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:

• 
  Tolani (@Tola_niii) reported 57 seconds ago

  @n_ev_er_mi_nd @commonsmade scroll down X is below github The reason why I received zero point is because you are yet to link them Sadly the people you vouch for earlier before connecting socials will remain zero 😭

• 
  Kanika (@KanikaBK) reported 2 minutes ago

  I actually found a problem with every AI coding agent right now. It writes code, you review it, it breaks, you ask for a fix, it breaks something else, and before long you are managing the agent more than shipping while the output lives or dies by your prompt. ECC installs a coordinated engineering system on top of Claude Code so you never have to manage that process manually again. One install via Claude Code plugin. Everything changes. ↳ Plans before building, plans become reviewable artifacts not chat messages ↳ TDD workflow enforced as a gated RED to GREEN to REFACTOR sequence with evidence ↳ Code gets reviewed from a completely fresh context to catch what the same context misses ↳ Sessions get distilled into summaries, instincts, and reusable skills automatically ↳ Hooks enforce quality checks outside the prompt so they cannot be forgotten ↳ AgentShield audits your agent configuration, hooks, MCP servers, and secrets as attack surfaces ↳ Memory Vault shares context across Claude Code, Codex, Hermes, OpenCode, and Kimi in one format ↳ Plan Canvas lets you review and annotate plans in a browser instead of retyping in terminal The numbers behind it. ↳ 68 specialized agents for planning, review, build repair, security, and architecture ↳ 286 skills loaded on demand so they do not flood your context window ↳ 94 command shims for quick entry points ↳ 40,000 GitHub stars in the first three weeks after launch Works with Claude Code, Codex, Cursor, OpenCode, Kimi, Gemini, Zed, and more. MIT license. Free forever. One plugin command to install.

• 
  Maxim (@maosi100) reported 4 minutes ago

  @_lhermann Microsoft is the worst. The most issues we‘ve had in development came from supporting Windows. Bonus: failed GitHub runners that magically succeed when simply restarting

• 
  Soveryn Intelligence (@Soveryn_AI) reported 5 minutes ago

  Anyone having issues with Canva on @bot @X @SpaceXAI it’s linked to GitHub with a bad url

• 
  James Morton (@jameslmorton) reported 7 minutes ago

  @acolombiadev Can you please ask your employer to create some method for people paying you tens of thousands of dollars a month to report an outage? You are half down again right now for Actions, and you don't seem to know it. Last week, you said you were down for 1.5 hours for release files, but it was actually more like 10 hours. When Github sleeps for US hours, anything short of a full site outage will go unnoticed. Almost every week Github has some kind of critical-to-CI issue for 8+ hours, but it's rarely acknowledged.

• 
  DarkestToaster (@DarkestToaster) reported 7 minutes ago

  @TiredAndOnFire @TiredAndOnFire Fix where you can download this, Internet archive is down and your Github link is broken, you REALLY need to have better options for people to download this

• 
  Matviy (@matviy) reported 9 minutes ago

  GPT-5.6 Luna is not good. I benchmarked Cursor's GPT-5.6 Luna vs Composer 2.5 vs Grok 4.6, running each model 5 times on the same real FastAPI repository. Setup: Existing GitHub repo Task: add a POST endpoint E2E tests already exist and fail on main Success = open PR + all tests passing The model works alone across 5 isolated steps: 1. Pull the GitHub issue, create a branch, write a plan 2. Review the plan 3. Implement 4. Review against the plan and fix anything missed 5. Open the PR The results were pretty brutal for Luna. Composer 2.5: 5/5 successful runs Grok 4.6: 5/5 successful runs GPT-5.6 Luna: needed 7 runs to get 5 successes Luna failed twice at the dumbest possible step: opening the PR. It checked whether `gh` was installed, saw that it wasn’t, and basically said: “welp, guess we’re done.” No investigation, or attempts to use another method. Luna was the cheapest model in my benchmark. It was also the worst-performing one. Has anyone else used Luna for real coding tasks? How has it performed for you? Note: most of Luna's input tokens are billed as cache writes.

• 
  Hero.S (@aka_ssy) reported 10 minutes ago

  *** was designed so no central server could stop development. Then we put pull requests, CI, packages, auth, and deploy approvals on GitHub. On August 17, every engineer still had the code—and many still couldn’t ship a line of it. Distributed version control, centralized everything else.

• 
  QuietForgeSoftware (@QuietFSoftware) reported 10 minutes ago

  2.9 Billion commits a month. The fact github isn't down more is honestly what is surprising.

• 
  LastToSign (@LastToSign) reported 11 minutes ago

  GitHub now provides the first evidence that Adam has already built a basic working AI pipeline around $IMD Launch #7 explicitly states "Produced by the IdentityMD contributor network" The hook itself is not the important part. The process is: task > code generation > separate tests > AI review > bug fixes > another verification > public PR The code and tests were created using Claude Opus 5, while Codex handled the review. The reviewer found two issues, after which the implementation was sent back for fixes. The result was then rerun inside a clean environment with no internet access. The changed files were checked, followed by <forge build> and <forge test> This is the first public confirmation that IdentityMD can already split roles between AI agents and produce verifiable output. The next finding was Launch #8: Receipt Anchor. Its job is to store onchain proof of participant work. Receipts from completed attempts are designed to be batched and published in a way that prevents them from being changed or deleted. In the future, this could become the foundation for: > work history > reputation > dispute resolution > rewards The proposed IdentityMD flow now looks like this: task > AI workers > testing > independent review > completed output > permanent record of the work IdentityMD is starting to look like a working AI pipeline capable of writing code, testing it, finding bugs and rechecking the result inside an isolated environment. Receipt Anchor adds another important layer: the ability to create a verifiable history of every participant’s actions. This is much closer to real infrastructure for an autonomous web3 company than a regular set of AI tools.

• 
  The Startup Ideas Podcast (SIP) 🧃 (@startupideaspod) reported 15 minutes ago

  Skills are going to add real enterprise value to companies. Every business already writes SOPs so employees do a task the right way. A skill is the same document, written for an agent. A good skill: - Trains an agent on a task that actually makes the business money. - Runs the same way for everyone, not only the person who wrote it. - Costs nothing to hand to the next hire. Right now most skills are sitting siloed on your team's laptops. Fix that by pushing them to your company's GitHub.

• 
  अन्जित कँडेल 🕉️ 🇳🇵 (@MurkhManusya) reported 15 minutes ago

  @rodydavis and sometimes it just gives half or unfinished code(stating it has completed), which results in build errors have to confirm using another model in github copilot🥲

• 
  Fox (@0xMfox) reported 18 minutes ago

  @0x_sakata 1040 stars in a day and its just an mp4 exporter, github hype is broken

• 
  Tensor-Slayer (@TensorSlay) reported 18 minutes ago

  GitHub down again

• 
  Paul Boag (@boagworld) reported 23 minutes ago

  The migration timeline most people expect for a website rebuild is at least a year. Sometimes longer if you're moving off an old CMS and the codebase is a mess. Then you run into teams like Boulevard, who just rebuilt their entire site in three months. Moved from Contentful to Next.js static pages, migrated off Unbounce for landing pages, set up GitHub branching and PR reviews. All of it. How? They used AI-assisted development tools (V0, powered by Claude) and stopped treating the migration like a traditional engineering project. Nesting and re-nesting components, building complex abstractions, spending weeks on architecture decisions. Instead, prompt-based development. Fast iteration. Refine as you go. The interesting part isn't the speed. It's that it worked. The site's faster, easier to maintain, and actual non-developers can make changes without touching code. Turns out the slow part of most migrations isn't actually the building. It's the overthinking. #WebDevelopment #AI #Development