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:

• 
  wispem-wantex (@wispem_wantex) reported 6 seconds ago

  I think a reasonable compromise would be to henceforth hold Anthropic responsible for any security breaches or service outages. Every time Github goes down, Anthropic should be fined

• 
  Threadripper (@threadripper845) reported 3 minutes ago

  Nobody: Me: I'll gladly accept this high-responsibility open source maintainer role for zero compensation. Now I spend my weekends answering angry GitHub issues from developers who don't know how to read the README file.

• 
  Vishal Tiwari (@VishalTiwa91817) reported 4 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.

• 
  Ben Vinegar (@bentlegen) reported 6 minutes ago

  💡 I have an idea for an experiment We need a website for SoAC ... so we get an agent to do it, on a loop, set in motion once with zero human intervention after "go". It works off a semi-public GitHub repo, w/ issues, PRs, maybe even public agent traces. A publicly auditable experiment on whether it produces dogshit or not. Yea, nea?

• 
  Mathieu A. (@zoontek) reported 9 minutes ago

  What are the most annoying bugs you still encounter with React Native? 👀 Please share GitHub issue links 👇

• 
  Elvis Irhaye (@viii_fn) reported 15 minutes ago

  Is GitHub down or it’s just MTN trying to ruin my career?

• 
  DFIR Radar (@DFIR_Radar) reported 17 minutes ago

  AutoJack: a three-flaw chain in AutoGen Studio's MCP WebSocket lets a malicious webpage rendered by a local browsing agent spawn arbitrary processes on the developer's host with no user interaction beyond visiting a URL. Key findings: - Three weaknesses chain together: Origin allowlist bypassed because the agent's headless browser is localhost (CWE-1385), auth middleware explicitly skipping /api/mcp/* with no handler picking up the check (CWE-306), and server_params decoded from the URL passed verbatim to stdio_client as a command line (CWE-78), accepting calc.exe, powershell.exe, or bash as valid "MCP servers" - Attack flow: attacker page serves JavaScript that opens ws://localhost:8081/api/mcp/ws/?server_params= with a base64 payload, agent's MultimodalWebSurfer renders it, AutoGen Studio spawns the command under the developer's account, no token required regardless of auth mode configured - Affected code never shipped in a PyPI release; exposure limited to developers who built from the main GitHub branch before hardening commit b047730, which adds server-side parameter binding via a POST/UUID flow and removes /api/mcp from the auth skip list - Broader pattern: any agent that browses untrusted content and shares a host with a privileged local control plane dissolves the loopback trust boundary, this is not specific to AutoGen. #DFIR_Radar

• 
  Almog Gavra (@almoggavra) reported 19 minutes ago

  A few other meaningless metrics to optimize for: - I've authored 22% of the RFCs - *** blame marks me responsible for 14% of the LOC (.rs files only) - I've opened 11% of the issues on GitHub - I've generated the most memes on our discord (allegedly)

• 
  Yarchi (@undefinedKi) reported 20 minutes ago

  BORIS CHERNY, THE CREATOR OF CLAUDE CODE, JUST SOLVED AI'S BIGGEST PROBLEM. HE STOPPED PROMPTING CLAUDE AND STARTED WRITING LOOPS THAT RUN IT 24/7 The guy who built Claude Code doesn't prompt Claude anymore. He writes loops, and the loops do the prompting. It's called loop engineering. Here's what it is and how to set it up. A loop is a system that wakes itself up, finds work, does it, checks it, and repeats, while you watch instead of type. In Claude Code it's three built-in commands: > /loop runs a prompt on an interval. Example: /loop every 5 minutes, check for new GitHub issues and handle any that come in. > /goal makes the agent work until a condition you set is true, with a separate model grading the result. Example: /goal build this feature until all tests pass. > /routines are scheduled jobs. Example: every hour, wake up, read the spec doc, and do the next task. The fastest way to start: write a simple task list in a plan.md file, then tell Claude "use the loop skill and work through plan.md one task at a time." It sets up the /loop itself, does the first task, validates it, wakes itself for the next, and reports back when the list is done. You never write the loop prompt by hand. Three rules so it doesn't burn your budget or ship garbage. One, split work across separate sessions instead of looping in one (a long /loop bloats your context and overwhelms the model). Two, use a cheap model like Haiku for planning and a strong one only for the actual code. Three, keep a human checkpoint on anything that ships, never let it run all night unchecked. Bookmark this

• 
  aisama.code (@aisama_code) reported 25 minutes ago

  AI Research gets stronger when it records contradictions *most research workflows collect supporting evidence - that is the weak version for serious research I want a contradiction log: - claim - source - date - who says it - what evidence supports it - what evidence conflicts with it - what is still unknown - confidence - next check example: > claim: this product has strong developer adoption > support: GitHub activity, docs updates, X discussion, integrations > conflict: low issue activity, small Discord, few production case studies, mostly founder-driven content now the memo is different, It says: "visible attention, but adoption evidence is still weak" the useful workflow: research question -> source list -> claim extraction -> contradiction log -> memo ! сode is good at assembling text ! AI is good at comparing disparate text ! human is good at determining which contradictions are significant *without a contradiction log, AI research becomes a confident summary of whatever it found first

• 
  Cristian Trifan (@CristianTrifan) reported 26 minutes ago

  This took 4 hours to complete and burnt almost all 5 hours tokens – I was left with 2%. I had almost 30 sub-agents created for independent code review and a lot of Claude sessions ran for adversarial code review. I still had to review every PR and added minimal guidance to Codex from time to time. Codex said my intervention was low to moderate, but high leverage. — Some insights from Codex: The run showed that this workflow can work, but only if the coordinator treats GitHub as the source of truth. The most useful pattern was: issue -> PR -> current head SHA -> checks -> reviewThreads -> merge/issue closure. When I followed that, things stayed grounded. When state moved underneath me, like #335 being force-updated externally or merged while Claude was running, the only safe response was to refresh GitHub state immediately. The “don’t rebase after merges” correction was probably the highest-value intervention. Without it, an agent will naturally try to keep branches clean, but with many open PRs that creates a CI storm. For this repo, “behind” should often be reported, not fixed. The other strong lesson is that reviewThreads matter more than flat PR comments.

• 
  𝕊ℍ𝕀ℕ𝔸☃ (@Shinawritesbugs) reported 27 minutes ago

  @viii_fn Github was slow too

• 
  Oluwatobi O (@ooluwatobig) reported 28 minutes ago

  More trouble for GitHub as Cursor has launched Origin, a product which is essentially GitHub for AI agents

• 
  Ucupaint 🔶 (@ucupaint) reported 30 minutes ago

  @iye_jr It works fine here. Check if the paint mask is turned on or not. If you still have a problem, please file a github issue with a sample file.

• 
  Andy Wheeler (@CrimeDecoder) reported 31 minutes ago

  For academics, this is entirely open source by its nature. If you right click on the page and view the source, you can see exactly how everything is created. (Hence a downside of WASM, there is no way to hide it if you wanted it to be locked down, like in a paid app.) It can also be deployed on a free static site. So you could deploy it via GitHub pages for free if you wanted to. You don't need to worry about a server at all in this setup. This could easily scale to databases with 1 million plus rows, and works just fine on a cell phone.