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:

• 
  Dominik Koch (@dominikkoch) reported 2 seconds ago

  @chronark @jacobmparis @vercel Yea github was down that night

• 
  Dylan Mikus (@dbmikus) reported 2 minutes ago

  @dexhorthy I haven't manually made a GitHub PR or pushed a commit in months I prefer spending $0.20 to tell Fable to `*** commit 'fix code' && *** push --force-with-lease`

• 
  CData Software (@cdatasoftware) reported 4 minutes ago

  These aren't theoretical. The advisory documents real incidents: tool parameter injection via unsanitized inputs, cross-server exfiltration through coerced MCP clients, GitHub MCP granting blanket repo access, output poisoning across chained agents.

• 
  A. Taha (@Purple_Pear_577) reported 4 minutes ago

  The new FlawCue will be a pre-launch code review for AI-built web apps. Connect a GitHub repo, then choose: • Security • Structure • One critical flow You get prioritized issues, exact code evidence, and fix prompts for your coding agent.

• 
  Sahaj (@iamsahaj_xyz) reported 5 minutes ago

  @gauravmandall @github fyi you can submit a support ticket. there's a "can't sign in" button on the login page but it's not obvious

• 
  David Xiong (@david_y_xiong) reported 5 minutes ago

  (1/4) Reproduce Issue: first, one LLM call extracts {observed, expected} discrepancy pairs from the GitHub issue text, then a tool-calling subsession (bash/read/write, up to 70 turns) writes one failing test per discrepancy. The subsession gets different prompt nudges at fixed turn counts: first explore relevant files, then grep for sibling call-sites, then consider edge cases, and lastly run pytest and confirm every written test fails and produces a traceback.

• 
  Matias Kiviniemi (@PragmaticZone) reported 6 minutes ago

  @Appyg99 Bigger change/opportunity could be in legacy operations. Big chunk of IT budget is spent on "men passively watching servers" with limited ability to solve problems. The moment you can just hook an agent to cloud account, github and ticketing and have it handle everything...

• 
  TheRegularJoeCEO (@RegularJoe_Ceo) reported 7 minutes ago

  Today, the Broad Institute at Harvard and MIT posted the story in the comments and the GitHub repo about the alpha fold process, and I forked the repo and improved it. You can just do things. I really like GitHub because you can go find somebody who's publicly publishing a problem, and you can just fork their repo and fix it and send it back to them. It's a really fair system. It either works or it doesn't, and if it works, everybody can see it.

• 
  Moysei (@0xMoysei) reported 10 minutes ago

  8 small gold boxes on a desk just ran a 440GB frontier model with 400,000 tokens of context. A datacenter doing the same job with GPUs would bill $150,000. This desk cost $36,000. The video shows the whole heist. 8 NVIDIA DGX Sparks wired into one brain: 1 terabyte of combined memory, 160 ARM cores, a 400G switch split into 100G lanes feeding every node. The model is GLM-5.2, full weight, no cut-down version. It loads sharded across all 8 boxes and answers at 24-26 tokens per second, with prefill near 880 tokens per second holding all the way to 300K context. He ran entire workdays on it before showing anyone. The comparison that breaks brains: matching this memory with RTX Pro 6000s takes 10 cards at $15,000 each. Same model, same quant, 4x the price, and the prefill chokes on CPU offload anyway. The desk wins on cost and holds its own on speed. Nothing worked out of the box. He patched vLLM by hand, forced static IPs across 8 nodes, tuned jumbo frames, and debugged alongside Claude Fable until 400K context ran stable. Every script is free on GitHub. Then the ending: the cluster ships to a datacenter next week. The desk was just its first job. $150,000 of capability, assembled where the coffee mug goes. The gap between "home setup" and "research lab" is now 8 boxes and a switch.

• 
  Rokko (@0xRokko) reported 11 minutes ago

  Shunyu Yao - creator of ReAct:"A language model is not very good at self-evaluation yet." Read that again. The guy who invented the agent loop is telling you the model cannot grade its own work. Yet everyone is out here letting it mark its own homework and shipping the garbage it gives an A. Here is what actually happens when you stop trusting the model and build the check yourself. ReAct with a HANDFUL of examples hit 40% and beat an RL agent trained on 100,000 samples. On real GitHub issues, plain models solved 2%. Bolt a reason-act-observe loop on top and it jumped past 10%.Same model. 5x the result. The difference was never the prompt. It was the verifier. Your taste, written down strict enough that a machine enforces it, is the whole game now.

• 
  RyDawgE @ Halfapps (@RyDawgE_) reported 17 minutes ago

  @gregceltiano Training was done by already public repos.. and is now done with people who agree to the terms and services of said hosting platforms. There will be a github migration soon. Ive already begun moving my private repos to my own *** server. I have faith people are smarter than this.

• 
  Ching-Yuen Huang (@Michael_Huang_W) reported 17 minutes ago

  Unintentionally created a duplicate Pro account due to @cursor_ai's GitHub/Google login architecture, which auto-renewed for 6 months with 0% usage. Since I am already an active paying user on my primary research account, billing support flatly refused to even transfer store credits. Any chance @ArVID220u can help a dev out?

• 
  NikoAoi (@NikoAoi) reported 23 minutes ago

  Reported two issues in @runfusion pertaining to mDNS/hostname and npm install. mDNS issue seems to have been there for a long time but npm is new-ish to me, though I haven't been on the latest release for quite some time. Check the latest issues on @github for details, thanks.

• 
  David (@dayonefoundry) reported 24 minutes ago

  This got resolved. An automated system flagged and suspended my account and a human reversed it. The problem wasn't the code repo, it was the automation scripts in github functions. I was scrambling to move all the cron jobs to a VPS while sleep deprived after a 20 flight. All companies mess up so I can't stay mad at Github. But it's also my responsibility as a business owner to prepare for the worst. I setup a 2nd repo on Gitlab. It took 15 min and now all my pushes go update on both. I already setup the redundancies for the cron jobs. I setup all the appropriate CLIs so I can bypass githubs CI quickly if needed. This used to be overkill for a small project but in the age of AI, it takes less than 1 day to setup all these redundancies. It's worth doing.

• 
  Aki 🇩🇪 (@h0rang1_5arang) reported 24 minutes ago

  so i don't need github, microsoft onedrive or google drive anymore. i have it all set up, on a 13 usd/month server up in helsinki. i even have an agent taking care of maintenace. the worst thing that tinkerer can experience is finishing a project, and i have just done that.