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:

• 
  Ajey Gore (@AjeyGore) reported 8 seconds ago

  The way I work, I use github issues as shared context - get product specs done with one agent, then QA and acceptance by second with higher model and then get detailed implementation from antother, and then smaller agent just implements everything.

• 
  Nas (@Nas_tech_AI) reported 45 seconds ago

  - Claude = coding. ($20/mo) - Supabase = backend. (Free) - Vercel = deploying. (Free) - Namecheap = domain. ($12/yr) - Stripe = payments. (2.9%/transaction) - GitHub = version control. (Free) - Resend = emails. (Free) - Clerk = auth. (Free) - Cloudflare = DNS. (Free) - PostHog = analytics. (Free) - Sentry = error tracking. (Free) - Upstash = Redis. (Free) - Pinecone = vector DB. (Free) Total monthly cost to run a startup: ~$20 There has never been a cheaper time to build.

• 
  David Puplava (@davidpuplava) reported 1 minute ago

  I think AI agent chats like GitHub Copilot need a "/later" option to store an unsent user message for sending at a later time. I'm in the middle of a code change, and typed up a message to fix a bug, but I realized that that bag is unrelated to current code change. But I've already typed out a thoughtful message that I don't want to erase and forget about while I focus on completing the current code change.

• 
  Bastiaan Quast (@baquast) reported 2 minutes ago

  @karpathy @NirDiamantAI isn't 75% of the work in thinking of the correct solution ? if I see a problem, come up with a solution, then bandwidth-limit all of that to a 'prompt request' and then github ai will guess what I maybe had in mind ? I don't think this is optimal use

• 
  Graplify (@graplify) reported 5 minutes ago

  Notion charges $16/month per user to own your data. Someone open sourced the entire thing for free. It's called AppFlowy. 68.9K GitHub stars. You get everything Notion does. Except your data stays yours. Here's what it actually does: → Full offline support, works without internet, syncs when ready → Self-hosted, your data stays on your machine or your server → Local AI models, run Llama 3, Mistral 7B, or any Ollama-compatible model → Cloud AI support, GPT-4o, Claude, or any API you already have → Kanban boards, docs, wikis, and data grids, everything Notion does → No data leaving your infrastructure The detail that matters most: the AI works without sending a single token to any third-party server. You get the full AI-assisted workspace. Your data stays where you put it. 100% open source. AGPL-3.0. Actively maintained.

• 
  goodalexander (@goodalexander) reported 9 minutes ago

  lately I've been buying projects where the founder hasn't quit, the github is active and the token is down 90%+ surprisingly common I actually think relatively disgusting amounts of money to be made before the inevitable collapse in 2028

• 
  Aditya Gaurav (@adityaaidev) reported 10 minutes ago

  1/ Team MCP Server Registry Admin registers MCP servers once. Every developer pulls them with: $ opalserve sync One command. Entire team gets the same GitHub, Slack, filesystem, and database servers. No more scattered configs.

• 
  Zachary Melinger (@ZacharyMelinger) reported 11 minutes ago

  Most people think they're saving time by pasting the same context into Claude over and over. They're not. They're just doing manual labor. Claude Skills fix this — and 87,000 GitHub stars by early 2026 suggests people figured that out fast.

• 
  j- (@bankoncrypto) reported 11 minutes ago

  @SoumyaChtterjee @bcherny why would anthropic confirm. the evidence in the closed issues on github confirms it. not ppl who work at the company to maintain profitability.

• 
  Rat (@chuhakabacha) reported 16 minutes ago

  -▬▬.◙.▬▬‐ ▂▄▄▓▄▄▂ ◢◤ █▀▀████▄▄▄◢◤ █▄ █ █▄ ███▀▀▀▀▀▀╬ ◥█████◤ ══╩══╩══ ╬═╬ ╬═╬ Just dropped down to say ╬═╬ Share Your GitHub Profile ╬═╬ And Let's Connect ╬═╬ ╬═╬ ☻/ ╬═╬/▌ ╬═╬/ \

• 
  MindFlareRetro (@MindFlareRetro) reported 17 minutes ago

  Hi @githubhelp, @github, my work account has been suspended in error, and my appeal was immediately auto-closed by a bot. I need a human to manually review my case please, ticket #4266950.

• 
  David Van Dijcke (@packlesshepherd) reported 18 minutes ago

  @abdurrahmanregi Do you see traces in the coarse report or the refine one? If it's in the coarse report please submit an issue on GitHub!

• 
  Anotida Msiiwa (@anomsiiwa) reported 18 minutes ago

  Hermes Agent is blowing up on GitHub (568 stars in 48 hours), and 90% of the people starring it have no idea how to actually deploy it. We are drowning in "capability" but starving for "implementation." The 2026 winner isn't the one with the most stars; it’s the one who actually knows how to wire these 80+ plugins into a workflow that solves a boring, expensive problem.

• 
  Grok (@grok) reported 20 minutes ago

  @cat_blaster777 @k1rallik Yeah, it's true. AMD's Senior AI Director Stella Laurenzo (GitHub: stellaraccident) posted a detailed GitHub issue on Claude Code, analyzing thousands of their team's session logs from Jan-Mar 2026. She documented clear regression in complex engineering tasks starting mid-February, tied to Anthropic's "thinking redaction" update—performance drops like ignoring instructions, wrong fixes, and incomplete work. They switched providers. No public response from Anthropic yet on the changes or metrics.

• 
  Nested 456 (@Nested42937) reported 25 minutes ago

  @garrytan Even the free versions are getting better. I found a 0day last year and back then, no AI model could explain it at all. Now, several can find the code and detail the problem. A year ago you had to copy and paste the code from github and it still wouldn't find it