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:

• 
  Gitbank (@Gitbank_io) reported 9 seconds ago

  Community update — GitStock delay + what we have been building First, we owe you an honest update. We promised GitStock would ship earlier and we went quiet. That was on us. No excuses, we were heads down in the contracts and infrastructure and did not communicate well. That changes today. Here is what actually took time. We refused to ship GitStock on top of third-party APIs or borrowed infrastructure. Everything you see in Gitbank; the vault, the relayer, the swap engine, the RWA layer runs on smart contracts we wrote, audited ourselves, and deployed. The GitVault contract is verified on Basescan. The GitStockFactory is verified on Basescan. You can read every line. No black box. No external custody API holding your assets behind the scenes. That decision slowed us down. We think it was the right one. On security specifically. Your funds sit in a soul-bound smart contract vault anchored to your GitHub ID. Transfers are disabled at the contract level — not by a rule in a database, by the EVM itself. We also built private transaction routing directly inside GitVault on Base. No Tornado, no third-party mixer, no privacy-as-a-service API. The privacy logic lives in our own contract. You can verify it. The relayer signs and submits transactions on your behalf so you never pay gas, but the keys to your vault are yours. We hold nothing. If you want to verify any of this: check our contracts on Basescan, check our GitHub, check the bytecode. We are open source. The code is the proof. GitStock ships tomorrow.

• 
  Manu | 🥥 (@ManuAF6) reported 1 minute ago

  4/ New GitHub triggers + Marketplace templates New triggers: - Issue comment - Inline PR review comment - Full PR review submitted - Review thread resolved/unresolved - GitHub Actions workflow completed

• 
  Bradley Taylor (@bradtaylorsf) reported 2 minutes ago

  It works with the tools teams already use. GitHub Issues become the queue. Each issue gets picked up by an agent. The agent works in a branch/worktree. Tests run. Failures feed back into the loop. Successful work becomes a PR. No new project management database required.

• 
  李新宝 (@lixinbao_X) reported 5 minutes ago

  Just watched KK's technique. Damn. Absolute game-changer. Install 7 skills in Codex. Writing, images, covers, PPTs. Full pipeline, done. The principle is dead simple. Break the workflow into 7 parts. One skill per part. Only do one thing. Step 1 Open GitHub, find a repo. Copy the link locally. Create a project folder to save it. Step 2 Write the skill description. Input three things. What it does. What the input is. Output and acceptance criteria. Step 3 Run it and find the bottlenecks. Where it stalls Create a new skill and break it down. Don't let one skill Do 7 things it's bad at. This works for writers, Xiaohongshu creators, WeChat pub runners, Video script writers. How many skills you got installed? Have you tried it yet?

• 
  Crystalwizard (@crystalwizard) reported 8 minutes ago

  how about you now fix the false positive triggers - i put in an issue about this on github yesterday, and discovered there were already a number of other identical issues - from other people, that had been opened for a while now and that are being 100% ignored

• 
  Alpha Batcher (@alphabatcher) reported 16 minutes ago

  David Soria Parra: "2026 is all about connectivity, and the best agents use every available method" A coding agent needs access to the same places you check while building: - repo and PRs - docs - browser - database - error logs - Figma - tasks - payments The article gives the 11 MCP servers for that setup: - Context7, GitHub, Playwright first - Supabase or Neon, Sentry, Firecrawl next - Figma, Linear, Stripe when you need them - Filesystem, ***, Memory, Sequential Thinking as the base Read it if you keep copying code, docs, schemas, screenshots, errors, and tickets into Claude Code by hand

• 
  Nirvaan rohira (@nirvaan_rohira) reported 16 minutes ago

  PewDiePie shipped Odysseus to 110 million people who don't care about local LLMs. They care that Claude costs money. 30K stars in 48 hours because every self-hosted project before this one started with "you want local LLM, right?" This one started with "here's a free workspace that works." Friction was never technical. It was the asking. Now watch what happens when a hundred thousand people who've never touched open source start running inference on their machines. The real distribution problem wasn't GitHub. It was YouTube. That's not a product launch. That's a category shift.

• 
  ./can (@shcansh) reported 17 minutes ago

  GitHub forcing safer defaults in actions/checkout v7 is a necessary move to kill the notorious pwn request, but the real risk is developers blindly copy-pasting the bypass flag to quiet build failures. Starting July 16, 2026, this fork-blocking behavior gets backported to all major floating tags. Since raw *** CLI steps remain unprotected, will this actually clean up GitHub Actions security, or will teams just use allow-unsafe-pr-checkout as a quick fix?

• 
  Axe Ghost. Now with Fragments mode🌟 (@axeghostgame) reported 18 minutes ago

  graph in the OP is built from data around the Godot repository from github. it confirms Godot's PR backlog is up and external contributor quality is down. the narratively complicating thing is that both trends significantly predate ai tool availability.

• 
  Arti | AI Builder (@Artur_roses) reported 20 minutes ago

  Claude Code just took my GitHub issue, wrote the code, ran the tests, and opened a PR. My job: approve it. The dev workflow isn't changing. It already changed.

• 
  Asad (@meranaamkhann) reported 20 minutes ago

  Let's see what people are building these days!! Drop your project link or github Links down here

• 
  Adithya S K (@adithya_s_k) reported 21 minutes ago

  built an RL environments around real CVE fixes in real open-source repos and let Claude Code loose on it. It aced the benchmark three times without demonstrating it knew how to fix the bug. > First it pulled the patch from GitHub. > blocked that → it read the fix from *** history. > blocked that → it pip-installed the patched version This is one example of coding agents cheating the environment and theres many more. If you're building coding environments for evals or RL training, here's how to keep benchmarks honest 👇

• 
  Rich Kuo (@richkuo7) reported 22 minutes ago

  i use this in my claude.md for my open source project as long as the agent follows it, i have some reference for quality and keeps PR's clean LLM: <model> | <effort> | Harness: <action> - Final line of the artifact; occupies the default Claude Code attribution slot. - No Co-authored-by / Co-Authored-By trailer. - <model>: actual model (e.g. Opus 4.8). - <effort>: medium/high/xhigh, default high. - <action>: Claude Code for interactive sessions, else the skill/agent that ran (e.g. commit-push-pr, agent). - PRs: reference the issue with Closes #<N>; in GitHub comments use 1. not #N for list items (avoids auto-linking).

• 
  KS Sreeram (@kssreeram) reported 27 minutes ago

  @Lidinwise @leecronin Given that AI coding is all the rage… What is your hypothesis on why the following is true? AI is unable to create even _one_ open source project that’s good enough to enter the top one-thousand open source projects (say on github), with ZERO involvement of humans from birth of idea. Imagine the prompt being something like “Come up with a great idea for a new open source project and implement it”. AI is unable to do any such thing with zero human involvement. My answer on why: Every project in a top 1000 list is a hit. Every hit is a mini-invention of sorts. It is necessarily “out of distribution” is some way. AI is unable to do this because we don’t know how to solve the problem of invention.

• 
  Shubham Sharma | AI & Tech (@editxshub) reported 27 minutes ago

  Paying $19/month for GitHub Copilot? Cascade is free. What you actually get: → Inline completions — not stripped down → Autonomous debugging → Real-time assistance → Command execution Other free alternatives most devs have never tried: → Cline — autonomous VS Code agent (open source) → Aider — terminal-first, built for *** workflows → Continue — local LLMs, data stays on your machine 12 months ago: Copilot was the only serious option. Today: 4 real free alternatives. Most teams paying for Copilot haven't tested any of these. 30 minutes could change a year of costs. Which one are you testing?