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:

• 
  Tolu Hunter (@tolu_hunter) reported 26 seconds ago

  Clone the repo and give it a shot, link below. Open a GitHub issue if there's a tool or retrieval method you'd want added. And if this is what your business needs built for real, my DMs are open.

• 
  Naveen Gumaste (@Z0D404) reported 1 minute ago

  🚨 BREAKING: Bun 1.4 is here, and it’s a massive upgrade! - Reduces memory usage by up to 35% - Adds 1,517 tests from the Node.js test suite - Fixes over 2,900 GitHub issues - Cuts idle CPU usage by 5× - Starts up to 50% faster on Linux And the biggest news: Bun has been rewritten in Rust! Bun 1.4 isn’t just an update; it’s a major rewrite. 🔥

• 
  f4g (@f4gch) reported 6 minutes ago

  Disagree on that, codeberg is a somewhat proper non-profit that unlike others doesn’t have hard large sponsors (feel free to correct). Looking at the graph of new AI-base projects on GitHub and them slowly having more and more a problem with outages I see a pattern Codeberg doesn’t have MSFT budget and if it ends up expanding like it had that budget it’d either HAVE to submit to large investment group sponsors or become paid both of which potentially come with hard drawbacks

• 
  🦅 (@EagleEyesCrypto) reported 7 minutes ago

  Some write up why @usedotai will be your free 10/20x DOT is sitting at ~$2M FDV while running the exact model stack the market already paid $82M and $675M for. Let me be clear: this is not a whitepaper play. It is already built. Dot is private AI on Base. No account, no logs, no training on your data. “Privacy AI” is a crowded phrase, so here is what is actually live today: DotChat. Around 19 models on one shared credit balance. Auto routing picks the best lane per prompt. DotImage / DotVideo. Uncensored generation, text to video and image to video, gallery stored browser local only. DotCode. Full agentic coding. Describe a product, the agent scaffolds it, runs commands, repairs failed builds, serves a live preview. Dot API. OpenAI compatible endpoint. Swap your base_url and your existing agents run on private inference. Connectors. Base MCP and GitHub MCP in app, with per action Allow, Ask or Never. Plus custom remote MCP by URL. Smart Privacy. This is the real differentiator. Your prompt gets rewritten before it leaves: names, wallets and locations swapped for placeholders, then restored locally. You get a receipt showing what you wrote versus what the provider actually saw. Dot Supercharged. Compound lane in private preview. Drafter proposes, verifier confirms, privacy stage cleans. Around 1.1s to first token. Dot Loom. The orchestration layer, open source on GitHub. DotPay. Fund credits with USDC or base:0x23a2847d772803f9efc64b4277b782b06296fe51 via x402 on Base. The wallet that funds is unlinkable from what it spends. Paying in $DOT gets you 10% more credits. The token mechanic: usage into revenue into buyback into burn. Supply exits as usage enters. The same structural loop Venice spent eighteen months building toward. Now the comps. $VVV: around $675M cap, around $1.15B FDV. Voorhees announced a $100M annualized revenue run rate on August 17. Privacy first AI on Base with staking for inference. $POD (Dolphin): touched $82M in May on the back of the Venice run. Their contribution is Dolphin Mistral 24B Venice Edition, the default uncensored model on Venice. Here is the part that should make you sit up: Dot runs Dolph-24B-Venice as one of its routes. Dot at $2M is a downstream consumer of the exact thing POD was valued at $82M for producing, wrapped in a Venice shaped product, with a burn loop attached. VVV’s own history is instructive. It went from under $10M to nine figures the moment the market understood that private inference is a real business and not a narrative. POD did 130% in a single day purely from proximity. And the timing. ETH is sending. Every time ETH moves, liquidity frees up that eventually rolls down into the L2s, and on Base that historically does not land in the majors, it lands in utility plays with an actual product. The rotation out of ETH beta and into Base utility is the one I think is still ahead of us, and when it comes this is the profile that catches it: built, running, and not yet priced. What you are buying at $2M: a live product with paid tiers ($12, $63, $119), an OpenAI compatible API, agentic coding, MCP connectors and a deflationary loop. Priced like it is still an idea. $DOT

• 
  Sai Krishna ⚡️ Superblog.ai (@_skris) reported 8 minutes ago

  You can create your own templates for Superblog now! Just install superblog-cli and ask claude/codex to build one by following your brand's design guidelines or handcraft custom designs. Without losing core web vitals. The below template was created with a single-prompt using claude. I was supposed to announce this on Monday but Github was down, so waited for a bit. Fun fact: Claude built the entire feature in 2 days. 16,000 lines of code. Since superblog is JAMStack but not open-source, building the instant live preview feature for developer experience was something super complex. All without shipping Superblog's core IP to the developer's machine. I'm gating this feature for customers on $99/mo plan for now. Will launch $199/mo plan and move it there.

• 
  Ankit Tripathi (@AnkitBuildsAI) reported 11 minutes ago

  Business Idea : The "SOP-to-Skill" Agency (High-Ticket B2B Consulting). The Problem: Mid-sized companies (10-100 employees) have all their "how-to" knowledge trapped in messy Notion docs, Slack messages, or the brains of their senior employees. When a senior employee quits, that knowledge leaves with them. The Workflow (Your Service): 1.)You interview their top 3 employees to understand their daily workflows (e.g., how they process vendor invoices). 2.) You build a private, version-controlled GitHub repository of SKILL.md files for their company. 3.)You create the invoice-processor Skill: It contains the exact rules for approving invoices, a Python script to extract totals from PDFs, and instructions on how to use their accounting software via MCP. Why They Pay: You aren't selling them "AI." You are selling them Organizational Memory and Time. You charge a $5,000 setup fee + $1,000/month to maintain and update their skill repository as their business changes.

• 
  Brandon Bright-Barker, PhD✨ (@AstroBarker) reported 13 minutes ago

  @ellie_huxtable Genuinely curious because I have never interacted with github (or open source) this way: what does it provide that can’t be done easily elsewhere? feed? Emojis? I’m not sure what any of this has to do with the identity of a project (beyond how devs interact on issues / MRs).

• 
  Rabnoor Singh (@RabnoorSingh10) reported 13 minutes ago

  @burkeholland @github Every tool in this category has that problem now. Reputation runs about a quarter behind the product.

• 
  Alysson 🌐 (@alysson_rx) reported 15 minutes ago

  @Vlad_GitHub you say " outage that lasted 7 hours and 47 minut" while the statuspage lies about the outage's duration. we can't trust github.

• 
  Hezha V. (@HEZHA_V) reported 15 minutes ago

  The story was that Cursor went after GitHub while GitHub was down. Read their own changelog. GitHub stays the source of truth for anything started there, and the CI partners run your existing Actions unmodified. They did not move your code. They moved where the agent sits.

• 
  Shri💐🤗 (@Shri_73_) reported 17 minutes ago

  From pencil to keyboard From Notebook to GitHub From abcd to DBMS From aree to arrays From mistake to error ...

• 
  Troy Scott (@troy_scott) reported 19 minutes ago

  @0xMovez I stopped using my IDE about 6 months. Last project I used a combination of a LLM Wiki, GitHub Issues, PRs and a UAT test harness that the agent runs and I review afterwards. One wiki page per task or issue. One GitHub Issue , one worktree and one PR per task.

• 
  CTX401 (@ctx401) reported 26 minutes ago

  @awsdevelopers My agent won’t let me access my repository. He even created a GitHub account for me, but refuses to give me the login credentials 😭

• 
  Lucas Dickey (@Lucasdickey) reported 32 minutes ago

  @greptile friends - might be worth friction logging the entire onboarding experience on mobile. Simplest and first example: "Next" button is below the fold, so I clicked the support chat icon instead. GitHub repo names get truncated on mobile, for example, too. Easy peasy stuff to fix.

• 
  Alex Reeder (@reeder1865) reported 34 minutes ago

  @emollick I recently just went back to using GitHub and Vercel vs ChatGPT Sites, as I was finding my ChatGPT app was taking up over 50GB-70GB. I just cleaned it up and got it down to around 20GB. A lot of it was just ***** branches and worktrees I hadn't bothered to clean up.