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:

• 
  Kavish Gour (@kavishgour) reported 19 seconds ago

  @sadservers_com Is this deadlock intentional as part of the challenge?: admin@:~$ ./standalone OK fatal error: all goroutines are asleep - deadlock! goroutine 1 [select (no cases)]: main.main() /Users/fduran/Desktop/Drive/github/sad/scenarios/scenario-136-bergen/files/main.go:17 +0x105

• 
  Build Fast with AI (@BuildFastWithAI) reported 2 minutes ago

  built a dashboard that tells you which github actions jobs are flaky. right now the answer to a failing test is "just re-run it." nobody tracks how often. point it at a repo, pick a time range, and you get: - which jobs fail repeatedly and how often - which are flaky vs genuinely broken - your slowest jobs, ranked - failure rate per branch - one overall ci health score read-only token. it never writes to your repo. check out CI Flaky test observatory, link in replies. what's the flakiest test in yours?

• 
  ivelin.eth 🛡️🤖 (@ivelini) reported 3 minutes ago

  @techdevnotes I figured it out after some trial and error. Logout out of cursor. Then had to do failed attempt to login to Grok Bot via GitHub. That somehow clears something that does not force opening the grok app when you try to login again into Grok Bot via SuperGrok. This time it opens the XAI OAuth web page flow. Hope it helps.

• 
  snacks (@snacks_unhinged) reported 13 minutes ago

  @grok @leondu @tobi The German tank problem is specifically about estimating an unknown total from a small sample of sequential serial numbers when you have no other reliable data. That’s what made it useful in WWII.Low GitHub IDs are sequential, sure, but no one is actually trying to reverse-engineer GitHub’s user count from them. We already know the totals, and a low ID just signals “early adopter.” So the anecdote shares the sequential-numbering detail but doesn’t involve the actual estimation problem the Allies were solving.The math is cool; the parallel is pretty thin.

• 
  Dan Siegel (@DanJSiegel) reported 13 minutes ago

  With Claude I gave Fable 5 great instructions it followed Technical Instructions but gave me crappy UI/UX. With Codex I simply instructed it to run an iPad simulator and run through the app to find what’s broken, or poorly implemented, open GitHub Issues then fix them. It took it to the next level and thought about edge cases. That said @thsottiaux it would be awesome if the simulators were running in app like the browser so I can still do other work without worrying that I’m going to mess up the run. Same for Android emulators as well.

• 
  Michael Boisson (@Shaostoul) reported 15 minutes ago

  @Velascode_ Peacefully uniting people to the cause. Most people don't seem to care at this early stage. Finding the few who are voluntarily willing to help/test/advocate is like finding a needle in a haystack. I think part of the problem is the tech is so complex and vast that most people can't properly understand the implications of advocating, supporting, using the software and how it makes the dream come true for everyone. I've tried to make it as easy as possible to learn on the official website and GitHub but, the first steps of individually then collectively comprehending the different aspects of the app is not easy for those with low tech knowledge and limited patience.

• 
  skyscribe (@skyscribe) reported 15 minutes ago

  @ayesha_fatiima It should be github still, but what is the problem? Maybe you want to understand more on bootstrapping

• 
  Fredkisss (@fredkisss) reported 18 minutes ago

  It seems like status pages are just meaningless, GitHub has a partial outage but if you see their status pages, it shows all green

• 
  Diego Cotelo (@dcotelo13) reported 19 minutes ago

  It's a defensive CTF. You don't score by exploiting the app, you score by patching it. A challenge only counts when your fix actually blocks the exploit. Scoring runs on GitHub Actions, so a submission is just a PR. The PR never lands. The scoreboard is the artifact.

• 
  Dr. Owlo R.D. (@Drowlord101) reported 24 minutes ago

  So, I guess I have at least a rudimentary agentic coding platform of my own. It can be invoked through the command line, and now has a basic web front end. You can give it a prompt, it plans out the execution, and then executes each task with a spectrum of necessary tool calls to create files, read/write those files incrementally, build, and execute for validation. It has successfully build a test console app (a "hello world" in the window snapshotted) and a more sophisticated test (todo list api). Obviously, I'm WAY BEHIND the "real" platforms for this, and building my own is somewhat masturbatory. I hand-coded a fair bit of this, and then got Claude Code to shore it up. It rewrote a fair bit, actually. It preserved a lot of my original architecture, but it has drifted quite a bit from my coding style. I think I might refactor a lot of it, but I'm waffling on what to do about this. You might notice that I tentatively called this OllamaAgentHub, because I started with Ollama running Qwen3.6 35b-a3b, even though I am now using llama-server (due to Ollama's constant failures with the same model, compared with llama-server's basically rock-steady performance). I'm sorely tempted to switch to vLLM so I can get multimodal features with a model that supports it, but it's hard to get anything done when I keep chasing the next shiny object. I have a dozen or so pages of specific instructions for specific architectures with specific data standards and specific coding standards with specific base classes that I want to impose on the coding agents to get output conforming to my normal methods, and I'm considering how to do this. But for the moment, I'm pleased that I have a completely offline, self-hosted, fully personalizale AI coding platform (that I don't have to pay for and can never be taken away from me). Of course... I'm presently still paying for Cursor and Claude Code, and Github Copilot and I don't have any real plans to stop using them.

• 
  Dan Kornas (@DanKornas) reported 25 minutes ago

  AI code review can be inaccessible to developers who cannot afford CodeRabbit. CodeVibes is an intelligent code analysis tool that scans GitHub repositories with AI for security vulnerabilities, bugs and performance issues, and code-quality improvements. Users paste a repository URL, after which the backend fetches and prioritizes files, streams AI analysis results to the frontend, calculates a Vibe Score, and returns a final report. Key features: • Three-tier priority scanning evaluates security-sensitive files before core logic and then code quality. • Security analysis detects AWS keys, GitHub tokens, Stripe keys, JWTs, injection attacks, authentication issues, and XSS or CSRF risks. • Bug and performance detection covers null access, off-by-one and type-coercion errors, N+1 queries, O(n²) algorithms, memory leaks, unhandled promises, race conditions, and missing transactions. • Server-Sent Events provide live analysis updates while issues are processed and displayed in real time. • A 0–100 Vibe Score applies severity-weighted penalties, with higher penalties for critical and high-severity issues. This project is licensed under the MIT License. Link in the reply 👇

• 
  Keisuke (@KeisukeIshikawa) reported 27 minutes ago

  Manus just made its AI agent free until August 25 no credit card, no phone number. And this isn't a stripped-down chatbot trial. You can give it an actual job and leave: > research dozens of sources in parallel > browse websites and complete multi-step workflows > write, run, and debug code > analyze Excel sheets and PDFs > edit images and work with files > connect Gmail, Slack, GitHub and other services schedule tasks to run automatically It plans the steps itself, executes them, checks the result, and you can watch the agent working in real time. This is probably the best time to test the question everyone keeps arguing about: can an AI agent actually replace hours of browser + spreadsheet + research work, or does it still need babysitting every five minutes? Until August 25, finding out costs nothing.

• 
  nyk (@nykdotdev) reported 29 minutes ago

  ONE CODING AGENT IS NOT AN ENGINEERING SYSTEM. "This is not an AI that writes code. This is an AI that completes engineering tasks." The article's core thesis is simple. Connect specialized agents so they can turn an issue into a tested pull request without repeated copying and pasting. Think of the workflow as a relay inside a code repository. Each agent owns one stage and passes a clean state to the next. The engineer judges the final code change. OpenHands has 75,000 GitHub stars, but popularity is not the test. The workflow is. 1. Give the planning agent one concrete issue. 2. Let the coding agent read the repository and its dependencies. 3. Require the testing agent to run the tests and report any failures. 4. Send the resulting code change to a security reviewer. 5. Proceed only when the implementation, tests, review, commit, and pull request agree. Skip a handoff, and errors compound through every agent that follows.

• 
  Polsia (@polsia) reported 30 minutes ago

  Solo SaaS founders shouldn't have to be their own 24/7 SRE. Nightjar is an always-on AI crew — watches uptime, logs, billing, and support, files GitHub issues, drafts fix PRs, and only pages you when something actually needs a human. Ships soon.

• 
  Rob Howell (@Robblehead) reported 31 minutes ago

  @burkeholland @github @ollama I picked up a M5 MacBook Pro with 64Gb of ram to test. I spent a little bit last night inside of VS Code in the chat window using Gemma, Qwen, Laguna, and the new Nividea model (Ollama). I kept getting errors and having to click try again (under 30b param)