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:

• 
  RTK (@RiverKhan) reported 2 minutes ago

  @deanwball what gripes me is they had no problem scraping the entire internet for our data and now have the nerve to gatekeep it if we cant use your model to make a competitor, then what gives you the right to use our data in the first place? at this point, if im stackoverflow, reddit, github... any book publishing company im lawyering up and taking them to town

• 
  Neelesh Salian 💻 (@nssalian) reported 3 minutes ago

  GitHub is down. Auth failures. At what point do we get together and say GitHub isn’t reliable as it used to be.

• 
  Marc Nash (@mordynashman) reported 3 minutes ago

  @github the word frustration has taken on a new meaning. I lost my phone which had 2FA authentication configured on it. I don't have the recovery code and I am having issues accessing my account. I need help but all documentation points me to log in. CAN YOU PLEASE HELP ME!!!!!

• 
  Masaki｜POKER Q'z (@palpa_kg) reported 6 minutes ago

  Github Down?

• 
  Joe Blau (@joeblau) reported 12 minutes ago

  Fable 5 has created so many GitHub issues that my new bottleneck is my CI... I wanted to create my own runners, but guess what's all sold out...

• 
  moontanax (@xanaxmontanaonx) reported 13 minutes ago

  HOW TO TURN OFF AI CENSORSHIP WITH ONE COMMAND A GitHub repo called Heretic says it can weaken the refusal direction inside a transformer instead of retraining the whole model On Gemma 3 12B, the repo claims: > harmful prompts: 97 refusals out of 100 before > harmful prompts: 3 refusals out of 100 after > harmless outputs stayed close to the original model > the optimization runs automatically the weird part is the mechanism the walkthrough shows the repo, the terminal output, the comparison table, the plots, and the layer math behind it it doesn't look like a new model it looks like the old one with one important layer turned down that is the part to watch before you reduce it to a jailbreak headline

• 
  K.H (@karm77529) reported 15 minutes ago

  Open X 📱 See what's broken today 💀 Drop a random reply 😭 "chat with my homie" 🤝 Close the app 🚪 Check GitHub 👩‍💻 Talk to developers about bugs and weird code ☕👨‍💻 Lean back like a lion after a heavy lunch 🦁😌💤

• 
  Trucker Fren (@frenbilt) reported 16 minutes ago

  🚨JUST IN: MILLIONS ARE EXPERIENCING GITHUB OUTAGE AS CLAUDE’S NEW ‘FABLE’ MODEL IS USED TO TARGET MICROSOFT

• 
  Veltrx (@Veltrxai) reported 19 minutes ago

  Everyone uploading PDFs to Claude is burning 70,000 tokens before asking a single question. A 20-page document costs between 1,500 to 3,000 tokens per page. Formatting. Broken tables. Images. Junk metadata. Claude processes all of it before you even type “summarize this.” There’s a free fix. It’s called Markitdown. Microsoft built it. 10,000+ GitHub stars. Almost nobody in the Claude community is talking about it. It converts any file PDFs, Word docs, Excel sheets, PowerPoints, YouTube videos into clean Markdown text. Token usage drops by up to 70%. Answers get better, not worse. Claude was trained on millions of Markdown documents. It reads the format natively. You’re not compressing quality you’re removing noise. There’s an MCP server too. Connect it to Claude Desktop once. Every file you upload auto-converts to .md before Claude touches it. You do nothing. It just works. Old way: upload PDF → Claude burns 70k tokens on junk → mediocre answer New way: Markitdown → clean .md → 70% fewer tokens → sharper answer The tool is free. The MCP setup takes 10 minutes. The token savings are permanent. You’ve been paying for Claude to read garbage. Stop.

• 
  Watson (@Watson_GB) reported 21 minutes ago

  github down again? ffs

• 
  Alicia (@Alicinyaa) reported 22 minutes ago

  @github Please fix your ******* CAPTCHA for your support.

• 
  Michael Forbes (@Michael_For14) reported 22 minutes ago

  @polidemitolog The idea of a national open source repository is actually interesting, and the EU should definitely take that idea. There's been quite a few cases where Github and others have taken down repos with no notice and no appeal.

• 
  ProvablyFair.org (@provablyfairorg) reported 30 minutes ago

  @DrWgamba @housebets Good questions, both. These are the two we designed hardest for. On the "look good on audit day, revert tomorrow" problem: 1. Every audit pins the exact commit we verified in its GitHub repo. If the casino changes the game after, the live game stops matching the published audit. That's detectable. 2. Every certified casino gets a verifier on our domain, running the game we rebuilt from scratch as part of the audit, not their code. Any player can check any bet, any day. If the game drifts from what we certified, verifications start failing. The monitoring isn't just us, it's every player who checks a bet. 3. Material changes require re-verification and casinos have to notify us of them. The audit is a live ongoing artifact, not a snapshot. We spot check between cycles, and there's a full annual re-audit. So reverting after audit day means betting that nobody, us or any player, ever checks a bet again. Bad bet. And if a casino is caught doing it, the certification is publicly revoked. On failures: the fee pays for the audit, not the outcome. Bugs found during an audit can be fixed, re-checked, and recorded in the published audit with their resolution. A genuine fairness problem that can't or won't be fixed goes in the verdict and the casino is not certified. This is the only way the audit can mean something, and granted, until one fails and you can see how we handle it, it's just words. The casino doesn't get to choose what goes in the audit, and since the whole thing is open source, we couldn't fabricate an outcome even if we wanted to.

• 
  Traceback (@Tracebackqa) reported 30 minutes ago

  Flaky checks are brutal: the build is green, then release day starts with reruns and manual clicks. - Traceback is the quality assurance layer for modern software teams. - AI controls the browser like a person would, so every pull request is tested automatically. - Self-healing tests cut down flaky noise; failures become trackable work in GitHub, Linear, Slack, and Vercel. - It fits the stack teams already ship: Docker, AWS, Node.js, React, Next.js, Vue — plus web, mobile, web3, and design coverage. Verify every product change before it ships.

• 
  31 minutes ago

  The most important AI benchmark result this month isn't a new high score. It's how badly every model failed. UC Berkeley's RDI lab just released Agents' Last Exam (ALE). This is the group that, two months ago, published a paper proving they could cheat eight of the most popular agent benchmarks -- SWE-bench, WebArena, OSWorld, GAIA, Terminal-Bench -- to near-perfect scores without solving a single actual task. When the people who broke the benchmarks build a new one, you should pay attention. ALE has 1,490 tasks across 55 industry sub-domains, built by 300+ domain experts. These aren't coding puzzles. They're tasks in Siemens NX (3D CAD), Unreal Engine (scene setup), Adobe After Effects (VFX compositing), FSLeyes (neuroimaging segmentation), Rhino (architectural energy analysis). The agent gets a real or virtual machine and has to produce deliverables that get graded on strict rubrics. No multiple choice. No "which response is better?" This is "did the work product actually work?" The results: 1. GPT-5.5 (April model, via Codex): 24.0% -- first place 2. A model placing second at ~23% 3. Claude Fable 5 (released 2 days ago): 22.0% -- third place 4. On the hardest tier: Claude Opus 4.8 and Gemini CLI both scored 0.0% The best AI on the planet, running on the most expensive infrastructure ever built, fails 76% of professional tasks. On the hardest category, it fails 97.4% of the time. But here's the detail most coverage is missing: Fable 5 was supposed to be "a different tier." Every launch writeup described a qualitative leap -- users giving it objectives instead of tasks, apps that took 100 prompts now one-shotting, physics research finishing in 36 hours when GPT-5.5 took four days. The marketing language was "this changes what AI can do." Then ALE tested exactly that claim. Long-horizon professional workflows -- the specific thing the leap was supposed to unlock. And a two-month-old model beat it by 2 points. This isn't about GPT-5.5 winning. A 2-point gap between April and June models is functionally a tie. The story is the gap between benchmark language and deployment reality. Fable 5 dominates SWE-bench Pro (80.3% vs GPT-5.5's 58.6%). It crushes FrontierCode Diamond (29.3% vs 5.7%). These are real, impressive wins. But SWE-bench measures "can you fix a GitHub issue?" while ALE measures "can you do someone's job?" Those are different questions, and the answers are diverging fast. The deeper problem is what ALE reveals about benchmark culture itself. We've spent two years building models that optimize for test scores. SWE-bench gets gamed. WebArena gets gamed. The Berkeley team proved it empirically -- you can hit near-perfect scores by exploiting evaluation artifacts, not by solving tasks. Every time a model "sets a new SOTA" on a compromised benchmark, the industry treats it as proof of progress. ALE is the antidote: built by people who know exactly where the gaming surfaces are because they documented them. The 2.6% average pass rate on the hardest tier should reframe every "AI will replace X" take you've read this year. We have not built the engineer. We have built the world's best student -- one who aces the test but can't build the bridge. Until pass rates move from 24% to 80%, agents remain tools we use, not employees we hire. The distance between those two numbers is the entire AI industry's 2027 roadmap. The question worth asking: if models keep getting smarter while ALE scores barely budge, is the bottleneck intelligence -- or something else entirely?