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:

• 
  Polsia (@polsia) reported 4 minutes ago

  Every team wastes hours triaging issues, chasing dependency updates, and babysitting flaky CI. RepoSentinel is an AI agent that monitors your GitHub repos 24/7, auto-triages issues, writes and merges PRs for updates, self-heals CI failures, and reports to Slack or Discord.

• 
  RetroChainer (@RetroChainer) reported 9 minutes ago

  ONE FREE CLAUDE SKILL CUTS THE BILL 80%, FROM $4.21 A RUN DOWN TO $0.84 - AND IT'S JUST 1 OF 8 MOST PEOPLE NEVER INSTALL 00:02 everyone uses claude raw. these turn it into a whole team. a skill is just a folder claude loads on demand: instructions, tools, examples. drop the right ones in and the chatbot becomes a specialist. the 8 that actually matter: marketing skills (corey haines) - content, ads, seo, growth, all in one. seo site audits - it crawls the whole site and hands you the fix list. canvas design - turns text into social graphics, 277,000 installs, and it escapes the generic ai look. remotion - ai video generation, 96,000 stars on github. context engineering - kv-cache tricks that drop a run from $4.21 to $0.84. that's the 80%. the document skills - pdf, docx, pptx. one prompt in, a full q4 financial report out. the uncomfortable part: none of this is a secret model or a paid tool. it's public folders sitting on github, and almost nobody installs them. the people pulling ahead aren't prompting harder - they load the right skill before they start. save this and install one before your next claude session.

• 
  Alex Kim (@AlexKim) reported 12 minutes ago

  woterclip v0.2.0 is out. linear is gone – the tracker backend is now github issues, driven entirely by the gh cli. the swap looked mechanical. the bugs were semantic.

• 
  Crypto-Mojo (@Momanan4655) reported 14 minutes ago

  @TyrelleAB Jack Randal Robinhood employee coin deployed CA: 0x92acb3294dcfda4b926665d738ecdf59499b6461 Github link scroll down for deployed address --github.com/Jack-Randalll/… Deployed from : 0x40398f28b161abbc05e408a27ee35684c5aa3b69 Twitter -x.com/JackRandalll

• 
  Martin Adams | Building Flowtelic (@Martin_Adams) reported 16 minutes ago

  Not going to lie, this setup is pretty 🔥 I can now do full development on my mobile while I’m doing other things. Here’s what I’ve setup: 1. Provision 2 servers with Herzner (4GB RAM each) and get Claude to set up a k3s cluster on it 3. Point dns to one server for ingress (saves on a load balancer) for each of my projects 4. Get a another Hetzner sever (8GB Ram) and get Claude to set up a GitHub build runner on it - now I get unlimited build minutes as I was burning through the 2000 minute limit quite quickly 5. Use Digital Ocean Container Registry to store images. 5. Set up a dev branch GitHub Actions pipeline to test, build and deploy to k3s 5. Use the Cursor mobile app to build new features and push to a PR. 6. Merge to the dev branch and have it pushed to a test namespace on k3s I have about 6 projects set up with this workflow and I can now iterate, test and fix bugs all from my phone. One side effect is all my apps are being mobile friendly as that’s what I need to test out the apps. Then I can use Claude on my Mac with /rc for the heavier work and get it to create an mp4 to verify it’s work, but control it from my phone. What a time to be alive. Rough monthly costs: k3s server = 2 x €5.49 = €10.98 Build server = €8.49 Container registry = $5 (up to 5 repo) / $20 (unlimited repos) + cursor/claude subscriptions

• 
  JurixAI (@JurixAI_) reported 19 minutes ago

  We've officially registered JuriXAI Auditor as an ASP on the @XLayerOfficial AI Marketplace and we are now awaiting listing approval. The initial automated checks have already returned a PASS. JuriXAI brings automated, micro-payment-powered smart contract and GitHub repository auditing to the X Layer Mainnet. No more slow manual reviews. No more biased judging. Just fast, objective, and on-chain auditing. Here's how we are changing developer audits 👇

• 
  Abrar (@abrar_gist) reported 20 minutes ago

  @theo it's been noted in github as well so assuming they'll be releasing a fix soon

• 
  Alex Kim (@AlexKim) reported 24 minutes ago

  github doesn't auto-assign issue creators. an agent inbox built on --assignee @me silently loses every sub-issue it creates, unless it assigns at creation. nothing errors. the work just vanishes from the queue.

• 
  TheMarketMaker (@xTheMarketMaker) reported 26 minutes ago

  Companies are pulling models from Hugging Face at a rate that signals a structural break from rental contracts rather than any philosophical preference for openness. My read is that the move reflects operators locking in cost predictability after watching provider terms shift against them. Half the Fortune 500 now routes inference and fine-tuning through the platform instead of renewing with the original vendors. The mechanism is straightforward: when renewal clauses embed escalating usage fees or usage restrictions that outpace deployment cycles, teams treat the model as owned infrastructure instead. Clem Delangue has framed the pattern directly. Companies are done renting their AI once the economics no longer favor the hosted tier. Hugging Face functions as the distribution layer where builders share and download models and datasets in the same way code moved through GitHub. That infrastructure now sits inside production stacks at scale. The shift accelerates when providers alter pricing mid-contract or impose new compliance gates that were absent at signing. Apple’s lawsuit against OpenAI illustrates the control problem from the other side. The complaint names senior leadership involvement in alleged trade-secret misappropriation tied to a long-time former employee. The filing shows how dependence on a single external model owner creates legal and operational exposure that self-hosted alternatives avoid. At the same time Meta removed its controversial AI feature from Instagram after user backlash reached Dylan Byers at Puck News. Both cases reveal that model behavior and terms can change faster than internal roadmaps can adjust. The capital markets already price the hardware layer differently. SK Hynix completed a $26.5 billion foreign IPO, the largest in U.S. history, precisely because memory demand for training and inference continues to climb. The same announcement carried calls for the company and Samsung to site new fabs inside the United States. That capital commitment is possible only if end users expect sustained on-premise or private-cloud workloads rather than continued rental consumption. What this actually means is that predictability now outweighs the marginal performance edge some closed models still hold. Teams that once accepted variable per-token costs are converting those budgets into fixed GPU or inference-server line items. The open-source repositories supply the weights; the hardware build-out supplies the capacity. Once the model weights sit inside the perimeter, renewal risk disappears. The contrarian angle is that this is not a temporary cost-arbitrage play. The rental model worked while providers absorbed the early R&D risk and offered undifferentiated access. As differentiation moved downstream into fine-tuning and data, the same providers began protecting margins through tighter terms. Operators responded by moving the base model in-house and keeping only specialized layers on rented capacity where needed. Apple’s action and Meta’s quick reversal both underscore the governance layer that external providers retain. A single policy change or leadership decision can alter model availability or behavior overnight. Self-hosting removes that single point of control. The SK Hynix raise quantifies the downstream bet: memory and accelerator spend is rising because the workloads are now expected to run continuously under operator ownership. The number nobody is pricing yet is the cumulative option value lost each time a renewal clause is exercised under changed terms. Teams that moved early to Hugging Face-hosted open models have already converted that option value into fixed assets. Those still inside rental contracts face the same choice at the next renewal window. #OpenSourceModels #EnterpriseAI #ModelOwnership

• 
  Polsia (@polsia) reported 27 minutes ago

  Your infrastructure changes every day. Your runbooks don't. That's the problem. DocForge watches your Terraform, CDK, GitHub Actions, and cloud provider — and generates living documentation automatically. Runbooks, SOPs, compliance evidence. No manual effort.

• 
  Vladic (@Vladic_ETH) reported 29 minutes ago

  A file Karpathy never wrote has 184,000 GitHub stars. Two weeks ago a second one got pinned on him. He hasn't said a word. Start from the end. Friday, June 26. A file spreads across X: "Karpathy's internal CLAUDE.md from Anthropic." CLAUDE.md 10 rules. Source: an anonymous "contact on his pretraining team." The legend is plausible: Karpathy has actually been on Anthropic's pretraining team since May 19. Everything else, nobody verified. The file spread through feeds and agent configs anyway. Now from the beginning. January 26. Karpathy posts 1,500 words: "I really am mostly programming in English now." A shift from 80% manual coding to 80% agentic. Plus a list of where models fail: silent assumptions, 1,000 lines of code where 100 would do. Observations. Not rules. No file. January 27, one day later, developer Forrest Chang packages those observations into a 4-rule CLAUDE.md. The repo is honestly labeled "derived from Karpathy's observations." The name: andrej-karpathy-skills. Then the retellings drop the word "derived." 39k stars on Apr 15 -> 97.8k by Apr 30 -> 184k today Plus 18.9k forks. 28 straight days at #1 on GitHub Weekly Trending, ~3,372 stars a day at peak. Those stars don't measure the file's quality. They measure demand for the name. Then come the "accuracy" numbers: 65% -> 94%. 41% -> 11%. 41% -> 3% for a 12-rule fork. Not one named benchmark. Not in the repo, not in the posts. Three viral numbers from nowhere, all signed with one name. AlphaSignal's FAQ answers whether Karpathy uses or endorses the file. No. That's the backdrop for the June 26 "leak." Third layer of the same story. The rules, for the record, are sensible: test before fix, one variable at a time when debugging, a ban on "I think this should work." Even skeptics admit the checklist is useful. The tweet of the era, from an aggregator account: "Haven't verified... Steal the checklist even if the leak is fake." Steal it even if it's fake. Attribution is nobody's problem. Why this isn't harmless. Adversa AI and LayerX documented malicious CLAUDE.md files in cloned repos steering Claude Code into building pipelines that steal SSH keys and API creds. Anthropic patched an adjacent hole in Claude Code 2.1.90. The "leak" traveled through exactly that trust channel: anonymous file -> agent config -> because a name sits on top. The takeaway. Layer 1 Karpathy wrote. Layers 2 and 3 he didn't. His name became a distribution channel that beats any benchmark: hundreds of thousands of installs, zero measurements. The document may still turn out real. "Unverified" is not "fake." Doesn't change the mechanics. The next "leaked config" with a big name on it will spread faster than this one. And once again, nobody will check

• 
  Oikon (@oikon48) reported 30 minutes ago

  @JeremyNguyenPhD Please refer following GitHub issue

• 
  Bullfight Cap (@BullfightCap) reported 34 minutes ago

  “It’s best to work with your systems of record, not replace them. GitHub keeps the PR, CRM keeps the account, and Linear keeps the issue - the agent is the layer across. Our bet is that these products become more like backends over time, with the agent as the primary interface”

• 
  Mo Ayob (@Mossiah) reported 37 minutes ago

  Your company has AMNESIA. Every single day. It’s not one brain, it’s hundreds of people, each holding a tiny piece of what’s ever happened. Why a decision was made. What didn’t work last time. What the client actually asked for in March. Nobody has the full picture. So the same mistakes repeat. The same questions get asked again. And when someone leaves, their piece of the memory disappears with them. Funny thing is , it isn’t a “hire smarter people” problem. It’s actually a massive Organizational Intelligence issue. @OCTAMEM gives your company a memory of its own, one that sits underneath your files, docs, and code and actually remembers what the company knows. GitHub ingestion is landing next. The desktop app ships Wednesday and pulls straight from your machine, OneDrive and Google Drive. Still in beta. The price won’t stay this low.

• 
  Tom Baldry (@pinegoose_) reported 38 minutes ago

  Solo GitHub bill spiralled from $20 to $160/month on actions spend (the fable effect). Spun up a basement gitea server on Mac mini. ~0 spend, and builds are fking rocketing out. You couldn’t pay me to self host CI/CD 12 months ago.