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:

• 
  Chris Schuchardt (@c_schuchardt) reported 7 seconds ago

  @erikzhang @kurubatermit @ngd_neo You’re the only one (besides superboyiii or Owen) with the actual permissions to remove someone from the core dev team in the GitHub repo. I was removed after my transparent 4-month school break — which I notified the entire team about in advance, with a planned return in ~90 days. This isn’t about money. It’s about getting $NEO to a proper coding and application standards. Here is the message from Discord `Neo Core Developers` server.

• 
  Chats 🇨🇦 (@ChatsFi) reported 4 minutes ago

  @ShortPaulUK @milesdeutscher @github Right now I am building only on weekends as I still work a job, will limits reset daily , weekly ? Co Pilot Pro plan mostly ran models older than Opus and GPT 5.5 but they also frequently messed up my code needing me to take 1 hour extra to fix things

• 
  Shravan Venkataraman (@theBuoyantMan) reported 4 minutes ago

  Github copilot outage? Transient API error since 2 hours ago. @github @Copilot

• 
  Tom Solid | AI Productivity (@TomSolidPM) reported 8 minutes ago

  I do not pay a penny for any app that lives in the cloud to run this. It is one folder of plain text and files on a disk I own. My journal goes back to 2017. My images to 2007. To back it up I use Time Machine, or any backup that copies a folder. No GitHub required. No vendor login. No black box. Plain text is boring, and that is the feature. It will open in thirty years. It will open in anything. No company can deprecate it, raise the price on it, or hold it hostage. What are you trusting a cloud you do not control to keep alive for you?

• 
  Asher Crowe 🪺 (@ashercrw) reported 10 minutes ago

  A 31-YEAR-OLD IN BELGRADE IS PULLING $8,400 A MONTH OFF FIVE MAC MINIS RUNNING IN A TOWER ON HIS DESK. The whole stack costs $19 a month in electricity to operate. The hardware paid for itself in week one. The setup is so quiet his girlfriend didn't notice when he turned it on. His name is Stefan. This is the cleanest example of the new solo operator economy I've seen all year and the numbers deserve a full breakdown. The hardware is five M4 Mac Minis stacked in a tower on his desk. Each one has a number written on it in marker, 1 through 5, so he knows which node dropped when one goes silent. A pink dumbbell sits on the shelf above them. A can of compressed air on the windowsill. The whole thing hums quieter than the mini fridge in the corner. The five machines are clustered with EXO into one virtual machine. EXO is the open-source framework that lets you string together consumer hardware into a distributed inference rig without needing a degree in systems engineering. The setup runs Llama 70B locally on MLX, Apple's machine learning framework optimized for unified memory. Nothing he runs ever touches a cloud server. No API costs. No rate limits. No latency tax. The model runs on his desk and answers in milliseconds. Here's the workflow he built around it. A client uploads a raw manuscript. Anywhere from 60,000 to 120,000 words. Indie author novels, self-help books, faceless YouTube channel scripts, the kind of long-form content that needs narration but doesn't have a studio budget. The Llama 70B model does the reading work first. It ingests the raw text, cleans the formatting, splits the chapters automatically, and tags every line of dialogue with the emotional tone it should be read in. Excited. Whispered. Angry. Resigned. Then it writes the chapter descriptions that faceless YouTube channels paste directly under their uploads. All of it done locally. All of it done in one pass. Then an open voice model on the same stack takes over and narrates the entire book in a single locked voice. The voice never gets tired, never asks for a re-record, never raises its day rate, never catches a cold the day before a session. The same voice across every chapter, every book, every client. Consistency that human narrators physically cannot match. A local audio mastering model handles the final polish. Compression, leveling, breath cleanup, room tone matching. The output is studio-quality audio ready for upload. The stack renders 28 hours of clean narration per month while he sleeps. He wakes up, exports the files, sends them to clients, invoices them, and goes back to whatever he wants to do with his day. Now the part that breaks people. The power draw across all five machines running at full load is 180 watts. He has a KUMAN meter plugged into the wall to track it. A single gaming PC idles higher than that. The entire AI studio he built consumes less electricity than a hair dryer on low. At Serbian residential rates that works out to roughly $19 a month in operating cost. Eight thousand four hundred dollars in, nineteen dollars out. A 442x margin on power alone before you account for the fact that the hardware paid for itself the first week he turned it on. His girlfriend asked why the power bill didn't move after he built it. He told her it can't, the machines barely draw anything. She asked what the whole thing cost to set up. He told her. She asked why he didn't build ten. That's the right question. A traditional audiobook studio has a narrator on a day rate, a booth, an engineer, and a monthly power bill that buries solo operators. The cheapest professional narrator in the US charges around $200 per finished hour. The cheapest decent one runs closer to $400. A 10-hour audiobook costs an indie author at least $2,000 in narration alone, plus mastering, plus mixing, plus the three week turnaround time while the narrator fits the project into their schedule. Stefan delivers the same product for a fraction of the cost, in 48 hours, with consistent quality across every chapter, and his only constraint is how fast he can find clients. The economics are completely deranged compared to traditional service businesses. He doesn't pay rent on a studio. He doesn't pay a narrator. He doesn't pay for cloud compute. His marginal cost per audiobook is approximately the electricity it takes to run the cluster for the duration of the render, which is measured in pennies. A few realizations worth sitting with. The frontier of AI economics is no longer in San Francisco. It's in apartments in Belgrade, Lagos, Manila, and Tbilisi, where operators with low overhead and high technical curiosity are quietly running businesses that look impossible from the outside. The geographic distribution of who actually makes money from AI is going to look nothing like the geographic distribution of who funded the labs. Local inference is the quiet revolution nobody on this app is talking about loudly enough. Every workflow that currently runs on OpenAI or Anthropic APIs has a cousin that runs on a Mac cluster for the price of an electrical outlet. The companies paying $30k a month in cloud bills are going to wake up in 18 months and find their margins eaten by operators paying $19. The audiobook market is just the beginning. Every service business with high human labor costs and predictable output requirements is about to get the same treatment. Voiceover work, transcription, translation, copywriting, image editing, video editing, customer support, technical writing. Each one of these has a local-inference version waiting to be built by someone with a stack of Mac Minis and an EXO config file. Stefan didn't invent anything. He just connected the right pieces. The pieces have been sitting on GitHub for over a year. The Mac Minis have been on shelves at every Apple Store. EXO is free. The voice models are open. The orchestration is a weekend project. The only barrier was knowing it was possible. Now you know.

• 
  Harman (@itsharmanjot) reported 16 minutes ago

  Open source NotebookLM alternative with no data limits and AI agents. Same idea as Google's NotebookLM. Same chat-with-your-docs. Same podcast generator. Same cited answers. Except this one has no source limit, no notebook limit, no 200MB file cap, and no Google login. It's called SurfSense. Google NotebookLM vs SurfSense: - Sources per notebook: 50 to 600 → Unlimited - File size cap: 200MB and 500K words → No limit - LLM choice: Gemini only → 100+ models via LiteLLM - Local LLMs: Not allowed → Full Ollama and vLLM support - Self-host: No → Yes, one Docker command - Price: $0, $19.99/mo Pro, or $249.99/mo Ultra → $0 forever Here's the wildest part: It connects to 27+ sources Google can't touch. Notion. Slack. Linear. Jira. GitHub. Discord. Dropbox. OneDrive. Gmail. Confluence. Obsidian. ClickUp. Microsoft Teams. Airtable. Your entire work life, indexed once, searchable from one chat box. 14.4K GitHub stars. 1.4K forks. 6,232 commits. Apache-2.0 license. One honest note: the README says it's not yet production-ready and still being actively developed. But it already does more than NotebookLM does, and the gap is widening every release. This is what NotebookLM should have been from the start. Repo in the first comment.

• 
  beiriannydd (@beiriannydd) reported 16 minutes ago

  Well I guess as of today, I am cancelling GitHub Copilot. It is worse than a chocolate teapot. At least you can eat that. Clean code to 2 errors to 20 errors in seconds flat. Is there a slow down and be more precise mode? I don’t know how anyone rates GPT for coding.

• 
  Gavinwang (@Gavin_Kollab) reported 19 minutes ago

  I’ll walk you through three typical Kollab workflows. The second is turning user feedback into GitHub issues, so non-technical teammates can participate in bug reporting and issue management. With Kollab, it’s not just engineers who can create GitHub issues. Product, operations, marketing, and other non-technical teammates can also report problems directly. They only need to describe the issue clearly and attach a screenshot, page link, or reproduction steps. Kollab understands the issue, evaluates its priority, and turns it into a structured GitHub issue. It can also assign the issue to the right developer based on the type of problem. For simpler issues, we can even ask Kollab to fix them directly inside an IM conversation. Because Kollab is connected to the full codebase, it does more than record the issue. It can understand the relevant code context, locate the problem, and propose or apply a fix. This turns bug reporting from scattered conversations and manual handoffs into a traceable workflow: Report the issue, create the GitHub ticket, assess priority, assign the owner, and in some cases, fix it directly.

• 
  David Hendrickson (@TeksEdge) reported 19 minutes ago

  🆕 Mistral Vibe (coding agent harness) just released some big coder updates! 🪝 before_tool & after_tool hooks Shell scripts in hooks.toml so you can deny, rewrite inputs, or append context around every tool call. Enable: enable_experimental_hooks = true 📬 Message queue while it worksType ahead freely. Esc = pause queue • Ctrl+C = drop last • Enter = flush 📝 Cleaner file edit diffs Syntax-highlighted + line numbers that match your terminal theme 🧠 Smarter compaction Re-injects your original messages after context reset so it stays on-task ✅ QoL winsTool results collapse by default • Read-only commands (ls, cat, pwd) run without approval GitHub issue automation via Skills + Studio connectors (Linear too) Open-source CLI • Web Code Mode • VS Code extension

• 
  FFmpeg (@FFmpeg) reported 21 minutes ago

  AI companies have open source initiatives. But critical infrastructure that doesn't fit the small-JS-lib-with-lots-of-GitHub-stars mold gets skipped. CC: @AnthropicAI @OpenAI @GoogleOSS your tools found real bugs in our code. Maybe help us fix the next ones before they happen?

• 
  Anthony Kroeger (@kr0der) reported 21 minutes ago

  i love how the Cursor agent window integrates PRs into the app so you don't need to open GitHub Bugbot comments all come with a "Fix with Agent" which automatically queues up a message in the chat to fix the PR comment with Cursor profiles recently being launched, and their native PR + Bugbot integrations, i actually wonder if they're building a GitHub competitor 👀

• 
  Kuzon (@Kuzon_0) reported 24 minutes ago

  @woody2788 @0xdinobtc @zuldotso Why would I take the time to entertain you when you can move your thumb down on their x page and look at the post, check the GitHub, and look at not only the deployer wallet but also major smart wallets from SOL team investors in the token. Just have some Convition. It’s not going much lower anymore.

• 
  Capafy (@Capafyai) reported 25 minutes ago

  @ar27111994 Thanks for the detail — let me make the model clear, since I think there's an expectation gap here. To host an Agent on Capafy, the publisher provides the complete runtime dependencies — i.e. the credentials and config the skill needs to run (Composio, Mem0, the LLM endpoint, etc.). The platform has no visibility into your runtime dependencies or your specific usage needs, so that part can only come from you. What the platform provides is the hosting and infrastructure to run your skill, payment settlement, and dispute/refund handling. The 20% covers those platform services — it does not cover model inference or third-party API usage. Those costs sit with you as the publisher, so it's worth pricing your skill to account for the usage it generates. So buyer runs going through the keys you supply isn't a bug — it's by design: the runtime dependencies are yours, and the platform just runs them for you. The only thing to fix is the personal keys. Don't publish with your personal Composio, local Mem0, or personal GitHub Copilot — provision a dedicated set of keys for this skill instead, so usage stays isolated and trackable and never touches your personal accounts or quota. Also, for any keys that should belong to the buyer (e.g. a token from the buyer's own account), leave those fields blank — the agent will then prompt the user to enter them, so the buyer supplies them at run time.

• 
  Phil Jeffcock 🇺🇦 🇪🇺 🐳🆔=🐘🆔=@mastodonapp.uk (@PJeffcock) reported 28 minutes ago

  🚨🚨 Yet another unplanned change to the live bodged FBA Inventory Report @amznsellerhelp @AmazonASGTG Sometime between 2026Jun11 08:36:58 BST and 2026Jun11 12:47:44 GET_FBA_INVENTORY_PLANNING_DATA (FBA Inventory Report) started including data on our entire ~1.5 million listings and the recommendations is higher than the Restock Report (including before it was broken per GitHub #5261) suggesting retirement of the Restock is nearer. Sadly it is generating files which are truncated even more frequently and probably has other issues which I don't have time to investigate right now.

• 
  Florian Darroman (@floriandarroman) reported 29 minutes ago

  @tarasshyn Or maybe skill issue? Look at the post above, GitHub is Dofollow if you know how to get there.