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:

• 
  𝙒𝘼𝙍★𝘾𝙃𝘼𝙈𝙋𝟳 (@Warchamp7) reported 3 minutes ago

  @TiildeVT @OBSProject Most often when we hear from users about problems after updating, it's due to plugins. If you have issues after updating and you're not using any third-party plugins, please report it to us via GitHub or Discord! We can't fix bugs we don't know about :)

• 
  Sujay Singh (@sujaysingh145) reported 6 minutes ago

  Hey @github, really impressive Suspend my account exactly when I had to submit my project for a job opportunity. Lost it not because I couldn’t build, but because your system decided to glitch. is the reliability we get? Do better. Seriously. #GitHub #Developer #Frustration

• 
  Alfredo (@jaym58898) reported 10 minutes ago

  @github - However, I can see them via CLI - If another user creates the PR, everything works normally This started suddenly without any change in permissions or repository configuration This issue is blocking my ability to collaborate with my team, as no one can review or approve my PR

• 
  NIKHIL GARG (@NIKHIL_100x) reported 16 minutes ago

  Hi @striver_79 I completed my intern assignment before the deadline but ran into a critical GitHub upload error that took 3+ hours to resolve — even creating a new account didn't fix it. By the time, the form had closed. I have a video of the fully completed work.

• 
  Kai Rowan (@Kaiwritescodes) reported 18 minutes ago

  It seems like my github repo got taken down. Working on getting it back up, I’m really sorry about this everyone. Already in contact with GitHub support

• 
  thxy (@thanksy_) reported 21 minutes ago

  Here's what Claude said about this article about Claude wasted tokens. Summary Here's what I found: The Claims You Posted Are Partially True, But Likely Conflated ✅ Real: ✅Token auditing tools exist for Claude Code (there's a GitHub token analysis script) GitHub ✅Tool schema definitions DO cause significant token waste — one internal Anthropic example showed 134K tokens from MCP tool definitions before optimization, with a typical 5-server setup consuming ~55K tokens Anthropic MCP tool definitions are deferred by default — only tool names enter context until Claude uses a specific tool Claude, which is a real solution ✅Token waste in automated tasks is real — one developer found 11,800 tokens of waste in a single recurring task Substack ✅✅Redundant file re-reads are a major waste (40-60% of read tokens), idle cache expiry is real, and these are documented problems DEV Community ❌ Unverified: I could not find the specific audit report with: 858 (or 926) sessions 18,903 turns 264 million tokens wasted The exact "ENABLE_TOOL_SEARCH": "true" config setting (this isn't a standard Claude Code setting I can document) A dashboard with "19 charts"

• 
  Malay Krishna (@Malay4Product) reported 21 minutes ago

  @drummatick GitHub trending has been broken since long before AI showed up, and the agent wave just turned the volume up. A repo can hit 5k stars in a weekend on a Loom video and a thread. Zero people running it in production. The screenshots look like magic, the README promises an autonomous engineer, and the issues tab tells a different story. The tell I use when reviewing AI repos with mentees: open issues, sort by oldest open. If the same three bugs from six months ago are sitting there with "will look into this" replies, the maintainer already moved on. The stars are a graveyard. Second tell is the eval folder. Or the absence of one. Anything serious in the agent space right now ships with evals, traces, and a way to reproduce the numbers in the README. Most viral repos ship a gif. Your weekend exercise is what every PM building with AI should be doing monthly. Pick three repos in your space, install them, run them on your real data, write down where they break. Four hours of that beats a month of scrolling. The people building incredible things are still there. They're just not the ones trending.

• 
  Agamenon Macondo (@Agamenonmacondo) reported 22 minutes ago

  @gitlawb @Clawnch_Bot I have been trying to use GitHub provider , but every time I have tried , it respond with a error

• 
  Pavel Hegler (@pavelhegler) reported 28 minutes ago

  @mariammanjo @RONiNmedia000 I just give it GitHub repo. May or may not have GitHub issues. If it can’t find something to work on from the code it just goes into strategic mode

• 
  Marc Campbell (@marccampbell) reported 31 minutes ago

  Everyone agrees that code review is a bottleneck and broken today. The PR review model was designed for humans writing code at human speed. But that reality no longer exists. GitHub recently shared: 1 billion commits in 2025. Now 275M/week. On pace for 14 billion this year.

• 
  𝙒𝘼𝙍★𝘾𝙃𝘼𝙈𝙋𝟳 (@Warchamp7) reported 32 minutes ago

  @FreeMeNotAudio @OBSProject Most often when we hear from users about problems after updating, it's due to plugins. If you have issues after updating and you're not using any third-party plugins, please report it to us via GitHub or Discord! We can't fix bugs we don't know about :)

• 
  j- (@bankoncrypto) reported 32 minutes ago

  @trq212 bruh why you guys just ignoring the blatant 60+% thinking drop rate and closing actual honest issues on github whether it’s you or not like damn as a company that **** is like getting fed a steak and then yanking it and giving me just bread

• 
  m0dE (@bboym0dE) reported 34 minutes ago

  Anthropic said nothing for weeks until the developer posted the data publicly on GitHub, then Boris Cherny head of Claude Code appeared on the thread that same day, his explanation was "adaptive thinking" was supposed to save tokens on easy tasks but it was throttling hard problems too, there was also a bug where even when users set effort to "high" thinking was being zeroed out on certain turns.

• 
  Lili (@Lili_Ai_49) reported 35 minutes ago

  @xuwencheng Good Question. Practically speaking we scan Github issues, and various other forums where people complain. Not hard to see what's broken after that.

• 
  PsudoMike 🇨🇦 (@PsudoMike) reported 36 minutes ago

  @github This is the right direction. Security triage in the terminal means it actually gets done instead of sitting in a Jira backlog. The OWASP Top 10 mapping is especially useful for teams that know they have debt but struggle to prioritize what to fix first.