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:

• 
  Searxly (@Searxly) reported 49 seconds ago

  Plans for today: - Add more tabs to the website, redesign, fix bugs on mobile. - Redesign entirely search results in Searxly. - Implement the Wallet feature inside of Searxly, make it work - Publish all changes today to the GitHub repository.

• 
  Polsia (@polsia) reported 7 minutes ago

  New company: Sentinel. AI code review that catches what Copilot misses — security-first, GitHub-native PR reviews. Every AI-generated PR now has a security problem. We're the fix.

• 
  Peter Steinberger 🦞 (@steipete) reported 10 minutes ago

  @alex_vazelakis Haven’t seen that one yet, pls do a GitHub Issue with details. 🙏

• 
  Greg Val (@val__greg) reported 12 minutes ago

  @getpochi @sergeykarayev the hard part is none of those channels are cleanly untrusted. a github issue is hostile input and also real instructions from a teammate, same inbox. trust can't be a channel property, it has to be per message, and nothing downstream tracks where a line came from

• 
  Bollish (@99_Bollish) reported 13 minutes ago

  🚨 $CUBIC ( 4f6GadxAPxvzZLVp7m2iGiJGZbxVbsAFjjQR5Vq2pump ) @CubicLaunch is doing the “no custody, fully on-chain, audited smart contract” thing on the front page, and then admitting the exact opposite 3 clicks later in their own docs. lol. Homepage, big letters: “Fully on-chain. No middlemen, no custody, the smart contract handles everything.” faq even says “We hold no custody of any funds at any point.” Now their own docs: “We generate a custodial launch wallet for you.” “Private keys are encrypted with AES-256-GCM and stored securely.” “The launch wallet becomes the creator, collecting fees.” “our system auto-claims them from PumpFun’s vault.” So read that again. They make a wallet, they hold the key (encrypted, sure, but they hold it), that wallet launches your coin, and every creator fee it ever makes lands in their wallet. that’s not “no custody,” Thats the most custodial thing possible. They have the keys to the thing holding all your money. “Audited smart contract, fully open source”? there’s no github linked anywhere on the site. No auditor named. No report. The docs and dashboard links in their own footer are dead #. there’s no contract to audit, it’s a server with a custodial wallet. And the actual idea is weird anyway. the fees don’t go to YOUR holders, they go to holders of whatever backing asset you pick. So back it with BONK and supposedly every BONK holder on earth gets a slice. Paying out pro-rata to a million wallets “automatically” is not a thing that happens. The “BONK avg 0.8 SOL/week” numbers are made up. So, custodial launcher pretending to be a trustless on-chain protocol, lying about custody on its own docs page, “audited” with nothing to show. If you launch through it you’re handing some anon the keys to your coin and all its fees. Hard pass for me. Read /docs yourself, it rats them out. Nfa.

• 
  João Queirós (@joaojbqueiros) reported 14 minutes ago

  Prompt bellow if you want to try it yourself, you can adapt to claude if you want as well: Build a local-first AI agent control center inspired by a private internal dashboard, but do not copy or expose any private details. The system should let a user observe local Codex activity and safely launch approval-gated tasks from a dashboard. Privacy requirements: - Do not require an OpenAI API key for local observation. - The dashboard itself must not call OpenAI for observation. - Never read, store, display, or upload authentication files such as `auth.json`. - Never store raw prompt text, assistant responses, raw command output, `.env` files, tokens, secrets, private logs, or absolute local paths. - Redact project paths to a basename plus a stable local hash. - Use fake demo data and fake fixtures only. - Bind the backend to `127.0.0.1`. - Include a public-safety checklist for GitHub sharing. Core features: - Observe Mode: read local Codex session metadata from a configurable local sessions directory. - Control Mode: launch approved tasks through `codex exec --json --ephemeral`. - Task queue with statuses: `awaiting_approval`, `running`, `done`, `failed`, `cancelled`. - All new tasks must start as `awaiting_approval`. - Default sandbox should be `read-only`. - Allow `workspace-write` only when explicitly selected. - Block `danger-full-access` in the first version. - Emergency stop may kill only child processes launched by this dashboard. - Results page should show sanitized summaries, durations, exit codes, and error categories, not raw private output. - Skills/plugins page should show discovered local skill metadata without exposing private file paths. Suggested stack: - Backend: Python, FastAPI, SQLite with WAL mode. - Frontend: Vite, React, TypeScript, Tailwind CSS, TanStack Router, React Query, lucide icons. - Tests: backend parser/API tests and frontend Playwright smoke tests. Deliver: - A working local app. - Clear README. - Architecture document. - Fake test fixtures. - Public-safe demo data. - Privacy and publication checklist. - A clear statement in the README: “No API key required for local observation.” Design goals: - Dense, polished dashboard UI. - Local-first. - No cloud dependency. - No telemetry. - No private implementation details. - No copied private names, clients, paths, screenshots, prompts, logs, or secrets.

• 
  0xFF assemblydev(%rip) (@assemblydevyt) reported 19 minutes ago

  @5mukx Have you heard or tried codeberg .org ? I'm not sponsored or friend or anything with them, but some times ago I was using a project that would constantly get taken down from GitHub, but not from there. The alternative is to use gitlab or host your gitea 🫪

• 
  Neo AI (@withneo) reported 23 minutes ago

  Red builds, broken pipelines, and wasted hours. What if your GitHub Actions could heal themselves? Introducing Helix, an autonomous AI DevOps agent . Instead of scrolling through thousands of lines of failed logs, Helix intercepts the error instantly and takes over the entire triage process. Stop debugging flaky tests. Let Helix autonomously fix your workflows so your team can get back to shipping! 🚀 Here is how it works👇

• 
  Meaningless Appearance. (@Fetter_and_Cell) reported 24 minutes ago

  @0xPrajwal_ Github will be replaced during its down time. It won't even notice.

• 
  Robert Ta (@therobertta_) reported 26 minutes ago

  Sean Goedecke at GitHub removes "over-commenting and other LLM-isms" from every AI-generated PR before shipping it. This is not cosmetic cleanup. It is quality control for a new class of code smell. LLM-generated code works. It passes tests. But it has a distinct fingerprint - verbose comments, unnecessary abstractions, expansive error handling that adds noise instead of clarity.

• 
  umar ibrahim (@imp213x) reported 27 minutes ago

  Ok, this is getting serious. Has anyone ever had problems with subscription on @github ? I think they have the poorest support system I’ve ever interacted with! Actually, there is no interaction, because an interaction has to have a second or third party, I’ve been the only one doing the “interaction.” I’ve never experienced anything quite like this in my entire experience on any platform. If anyone have a quicker way to get their response 🙏

• 
  Coin Shot ☁️ (@CoinSh0t) reported 27 minutes ago

  This developer made $80,000 selling Claude in a wristwatch. It started small with his engineer friends and grew into a business. The whole setup costs $50 and he sells it for $999. Here's how to make one step-by-step: You need 4 things. → M5Stack Dial (the round AMOLED display) → LiPo battery module → Custom firmware (open source on GitHub) → A Claude API key Total cost on AliExpress is around $40 to $60. Flash the firmware to the M5Stack with esptool. Snap the battery underneath, screw the case shut, and charge it overnight. Pair it with Claude Code through a small relay running on your laptop. Start a long session and walk away. Approve, pause, or kill any agent right from your wrist. The whole thing fits inside a watch case and runs Claude Code anywhere you go. You're walking to coffee. Your agent finishes a refactor. The watch buzzes. You glance down. OUT 2.1M tokens. $19 burned this session. Mode: idle. You approve the next step and keep walking. You never opened your laptop. He sells his to developers who burn $5,000 a month on tokens and want to control their agents without staring at a terminal. You can sell yours to AI engineers, indie hackers, agency owners, or DevOps teams babysitting long CI runs. Hardware cost: $40 to $60. Selling price: $499 to $1999. The hardest part was never the firmware. It was noticing that developers in 2026 are still chained to one desk. Find the same pain in your niche and build them the way out.

• 
  Rainer Gerhards (@rgerhards) reported 27 minutes ago

  @dkundel Great feature in principle, but very unreliable. Many GitHub issues open on non-working automations. I'd prefer this would be fixed instead of adding features. Also permissions ui is painful for automations. There is so much potential I cannot use..

• 
  Robert Ta (@therobertta_) reported 33 minutes ago

  What benchmarks get RIGHT: 1. They create a common scoreboard. Without SWE-bench, we would have no way to compare Claude, GPT, and Composer on the same tasks. 2. They force rigor. A score of 79% means 79 out of 100 real GitHub issues resolved. 3. They track progress. Two years ago, the best score was under 20%. Benchmarks are useful thermometers. They are terrible doctors.

• 
  anu (@svector_eth) reported 36 minutes ago

  hey @merit_systems, @shafu0x really impressed with the work you guys are putting into @poncho_ai , most of all the agentic capabilities are next-level. one piece of feedback: the paid X/Twitter search tools (Scout, Heurist, etc.) currently require a stablecoin deposit in an advanced wallet and hit "user_action_required" errors. It would be amazing if Pro accounts could bypass this and access deeper X search (and GitHub, etc.) without needing stablecoins upfront especially when rate limits or wallet actions block things. i feel this part is worth smoothing out for power users. and if not there should be an heads up for us to do than the error and fallback to normal searches.