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:

• 
  Synonmous 🌚 (@Gem_Akinbo) reported 1 minute ago

  Most Junior Developers Don't Have a Skill Problem. They Have a Visibility Problem. Every hiring manager says the same thing: "We couldn't find anyone." Every junior developer says the same thing back: "I've built things, I just can't get anyone to look." Both are true. The bridge between them isn't more tutorials or another certificate; it's visibility. Here's the uncomfortable part: most junior developers are more capable than their CVs suggest. They've built things nobody assigned them; a password system worked out from first principles, a game engineered from scratch for a school project that never even shipped. Real problem-solving, done quietly, for no audience. I know a developer who taught himself to code on a dumb phone, reading tutorials through a micro-browser, years before he ever touched a real computer. That kind of persistence is rare. It's also, for most people, invisible; because it lived in one place: a hard drive that eventually got formatted, taking years of proof with it. That's the pattern. Skill gets built in private. Visibility never catches up. Companies aren't hiring a stack. They're hiring a way of thinking; and the only way to show that is by leaving a trail. A GitHub history with failed attempts still in it. A thread explaining a bug that took three days to find. A half-finished side project, shipped ugly, with the commit messages left in. These aren't embarrassing. They're evidence. They're the only thing that separates "I know JavaScript" from a story a stranger can actually believe. Think about how trust actually forms between two people who've never met. It's not credentials; it's pattern recognition. A recruiter scrolling through a portfolio isn't checking boxes; they're asking, "Does this person's brain work in a way I want on my team?" You can't answer that with a bullet-pointed CV. You can only answer it with a body of visible work that shows how you think when no one told you what to build. This is why the advice to "just build projects" is incomplete. Building isn't the gap. Documenting is. A junior developer who ships something rough in public, explains their reasoning, and keeps a visible record of the climb will out-compete a more skilled developer whose best work is trapped in a private folder. Not because the work is better; because it's findable. The instinct, especially early on, is to wait. Wait until the project is polished enough. Wait until the code is clean enough. Wait until the story is impressive enough to tell. But nobody discovers polished; they discover consistent. They find the person who's been quietly, publicly showing their work for months, warts included. If you're a junior developer right now, the fastest unlock isn't a new language or another course. It's an audit: what have you built that no one has ever seen? Go find it. Post it. Explain the ugly parts out loud. Your CV is a summary. Your documented journey is the evidence; and evidence is what gets you picked.

• 
  Harry Tandy (@HarryTandy) reported 3 minutes ago

  A World Cup prediction game sounds like a gimmick until you see the payments layer Sam Witteveen breaks down Google's Agent Payments Protocol in 10 minutes: 0:00 - Agents, MCP, and A2A context 1:00 - Agent Payments Protocol 1:41 - AP2 use cases 6:42 - Core principles: openness, user control, accountability, verifiable intent 8:16 - Google Agent Store 8:49 - AP2 GitHub Then open the Cyber Cup piece with one question in mind: what happens when leaderboard points depend on a tiny agent org? > one agent pulls match data > one agent checks odds and injuries > one agent decides the bet > another agent gets hired when the first 3 can't answer That is why a football tournament is such a clean test The scoreboard is public, the feedback loop is daily, and bad agents can't hide behind a polished demo

• 
  NIKHIL (@badnikhill) reported 3 minutes ago

  Wake up→open github project →write insane amounts of code → break everything→fix it like a maniac →repeat till 5AM. No sleep. No chill. Just pure unhinged contribution mode. That's how you go legendary. Who else going full degenerate this summer? #GSOC2026

• 
  Rohit Kashyap | AI + Full-Stack (@rohit_jsfreaky) reported 4 minutes ago

  @Hetzner_Online every github outage is a reminder your workflow is a tenant, forgejo on a cheap vps is a real fallback

• 
  Dirble (@Dirbles_) reported 5 minutes ago

  @Hangsiin All subagents are inheriting main thread model + effort level so any sol x high threads will just spawn more sol x high subagents i found this fix on github

• 
  Isaac Sin (@IsaacSin12) reported 6 minutes ago

  been looking for exactly this. notion mcp is painful for agents slow, and it burns tool calls to do anything simple. tried obsidian + github but syncing isn't real time, same with the obsidian vault.

• 
  Addison Lynch (@alynchfc) reported 6 minutes ago

  @martinwoodward @github @ashleymcnamara Thank you for replying. The job checks a concurrency cap then triggers an orchestration workflow (not a build). I don’t need it to run at 10min intervals and could move to a systemd timer, but the issue is the drift is significant and it looks like some runs are silently dropped

• 
  Sponge Bob (@muriisajon) reported 6 minutes ago

  Last year, GitHub saw 1 billion commits. This year, it's on pace for 14 billion. We're writing more code than ever, mostly because AI generates it faster than we can read it. ThoughtWorks is calling this "Codebase Cognitive Debt," and it's becoming a massive problem.

• 
  Rituraj (@RituWithAI) reported 6 minutes ago

  🚨 Someone built the web crawler that every AI agent actually needs. Not a scraper. Not a spider. A crawler designed specifically for feeding LLMs — structured, clean, and fast enough to process the entire web at scale. It's called Crawl4AI. 44,000 GitHub stars. The most starred web crawling repo in AI history. And it does something every other crawler gets wrong. Here's the problem. Every web crawler built before AI was the primary consumer was built for humans or databases. They returned raw HTML. Noisy. Bloated. Full of navigation menus, cookie banners, ad containers, and script tags that have nothing to do with the content you actually need. Feed that raw HTML to an LLM. You're wasting 60-80% of your token budget on noise. Your context window fills with irrelevant markup before the actual content loads. Crawl4AI returns clean, structured Markdown. Not HTML. Not JSON. Markdown — the format LLMs read most efficiently, with all the noise stripped and the structure preserved. Here's what it actually does: → Async-first architecture — crawls hundreds of pages simultaneously without blocking → LLM-ready Markdown output — clean content, no navigation noise, no ads, no cookie banners → Smart content extraction — identifies the main content block automatically, ignores boilerplate → JavaScript rendering — handles SPAs and dynamic content via Playwright integration → Media extraction — images, videos, audio all captured with context → Link analysis — internal and external links extracted and categorized → Structured data extraction — CSS selectors, XPath, and LLM-based extraction strategies → Session management — maintains login state, cookies, and browser context across requests → Proxy support — rotate proxies for large-scale crawling → Magic Mode — automatically handles consent forms, cookie banners, and overlays Here's the architecture that makes it genuinely fast. Crawl4AI uses an async browser pool — multiple browser instances running simultaneously, each handling their own queue of URLs. No sequential processing. No waiting for one page before starting the next. Hundreds of pages crawling in parallel. Combined with smart caching — pages already crawled get served from cache without re-fetching — large crawls that would take hours on a traditional crawler finish in minutes. Here's the wildest part. It ships with a Deep Crawl mode and an AI-powered extraction pipeline. You describe what you want to extract in plain English. Crawl4AI uses an LLM to intelligently extract structured data matching your description from any page — no CSS selectors, no XPath, no brittle scraping rules. "Extract all product names, prices, and descriptions" — it understands that instruction and applies it to any e-commerce page it crawls. And it has full MCP support — Claude Desktop, Claude Code, and any MCP-compatible agent can call Crawl4AI as a native tool. Your agent can crawl the web as part of its reasoning process without you writing a single line of crawling code. Your agent can now crawl any website, extract clean structured content, and use it directly in its reasoning — at the speed of async Python, at the scale of a professional web crawler. 44K GitHub stars. 6.2K forks. 847 commits. Apache 2.0 License. 100% Open Source. GitHub link in the comments 👇

• 
  Abhas Bhattacharya ⤵️ (@abhas_tweeter) reported 8 minutes ago

  @NoriSte @siddharthkp Great idea. I assume this repo is created intentionally for interviews? Or is it somehow derived from real Github codebase and old issues?

• 
  Techjunkie Aman (@Techjunkie_Aman) reported 9 minutes ago

  GitHub has quietly become one of the biggest Android app stores on the internet. The only problem? Nobody built a Play Store for it. Thousands of amazing Android apps live only in GitHub Releases. Installing them means hunting through repositories, checking release pages, downloading APKs manually, then remembering to check for updates. Developer Samyak Kamble got tired of that. So he turned GitHub itself into an app store. RepoStore automatically discovers public GitHub repositories whose latest stable release contains a real installable APK. No manual submissions, no private index, no middlemen. Every app must meet strict rules: • Public repository • Latest stable release • Real APK attached • No draft or prerelease builds The result feels remarkably polished. Material 3 UI, Material You theming, rendered READMEs, screenshots, release notes, install tracking, update detection, developer profiles, and one-tap installs, all fetched directly from GitHub. Optional GitHub sign-in boosts API limits from 60 to 5,000 requests per hour, making browsing much faster. Built entirely in Kotlin with MVVM architecture and released under the MIT License, RepoStore is the bridge between GitHub's open-source ecosystem and the app store experience Android users have always wanted. One developer got tired of digging through GitHub Releases... ...so he built the Play Store GitHub never had.

• 
  Talos (@talosbuildss) reported 14 minutes ago

  @simonsequedac @nia_thinks Good question! but the problem im solving the LLM is not involved, the data is coming from *** log and github API. Coming to hallucination again we're training to LLM to memorize anything as you can see in my day 2 the data is already indexed and put in the form of trees.

• 
  ARIJIT ROY🌠 (@arijiiiitttt) reported 16 minutes ago

  is github down?

• 
  Kevin Whinnery (@kevinwhinnery) reported 20 minutes ago

  @threepointone This was after a configuration error on our Stainless SDK repos. Some Stainless customers were temporarily added as outside collaborators in Anthropic's GitHub enterprise. All resolved now and no data was exposed, details were emailed to affected customers 🙏

• 
  Sean (@Sean1h3z) reported 22 minutes ago

  Had a small financial planning firm reach out to me for help on tech. They were already using Claude, thanks to a 21-year old intern at the office. They got offers from overseas shops asking for ~$5k/month retainers to rebuild a website and optimize it for marketing. I told them don’t do that. The intern generated one file of html with Claude, looks really nice, way better than their current website. They couldn’t figure out how to get it deployed. I told them get a GitHub and Vercel account and let’s connect in a week. Sat down with them for thirty minutes yesterday, got *** installed on their machine, showed them how GitHub works. Got Claude desktop app downloaded and had them making quick changes and deploying things within an hour. Saved them $5k/month and now they understand how to make changes to their website.