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:

• 
  Ife (@ifedapolarewaju) reported 19 seconds ago

  Is GitHub having problems?

• 
  Adel Bucetta (@adelbucetta) reported 3 minutes ago

  @tanujDE3180 your hard drive search issues are a symptom, not the problem. github doesn't have 1 billion files like windows does.

• 
  Vaibhav Sisinty (@VaibhavSisinty) reported 3 minutes ago

  Researchers just replaced $100,000 consumer surveys with an AI model and a demographic persona. The accuracy hit 90% of human reliability. 🤯 Here is what they actually did. Colgate ran 57 real product concept surveys. 9,300 human respondents across toothpaste and personal care products. Then replicated the entire thing using AI. But here is the problem they had to solve first. When you ask an AI to rate something 1 to 5, it always picks 3. Safe. Middle of the road. Useless for real market research. So they built something called Semantic Similarity Rating. Instead of asking the AI to pick a number, they asked it to explain its purchase intent in plain text first. Then they mapped that response against anchor statements using embeddings. The result was a realistic distribution of ratings that matched what real humans actually said. 90% of human test retest reliability. Distribution similarity of 0.88 versus 0.26 for standard AI prompting. It even reproduced demographic nuance. Lower income personas rated premium products lower. Mid age groups showed more interest in familiar products. Without personas the whole thing collapses. With them it works. The global market research industry is worth $76 billion. Most of that money goes to panels, surveys, and waiting weeks for results. This runs in hours. On GPT-4o or Gemini. Code is open source on GitHub.

• 
  ClassicMain (@ClassicMain) reported 5 minutes ago

  @rodydavis @shengzheyao The "submit feedback" is a black hole. you never hear back. and stuff never get fixed. even if the app is actively bricking dozens of other programs on your machine, notably, on hundreds of other users as well as we can both gauge by the scope of the github issue.

• 
  Shamus Aran (@ShamusAran) reported 8 minutes ago

  @ZaxBit @AntonHand Joining the peanut gallery in saying this guy is absolutely right. There's a difference between being able to fix your family's router and being able to compile a github repo.

• 
  Apurv Sheldiya (@Apurv35904761) reported 9 minutes ago

  I noticed a new problem after using AI tools. Earlier, when I got stuck, the problem was: “I don’t know what to do.” Now the problem is different: “I have 10 possible things to try, but I don’t know which one is actually worth trying first.” ChatGPT gives answers. Google gives docs. YouTube gives tutorials. GitHub gives issues. StackOverflow gives fixes. But the real time waste is choosing the right path and switching between all of them. How many of you face this too? When you get stuck while working/learning/building on your computer, what wastes more time? 1. Not knowing the answer 2. Too many possible answers 3. Context switching 4. Not understanding the real root cause

• 
  Tobias (@tobwen) reported 12 minutes ago

  @cyberswayam @github @Kimi_Moonshot Github CLI was cut down to 4 old models with auto-selector.

• 
  Vikram (@vchennai2) reported 12 minutes ago

  So this is how I'm supposed to access my repo when github goes down

• 
  Abhishek Singh (@0xlelouch_) reported 17 minutes ago

  Loop engineering is simply moving from: “AI, write this function for me” to: “Here is the goal, rules, tests, tools, and stop condition. Keep working until the result is actually correct.” Instead of you manually prompting the AI 20 times, you build a repeatable loop where the AI can: 1. Understand the task 2. Inspect the codebase 3. Make a change 4. Run tests/lint/build 5. Read the failure 6. Fix itself 7. Stop only when checks pass Simple example: You want to add rate limiting to a Go API. Normal prompt engineering: “Write rate limiting middleware in Go.” AI gives code. You paste it. Build fails. You send error. It fixes it. Tests fail. You send another prompt. You are the loop. Loop engineering: “Find all public API routes. Add per-user rate limiting of 100 requests/minute. Use Redis. Do not change existing response formats. Run unit tests and integration tests. Fix failures until all tests pass. Create a PR only when coverage does not decrease.” Now the agent has a loop: Goal: Add rate limiting while requirements are not verified: inspect codebase implement smallest safe change run go test ./... run integration tests inspect errors fix errors check security/performance constraints stop when: tests pass coverage is not lower API contract is unchanged The important part is not the AI prompt. The important part is the feedback system around it. --- Remember, Good loop engineering needs: - clear goal - access to tools: code, logs, tests, GitHub, database sandbox - rules: what it must not break - verification: tests, lint, benchmarks, review - memory: what it already tried - stop condition: when to stop spending tokens and touching code Think of it like hiring a junior engineer. Bad setup: “Build something good.” Good setup: “Fix this bug. Here are the logs. Here are the tests. Do not touch payments. Run the test suite. Show proof before merging.” AI agents become useful when they are not just generating code, but are forced to observe reality and correct themselves. So prompt engineering is asking better questions. Loop engineering is building a system where the AI keeps asking itself the next useful question until the work is done.

• 
  PsudoMike 🇨🇦 (@PsudoMike) reported 17 minutes ago

  @github Finally a backup strategy that survives an S3 outage. Though knowing me I would still find a way to scratch the disc.

• 
  Ray (@Rayghattas_v) reported 18 minutes ago

  I just got my 5 hours limit reset opened codex didn’t do any coding asked to install the current working version of macOS app I’m working on and push it the app GitHub repo and I’m down to 69% how?! @OpenAIDevs @thsottiaux is such request can consume 31% of my limit?!

• 
  Mallow 🦊 (@hexandcube.com) (@hexandcube) reported 18 minutes ago

  GitHub presents a solution to their terrible uptime:

• 
  Peter (@pjmfinn) reported 20 minutes ago

  @thogge The big issue is that corporations today are basically giving away their IP to these model companies. It’s an issue with github as well, but at least you can technically request they do not use your code for model training. Not on by default btw.

• 
  void (@voidcooks) reported 26 minutes ago

  @github Why ******** would anyone ever need this? If you wanted it…even more local…?… why wouldnt you just throw it on a flash drive? The whole point of github is to have your code not local..? Please fix downtime on actions or something remotely useful lol. What a joke.

• 
  Nainsi Dwivedi (@NainsiDwiv50980) reported 26 minutes ago

  In October 2025, a builder with 30 years of experience posted a Reddit thread about AI agent personalities. By late December, the repo had 938 stars and 51 agents. Cute side project. Today it has 124,000+ stars, 20,000 forks, and 232 agents. It's called The Agency. And it stopped being a prompt library months ago. It's an org chart. 16 divisions. Engineering. Design. Marketing. Sales. Finance. Security. Product. Testing. Legal-adjacent support. Even a Game Dev division split by engine — Unity, Unreal, Godot, Roblox. Each agent isn't "act as a developer." Each one ships with an identity, critical rules, workflows, deliverables with code examples, and success metrics. The roster gets weirdly, wonderfully specific: → A Whimsy Injector who adds "celebration animations that reduce task completion anxiety" → A Reality Checker who refuses to certify anything without visual proof → An Evidence Collector who defaults to finding 3-5 issues in your code → An Anthropologist and a Historian — for world-building with actual scholarly rigor → A Korean Business Navigator. A Medical Billing Specialist. A Grant Writer. A CFO. The framing is the breakthrough: stop building one god-agent that does everything badly. Structure it like a company — specialists, clear responsibilities, handoffs between them. Deploy a squad: Frontend Dev + Backend Architect + Growth Hacker + Reality Checker, and ship an MVP with a quality gate at the end. And installing it went from "clone and copy files" to a native desktop app — macOS, Linux, Windows. Browse the roster, click, and it installs into Claude Code, Cursor, Codex, Gemini, and 10 other tools. Auto-updates included. The community translated the entire thing into 8 languages. The Japanese fork alone has 97 Japan-market original agents. MIT license. Use it commercially. Strip the branding. No attribution required. Nine months from Reddit thread to one of the fastest-growing repos on GitHub — because one person decided AI employees deserved job descriptions. Your dream team is a *** clone away. Or now, just a download. (Link in the comments)