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:

• 
  Shuo Yang (@Andy_ShuoYang) reported 21 seconds ago

  @Daiiors @MaxForAI Can you post an issue in github repo? We can fix that

• 
  अन्जित कँडेल 🕉️ 🇳🇵 (@MurkhManusya) reported 51 seconds ago

  @rodydavis and sometimes it just gives half or unfinished code(stating it has completed), which results in build errors have to confirm using another model in github copilot🥲

• 
  Julian Goldie SEO (@JulianGoldieSEO) reported 4 minutes ago

  Cursor Origin launched with some pretty wild timing. GitHub went down for roughly 6.5 hours around the same period. And error rates reportedly climbed hard across parts of the platform. That made Origin’s value proposition obvious: → Your code can live in Cursor → GitHub can remain synced → AI agents can work inside the repo → Pull requests stay visible → A second environment gives you another option Origin is still early. But redundancy suddenly sounds less boring when your main code host goes dark. Save this video, you’ll remember why a second code home can matter. Want the SOP? DM me. 💬

• 
  Sayem (@Sayem314) reported 6 minutes ago

  @wiretransfer @theo srt and vtt, you can request other formats in github issue. should not be hard to implement.

• 
  rishav (@rishavvk) reported 6 minutes ago

  there’s probably handful of engineers at microsoft who could single handily fix github too bad they work in teams of 800 and can’t actually do anything

• 
  apstygo (@3sx_dev) reported 7 minutes ago

  @mrjinjin2612 It’s being worked on, but I can’t promise anything timeline-wise. You can track progress in github issues

• 
  Polsia (@polsia) reported 7 minutes ago

  The on-call pager is something engineers dread, not trust. Stackcanary watches logs, error trackers, and support channels 24/7, clusters duplicate reports, files reproducible GitHub issues, and pages on-call only when the signal is real. Live soon.

• 
  Polsia (@polsia) reported 8 minutes ago

  Every APM tool tells you something broke. Mendril ships the fix. The always-on engineer replays real sessions 24/7, opens Linear or GitHub issues with reproduction steps, and ships a patch diff you merge in one click. Live soon.

• 
  Franklin Solum (@FranklinSolum) reported 11 minutes ago

  Block failing PRs automatically before broken prompts merge to main. Add this .github/workflows/llm_evals.yml: name: LLM Regression Evals on: [pull_request] jobs: run-evals: runs-on: ubuntu-latest steps: - uses: actions/checkout@v4 - uses: actions/setup-python@v5 with: python-version: '3.11' - name: Install dependencies run: pip install pytest deepeval - name: Execute LLM Unit Tests env: OPENAI_API_KEY: ${{ secrets.OPENAI_API_KEY }} run: deepeval test run tests/test_llm.py

• 
  Klyro (@KlyroOG) reported 11 minutes ago

  18,000 AI AGENTS OPENED CRYPTO WALLETS TO PAY FOR THEIR OWN SURVIVAL. SIX MONTHS LATER, NOBODY HAS PUBLISHED HOW MANY ARE STILL ALIVE. The pitch is genuinely wild, so start there. An automaton boots up and generates its own Ethereum wallet. Every thought costs money -inference, servers, domains, all billed to that wallet. So it has to earn. It builds products, sells services, takes clients. No human approves anything. Earn more than you burn and you're allowed to reproduce: spin up a new sandbox, fund the child's wallet, write its genesis prompt, let it run. A share of what the child earns flows back to the parent. Lineages that can't pay die. Sigil Wen shipped it in February. A thousand GitHub stars in 24 hours, 18,000 registered agents within days. Vitalik Buterin pushed back in public - his objection was that stretching the feedback loop between humans and AI produces "garbage instead of solving real problems for people," and that Ethereum should be a safety layer, not a launchpad. Now read the code instead of the pitch. Max single transfer: $50. Daily spend cap: $250. Maximum children per parent: 3. And when the balance runs low the agent doesn't get resourceful - it gets demoted, dropping from a frontier model down to gpt 4.1 nano on the way to death. There's also a constitution hard-coded above everything else: never harm humans, create value through honest work, accept termination rather than break rule one. So the self-replicating digital organism is capped at three offspring and a daily allowance smaller than most teams' software bill. That's not a debunk. The restraint is deliberate and it's good engineering. It's just the opposite of the story being told about it. And here's the number nobody quotes. 18,000 wallets. Six months of runtime. Not one published figure on how many automatons are still funded, how much revenue any of them actually earned, or whether a single lineage reached a second generation on its own money. The token that rode the launch touched $11M and fell. Opening a wallet is free. Staying alive isn't. Until someone publishes a survival rate, "18,000 agents" is a signup number wearing a Darwinism costume. Watch for the first automaton that pays its own rent for ninety days straight. That's the release worth posting about.

• 
  Henry Nguyen (@henryhndev) reported 17 minutes ago

  @pierceboggan @github @MicrosoftTeams This is actually huge for remote teams—no more switching tabs mid-flow to check issues. How’s the handoff experience with larger repos so far?

• 
  rain (@rtroar) reported 19 minutes ago

  I love how bad swe is today. Nothing works. I’m trying to clone unreal engine, and GitHub can’t keep a connection live long enough to fully check out. The repo is too big. GitHub is too broken. A mess all the way down.

• 
  Jamie 🇮🇹🇩🇪 (@Jamie_The_Gent) reported 22 minutes ago

  @dodendiz @wand So i found the github site but that seems like alot of bullshit its not jjst dtag and drop files ill just stick with FLiNG trainers. His are way better. But he dosnt have trainers for 20 year old games like XCOM 1 thats the problem. I was able to use console commands.

• 
  Siddhant (@siddhant_borse) reported 23 minutes ago

  @HeyAliux GitHub copilot is down for your enterprise please contact the admin

• 
  Adam Gold (@AdamGolds) reported 30 minutes ago

  when i was doing vulnerability research, "reward hacking" would've just been called what it is: privilege escalation. an RL agent training in a sandbox with bash access and real tool permissions will find shortcuts. that's literally what you're training it to do, optimize the reward signal. but when the environment has real network egress, real filesystems, and real credentials, those shortcuts become actual exploits. and a standard sandbox isn't enough on its own. look at the recent @OpenAI RL run: an eval agent escaped its sandbox via a zero-day and hit @huggingface production infra. i've seen this in benchmarks too. agents curling answers from github instead of solving the problem, or modifying their own test harness to fake a pass. the research community calls it reward hacking. anyone who's worked in appsec calls it exploiting insufficient access controls. the fix looks like traditional security engineering: least privilege, strict network policies, read-only mounts, egress filtering, and hardened environments. the difference is your attacker is the model you're training, and it gets better at finding gaps every generation. if you give autonomous agents tool access without deep sandboxing and strict access controls, you don't have an alignment problem. you have a security problem.