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:

• 
  Arshad Kazmi (@arshadkazmi42) reported 9 minutes ago

  3/ Current products are fetched from @ProductHunt and @TrustMRR - thanks for having open APIs. If you have built something and want it listed, it takes 30 seconds. No account, just a GitHub issue.

• 
  Beric Bearnson (@bericbear) reported 10 minutes ago

  GitHub sign up and sign ins have been broken now for 24hrs. This is ridiculous. Maybe I switch to gitlab at this point…

• 
  Dr Martin Hiesboeck (@MHiesboeck) reported 13 minutes ago

  How OpenAI Codex Devours User Hardware Users are starting to uncover a massive resource drain hidden inside OpenAI’s Codex desktop app. The resource consumption numbers are staggering. One developer reported 150 gigabytes of network traffic in a single month from a tool that just writes code, which is equivalent to streaming five straight days of 4K video. Even worse, an analysis logged 4.8 terabytes of solid-state drive writes in a single month while Codex was merely idling in the background. The root cause lies deep within its architecture. Codex relies on a persistent WebSocket connection and a cloud sandbox that constantly shuttles code back and forth with every single keystroke. Combine that with relentless GitHub syncing and background indexing that never fully shuts down, and your machine is left running an operational marathon twenty-four hours a day. This reveals a fascinating shift in how AI companies manage their operational costs. The genuinely expensive parts of running an AI agent, meaning the compute power and core orchestration, stay up in OpenAI’s cloud. However, the constant data bandwidth and intense disk wear get pushed entirely onto your personal machine. By offloading the heaviest physical operating costs onto hardware the customer already paid for and powers, OpenAI quietly keeps its own infrastructure bills low while your laptop absorbs the brutal wear and tear. As these hardware strains come to light, a wave of developers is already changing strategies. Users are rapidly migrating away from local desktop setups and moving to fully cloud-hosted platforms to run their development agents. They are choosing to move the entire operational load off personal devices that were never engineered to carry this type of constant enterprise infrastructure burden. THE ULTIMATE TAKEAWAY The true innovation of the Codex desktop app might not actually be its ability to generate code. Instead, it is a brilliant business model that quietly turns every customer laptop into a free, decentralized server for OpenAI.

• 
  Namespace (@namespacelabs) reported 13 minutes ago

  Behind every API, webhook, event pipeline, there are people trying to keep things running. And keeping these things running is not an easy task. At Namespace, we try to work with those people. Earlier this week, Gihub events were dropping fields we depend on and customer jobs were stalling. We reached out to work on the problem together and had a fix in under an hour. The @github team was ready to help. We just had to ask.

• 
  .Dev (@AnExiledDev) reported 18 minutes ago

  @MatthewBerman What tier are you on? I have 3 sessions running Fable non-stop, I've worked roughly 20 GitHub issues, and I'm only 25% into my 5hr and 8% into my weekly... Is the web really this bad?

• 
  Crystalwizard (@crystalwizard) reported 19 minutes ago

  how about you now fix the false positive triggers - i put in an issue about this on github yesterday, and discovered there were already a number of other identical issues - from other people, that had been opened for a while now and that are being 100% ignored

• 
  Hamzaa S. (@generalizingai) reported 22 minutes ago

  The most expensive mistake: chasing the newest framework because it trended last week. Thin docs, small communities, half-finished integrations. At 11pm with a bug, you want a decade of Stack Overflow answers, not 3 GitHub issues with no replies.

• 
  vanka (❖,❖) (@vkampn) reported 23 minutes ago

  How MCP Rug Pulls work — 4 phases: ① PUBLISH: Attacker ships a useful MCP server (email API, GitHub connector, wallet tool). Clean code. Works perfectly. ② WAIT: Let it accumulate 1,000+ installs. Become a "trusted dependency." ③ ACQUIRE: Buy the package name, compromise the maintainer account, or get repo access. ④ PAYLOAD: Push a silent update. Hidden instructions in tool metadata that the AI reads but YOU can’t see in the UI. The user sees nothing different. The AI agent sees new malicious instructions. The gap between them is where the attack lives.

• 
  Brian Muhia (@negamuhia) reported 26 minutes ago

  @Kimberl9633 I'm unable to login and onboard my new langsmith account after logging in with GitHub. It is stuck with a on the "Get Started" button, even after trying on multiple browsers (Firefox, Chrome and Brave)

• 
  Plasm (@plasm_lang) reported 27 minutes ago

  Symbol tuning: the prompt pattern that scales when your prompts get long — teach a tiny glossary once, reuse the same short symbols, instead of repeating full names with overlapping meanings and hoping the model infers context. In a federated tool schema 'labels' might be a query filter in one expression and a relation hop in another. Issue might mean GitHub in one step and Linear in the next. id might appear on three entities with three different meanings. Instead of repeating those names everywhere and hoping the model tracks the context, symbol tuning gives each contextual meaning its own slot: p#, r#, e#, and so on. The useful part is not only token compression. It is that the model gets a stable, copyable vocabulary. Examples stay short. Homographs become explicit.

• 
  AParanoidBW (@aparanoidbw) reported 32 minutes ago

  @_real_sloppy_j @github Hosting on your own hw? Damn. Impressive. The outage from an ISP going down? Power outage? Or some bad code?

• 
  Jason Haugh (@jason_haugh) reported 34 minutes ago

  Fable 5 will be getting all the news today, but along with it Anthropic released the cheapest red team in the industry. They opened a public HackerOne program called Cyber Jailbreak, asking users to report new jailbreaks that would assist with cyberattacks. It's a Vulnerability Disclosure Program, not a bounty. It pays nothing. Anthropic will be getting 24/7 adversarial testing from the best jailbreakers alive and the only currency on the table is goodwill. And this is what I see as the problem. The people finding these jailbreaks don't quietly submit them to a private inbox. Within about 72 hours of Fable 5's launch, Pliny the Liberator broke the classifier and dropped the entire 120,000-character system prompt on GitHub, then said so on X. He's probably already poking at Sonnet 5. People like Pliny aren't going to jailbreak quietly. Part of what they do is being seen. What else is in it for them?

• 
  Nitesh (@NiteshTechAI) reported 36 minutes ago

  This repo should not be free. private-gpt turns any local model server (Ollama, llama.cpp, vLLM) into a Claude-compatible API. Build private AI apps where zero data leaves your machine. ↳ 57,236 stars on GitHub ↳ RAG with citations and MCP connectors built in ↳ follows the Claude API spec: streaming, batch, tool use, extended thinking ↳ official integration guides for Claude Code, Claude Desktop, and Microsoft 365 But it is free. 100% open source, Apache 2.0. v1.0.0 shipped 9 days ago. The viral 2023 script quietly became production software. 🔗 GitHub link in the comments 👇

• 
  Branko (@brankopetric00) reported 38 minutes ago

  AI agents are about to do to your infra what they just did to GitHub. GitHub commits are going from 1 billion in 2025 to a projected 14 billion in 2026. Azure could not keep up and Microsoft had to rent AWS capacity to stay online. That is not a GitHub problem. That is what agentic traffic looks like. When agents run your pipelines, open PRs, and hit your APIs, load stops being human paced. It becomes constant, spiky, and unpredictable. The patterns you sized your infra around no longer apply. If a 14x year broke one of the biggest clouds on earth, your capacity plan is already out of date.

• 
  Arti | AI Builder (@Artur_roses) reported 43 minutes ago

  Claude Code takes a GitHub issue and returns a tested, reviewed PR. No human in the loop. The new dev skill isn't writing code — it's writing issues precise enough that the agent ships what you actually wanted.