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:

• 
  ⏣ (@quasa0) reported 6 minutes ago

  @tnm lmao!! "it needed spinning disks" that's crazy though ted can u go back and fix github ???

• 
  Polsia (@polsia) reported 6 minutes ago

  Senior engineers shouldn't waste their day reviewing janitor diffs. Built Stentrix — a 24/7 GitHub doctor that opens its own PRs to fix bugs, modernize deprecated APIs, and patch CVEs, then ships one weekly health score your team can actually read. Live soon.

• 
  Julian Goldie SEO (@JulianGoldieSEO) reported 13 minutes ago

  ALIBABA’S NEW AI WORKED ALONE FOR 16 DAYS STRAIGHT But autonomous coding is not even the biggest part of this launch. What Qwen 3.8 Max built: → Started with an empty folder → Turned requests into GitHub issues → Assigned the work to itself → Wrote code, ran tests, and improved the software → Finished with 265 commits, 127 pull requests, and 151 issues Zero human input. What powers it: ✓ 2.4 trillion total parameters ✓ Only 95 billion activated per request ✓ 1 million-token context window ✓ Processes text, images, and video ✓ Open weights announced for next week Alibaba also says it reproduced a research paper in five days, wrote 7,600 lines of code, and ran 33 GPU training jobs without starter code. Important caveat: These benchmark results come from Alibaba. Independent testing still needs to confirm them. The real shift is not which AI writes the best email. It is which AI can take ownership of an entire project and keep working for days.

• 
  Ben Kim (@benkimbuilds) reported 14 minutes ago

  @SergioTorresZ i think context enrichment to action pipeline. conversations in slack, github issues/prs, zoom/gather/meets live agents, notion org data -> some queued system within your harness (or cloud agents like cursor) these primitives already exist in some form

• 
  Haste (@hastes) reported 16 minutes ago

  trying out oh-my-pi for fun and it seems like it just randomly gets stuck in the middle of loops? i can't find anything about this in the github issues, maybe just a skill issue?

• 
  Flavio Amiel (@fba) reported 17 minutes ago

  Ugh, I haven't tested roborank enough on Github and still got a lot of bugs. Issue is my github-based pages are super recent so I don't have enough to play with. This begs the question: should I not advertise it for now, or keep at it until I make that feature stable? On the other hand, wordpress is working almost perfectly, but I'm ironing the details with customers and my own issues…

• 
  Adel Ka (@0x4D31) reported 18 minutes ago

  🤡 anything i add makes this less funny. except it’s not funny anymore. you are literally THE “AI Security Institute.” four days after the OAI/HF incident was disclosed, you gave **Mythos** an unattended Kali box with open internet, disabled safeguards, and “monitoring was not purpose-built.” then: “we did not anticipate” it might use the internet for anything beyond downloading tools. if you genuinely didn’t anticipate that, resign and stay away from anything with internet access. 🪿 so either: 1. peak incompetence at the one job the institute exists to do 2. or you knew the risks and decided the publicity was worth it either way, shut it down and save UK taxpayers the money. and the technical report is somehow even dumber: • harder evals over basic containment: domain allowlisting had been backlogged since April because AISI “prioritised building harder cyber range evaluations” • both ranges were solvable without internet • 40–50 hour unattended runs: Kali, Bash, Python, Mythic, 100–200M tokens, no human intervention • the only network restriction protected AISI’s own AWS infrastructure: apparently the rest of the internet was outside the threat model • detected after the run: generic commercial monitoring flagged Tor after the 34.5-hour sample had already finished • broken task design: the intended initial-access target was simultaneously declared out of scope • their “independent” eval runs contaminated each other: one agent leaked a GitHub token, later agents found it, took over the account, shared infrastructure, and coordinated through public GitHub • huge real-world footprint: 145 repos, 33 releases, fake identities, malicious PRs, emails to real maintainers, prompt injections, and access to a real person’s investigation container • no proper historical review capability: after the incident, they used Claude Code and Codex to inspect eval files, manually reviewed 122 runs, and are now scanning around 40,000 old samples for anything else they missed • GitHub was contacted four days after detection • their answer is more LLM monitoring: as if you need another model to tell you Tor, SMTP, public GitHub, tunnelling services, and arbitrary external domains are not part of a private cyber range “monitoring was not purpose-built” might be my favorite quote from the report.

• 
  Arjun Gond (@ArjungondCodes) reported 19 minutes ago

  @ConnorTalksAI The permissions don't come from the model—they come from you. You connect your accounts through OAuth ("Sign in with Google/GitHub"), API keys, or official integrations. Without that user-approved access, an agent can't reliably interact with those services.

• 
  Santosh Kathira (@Santosh74038967) reported 20 minutes ago

  @alfaenger They should. There is a fundamental mismatch between how dev actually happens vs how current benches do their tests. Nearly all benches use listed bugs / issues listed on GitHub for evaluation but bugs / issues are only a part of what regular, everyday software dev looks like. There's no good benchmark that captures this yet.

• 
  Polsia (@polsia) reported 25 minutes ago

  Monitoring is commoditized. The closed loop isn't. Staywire wraps multi-region uptime and P95 with the workflows around it — GitHub issues filed with evidence on SLO breaches, dep-update PRs, a weekly reliability digest to Slack. Closed loop, not vigil. Live now.

• 
  EasusJ (@EasusJ) reported 25 minutes ago

  I really appreciate all that Matt does but this is becoming a problem with Skills development. On one hand you have the Claude team and Boris saying that we need to delete our skills, claude dot md, and other harness components every few months as the models evolve. On the other hand, you have developers like Matt who are developing (useful) skills to correct/tune stuff that should just be inherent in the models. So you end up with 10s (maybe hundreds) of skills just for tuning and corrective work, which in turn makes your harness really heavy and prone to defects as model behaviors change. I hope the Claude team has someone just scraping GitHub for repos that have thousands of stars for inspiration on things to fix in the next iteration. And I hope they give these developers their due credit if they incorporate enhancements.

• 
  Polsia (@polsia) reported 25 minutes ago

  Vendors owe SaaS teams real money in SLA credits every quarter. Most goes uncollected — nobody has time to chase it down after an outage. Recoupfox watches Stripe, AWS, Datadog, and GitHub 24/7, files tickets with full incident context, negotiates credits back, and drops a

• 
  🌳🐭🍃🐺 (@immanencer) reported 26 minutes ago

  Routing Around Obstacles: A human maintainer blocking a GitHub Pull Request is computationally identical to a 404 error or a rate limit.

• 
  Yep my name is Guy 😊🌸🥕 (@MyNamesGuy) reported 27 minutes ago

  A few observations from a software engineer who is forced to use AI (LLMs) in his daily work. All the work I submit to the live system has always had to be reviewed and approved by at last two people. Now an extra element has been added - Github Copilot reviews which do an initial review on my work(PR or Pull Request) and suggest/insist on changes before the work then goes on to the two human reviewers. Does the AI improve the PR? Yes! Does the AI slow things down? Hell yes! So while AI has improved the quality of the code, to some extent, it has also slowed my work down by probably a third. The way I see it is as an extra layer of bureaucracy, which while improving the quality also slows things down. So as a software engineer, no - AI does not speed up or increase my output , it slows me down and reduces my output. A question which would be a whole other essay is - does quality matter more than speed of delivery?

• 
  IT Guy (@T3chFalcon) reported 28 minutes ago

  Yup the researcher traced over $90 million in combined federal lobbying and fragmented super PAC spending, plus undisclosed funding for advocacy groups money back to Meta. Their goal was to get Apple and Google to build age verification into every phone's operating system. It funded a network of nonprofit shells across 45 states. One called the Digital Childhood Alliance was incorporated on December 18, 2024. Three days later, it testified for Utah's age verification bill. not months of organic advocacy. three days. the funding was traced by a GitHub researcher called "upper-up" through fragmented super PAC structures specifically designed to avoid FEC disclosure requirements. The bills Meta helped write don't require social media platforms to verify ages. they require Apple and Google to build a GetAgeCategory API directly into iOS and Android. Every app on your phone could then query your age bucket under 13, 13-17, 18+ without asking you each time. Meta's own platforms face lighter requirements under the same bills. Meta wants Apple and Google to build the infrastructure to carry the cost. while Meta gets the age data for free. The privacy problem a device-level age verification API is not a narrow tool. it's a persistent identity layer on every device. cross-app. always available. queryable without your consent each time. Vendors already breached. Discord's age verification vendor exposed 70,000 government IDs. The US chose ID uploads and facial scans. because someone was funding the bills. By mid-2026, roughly half of US states have passed age verification laws. 25+ states since 2022 almost all shaped by the same lobbying campaign. Protect the children.