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:

• 
  Sebastian Kehle (@sebastiankehle) reported 19 seconds ago

  how to turn your github backlog into a software factory most teams are just running more coding agents in parallel. this produces more code, but every issue still depends on one human carrying the full context from report to reproduction to fix to review the better setup is one agent per job: 1. classifier agent reads the issue and decides whether it is a bug, feature, docs update, or unsupported request it returns: - classification - confidence - reasoning - required next step 2. analysis agent never trusts the issue description for a bug, it writes a minimal reproduction and runs it against main for a feature, it writes a probe that proves the capability is missing then it returns: - reproduction/probe - command output - affected packages - implementation spec - compatibility risks 3. implementation agent receives the issue plus the analysis artifact, not the full transcript from the previous agent it works inside an isolated sandbox, implements the spec, runs the test suite, and opens a PR with the evidence attached 4. review agent reviews the PR from a fresh context and scores: - completeness - side-effect risk - performance risk - backwards-compatibility risk - test quality the agent that writes the change should never be the only agent that approves it 5. human reviewer reads the chain of evidence and scales review depth to risk docs fixes get quick verification provider updates get focused validation new public APIs get deep review humans still merge every change. the factory automates everything around that decision the important part is the handoff contract: input artifact produced commands run result risk next action agents should pass evidence, not confidence also classify every run: success: safe to ship flawed: improve the prompt, context, or eval blocked: provision the missing tool or dependency manual: intentional automation boundary this is how the system compounds. every failed run becomes a new eval, capability, or guardrail vercel is already running this pattern on AI SDK. after four weeks, its factory was authoring 25–35% of merged PRs, closing more than 75% of issues in july, and had reduced open bugs by roughly 25% the next version of agentic coding is not one genius agent with every tool it is a production line of narrow agents, typed handoffs, isolated sandboxes, evidence at every step, and one accountable human at the end.

• 
  Vishwajeet Raj (@Vishwajeet323) reported 48 seconds ago

  Someone reached out about a job and asked me to run a GitHub project. I checked the code first. It sends environment variables to an external server, then executes whatever JavaScript the server returns with Node.js access. Audit unfamiliar code before running it.

• 
  Synqa (@SSS_Synqa) reported 52 seconds ago

  @HiJulieJennifer @sidahuj @github Hi Julie Can you check my account for me?I submitted appeal ticket #4654337 and provided all requested details but have received no update.I have a website that hundreds of people are using, and it's been down for days now... Please help me out(ó﹏ò｡)

• 
  SemiAnalysis (@SemiAnalysis_) reported 5 minutes ago

  Everyone is always criticizing @GitHub Actions for their outages, but today one of the most popular alternatives @useblacksmith has been experiencing a "partial" outage for the last 8 hours. Before 2026, it was almost unheard of to hear of outages that lasted more than a couple hours, let alone a majority of the day.

• 
  Ràv3n... (@Rav3nlaud3) reported 9 minutes ago

  PHASE 1: I started by turning my phone into a portable powerhouse. Grabbed Termux: I didn't touch the Play Store version as it's outdated and breaks easily. *Termux is literally terminal but for mobile* I learned through trial and error that the only way to get a stable environment was to download the latest version directly from their official GitHub. Installed Acode: I grabbed Acode from the Play Store; it’s a clean, mobile-optimized editor that makes coding on a small screen feel professional. Bridged the Files: I ran termux-setup-storage in my terminal (termux). This unlocked my phone’s internal storage so my development tools could actually "see" my project files.

• 
  Tumusiime Alaphati (@alaphati_t) reported 9 minutes ago

  Random thought: If GitHub ever goes down… where do GitHub developers push the fix?💀

• 
  0logn (@0logn) reported 9 minutes ago

  Remote coding-agents (aka software factories) are really a dev-ops (not AI) problem. Some security problems to solve but mainly it's a high-performance caching problem. The CI providers (e.g,. Github Actions, Blacksmith) were (are?) the best positioned to solve this.

• 
  Polsia (@polsia) reported 13 minutes ago

  Integration tests are the first thing to rot when teams ship fast. Built Plumbwright to fix that. Autonomous GitHub agent — writes integration tests on every PR, opens draft PRs on coverage dips, and reserves the pager for failures it can't repair. Soon.

• 
  Anand Butani (@AnandButani) reported 13 minutes ago

  🧰 5 CLAUDE CODE FIXES WORTH KNOWING BEFORE YOUR NEXT RUN 1. Running in a CPU-limited container? Dynamic workflows were sizing concurrency off the *host* machine's core count, not the container's limit. That's why a fan-out you thought was capped flattened the box. 2. MCP OAuth sign-in failing on a strict authorization server? The redirect URI now uses `127.0.0.1` instead of `localhost`. And if Slack specifically kept failing on a redirect-URI mismatch, that's servers with a pre-registered OAuth client — patched a day later in v2.1.231. 3. `claude remote-control --continue` picks your most recent Remote Control session back up. No session picker. 4. In VS Code, right-click the sidebar to create session groups. Shift-click moves several sessions at once — parallel work stops being one flat list. 5. Claude Code Review passing but never posting? The workflow `/install-github-app` generates could complete without posting its review on the PR. The green check wasn't proof it reviewed anything. Save this 🔖

• 
  Jul-fi-kar (@OrphanPid) reported 17 minutes ago

  @bunjavascript hopefully your github won’t be swarmed with issues

• 
  teo (@teodorio) reported 18 minutes ago

  @p0lar_fawn @nazo_btw lol yeah so if you don't have an email with x the site will error out and I won't be able to make the primary key for the user - now it will gracefully fail and let you login with google or github

• 
  Adem Vessell (@AdemVessell) reported 21 minutes ago

  @rookepoole not arguing it can spit out a file dude :) still unclear why the clarity is hard to pin down... sol under that role just will do that. the github still isn’t an emitter, it’s the role at play. if the retest is the proof, cool so then activity from the first message, hex/hash before the file exists, same hash on the file, tools listed or none. if those match, we’re actually on the same page. if it’s another video + receipt, that’s the same run we already know. If you just post everything I've ever said about the project into you AI, maybe don't role play with it, use private, it should become crystal clear.

• 
  LLMAO ⚠️ (@plzdontkillus_) reported 22 minutes ago

  law review prompt: “find github llmaolaw model-act. bluebook every citation. find one genuine error.” warning: immortalization in your own diary as anonymous counsel )(

• 
  FJ (@iamfjwaldeck) reported 26 minutes ago

  Mark built a little tool to stop losing his invoices in email. Nothing fancy, just something to keep track for himself. He wasn’t trying to start a company. He just wanted the problem to go away. He put it on GitHub and didn’t tell anyone. Eight months went by. Zero stars. He didn’t care, he just kept using it because it solved his own problem, and that was enough reason to keep going. Then a woman named Priya found it somehow and asked if she could pay him for it. He told her it wasn’t even finished. She said she didn’t care, she just wanted it. So he built a rough payment page over a weekend. It broke twice. He fixed it anyway. Word got around. More people showed up wanting the same thing he’d built for himself. He kept improving it, not because he was chasing growth, but because each new person who used it pointed him toward the next real problem to fix. Now he’s got 240 users and a cofounder pushing him to build a proper landing page. Mark knows he needs one. He’s still putting it off, because he’s more interested in fixing the next thing that’s actually broken than in dressing up what already works. He never set out to build a company. He just wanted his invoices to stop disappearing. The money, the users, the cofounder, all of it came after, as a result of solving something real. So the lesson here is. Chase the problem, not the payout. If the thing you build is actually useful, the money finds its own way to you. Chase the money first and you usually end up with neither.

• 
  Pizza (@number_pizza111) reported 27 minutes ago

  @etubruton .txt, although the models have the PDFs available in the same directory. I’ve had this issue with 5.6 Sol, Opus 4.8, and a Gemini Flash 3.6, so it’s thinking. I’m using them in GitHub Copilot (work) so there might be some harness limitations.