMission-specific bug. My comment: I meanwhiile tested. All the recommendations and the Ask User tool are now broken, even in completely unrelated new missions and new repositories. Planning also can't go to execution; it's always the same error. Droid seems to be broken for good now, at least on my computer.

• 
  Arti | AI Builder (@Artur_roses) reported 12 minutes ago

  Claude Code just took my GitHub issue, wrote the code, ran the tests, and opened a PR. My job: approve it. The dev workflow isn't changing. It already changed.

• 
  𝕊ℍ𝕀ℕ𝔸☃ (@Shinawritesbugs) reported 13 minutes ago

  @viii_fn Github was slow too

• 
  Tim Spalding 🇺🇦 (@librarythingtim) reported 16 minutes ago

  @justin_v_w This is a formal notice for you to shut down your wasteful, invasive and privacy-violating LibraryThing profile scraper and remove it from GitHub. Please reply to confirm that you have done so.

• 
  Arti | AI Builder (@Artur_roses) reported 17 minutes ago

  Claude Code takes a GitHub issue and returns a tested, reviewed PR. No human in the loop. The new dev skill isn't writing code — it's writing issues precise enough that the agent ships what you actually wanted.

• 
  Branko (@brankopetric00) reported 19 minutes ago

  AI agents are about to do to your infra what they just did to GitHub. GitHub commits are going from 1 billion in 2025 to a projected 14 billion in 2026. Azure could not keep up and Microsoft had to rent AWS capacity to stay online. That is not a GitHub problem. That is what agentic traffic looks like. When agents run your pipelines, open PRs, and hit your APIs, load stops being human paced. It becomes constant, spiky, and unpredictable. The patterns you sized your infra around no longer apply. If a 14x year broke one of the biggest clouds on earth, your capacity plan is already out of date.

• 
  pratik.eth (@eth_ethpratik) reported 20 minutes ago

  @Shahules786 @VibrantLabsAI Hello @Shahules786 , I am trying to report a security vulnerability over the email id provided over GitHub Security.md file but apparently its wasn’t delivered. Please share an alternative email or open the advisory for reporting the issue.

• 
  Arti | AI Builder (@Artur_roses) reported 25 minutes ago

  Claude Code just closed a GitHub issue, wrote the tests, passed CI, and opened a PR. No human touched the keyboard. This isn't AI autocomplete. The dev loop just got rewritten.

• 
  Vishal Tiwari (@VishalTiwa91817) reported 27 minutes ago

  @AlfieJCarter I am a Computer science student . I have given a brief introduction about MCP server in my college and explained them how to connect your GitHub repositories with MCP and your local system with MCP SERVER . I would love to connect you.

• 
  Iman (@RealKingiman) reported 28 minutes ago

  @ClaudeDevs Fix the auth bug with GitHub where I have it keep disconnecting and reconnecting GitHub every time

• 
  Yiqing Xu (@xuyiqing) reported 28 minutes ago

  @Faylosophe Certianly. Could you file an issue on the Github page?

• 
  𝑵𝒐𝒙𝒊𝒆 🥐 (@noxiepup) reported 29 minutes ago

  @softgaypaws @sillyandsunny no idea tbhhh, i found it like 2 years ago lurking thru github, so far it never gave me problems, at least none that i noticed

• 
  李新宝 (@lixinbao_X) reported 30 minutes ago

  Just watched KK's technique. Damn. Absolute game-changer. Install 7 skills in Codex. Writing, images, covers, PPTs. Full pipeline, done. The principle is dead simple. Break the workflow into 7 parts. One skill per part. Only do one thing. Step 1 Open GitHub, find a repo. Copy the link locally. Create a project folder to save it. Step 2 Write the skill description. Input three things. What it does. What the input is. Output and acceptance criteria. Step 3 Run it and find the bottlenecks. Where it stalls Create a new skill and break it down. Don't let one skill Do 7 things it's bad at. This works for writers, Xiaohongshu creators, WeChat pub runners, Video script writers. How many skills you got installed? Have you tried it yet?