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:

• 
  Shubh (@TheSuperEng) reported 7 minutes ago

  Tech twitter is filled with demoralizing content for people starting out in tech. Let me fix that. 1. Feeling stupid is part of the process. Seriously, you won't know everything on day one and don't even try. 2. Stop waiting to “feel ready” before building. Please just start building. People think they'll first become ready and then build, which is WRONG. 3. Build small things, not dream projects first. If you just wrote your "hello world" code right now, there is no point in comparing yourself with someone who has made computer vision and LLM projects. Don't be stupid here, please. Do things at your pace. Compare yourself to yesterday. 4. Consistency beats intensity. Don't follow Elon Musk's (work 20hrs a day). You'll hurt yourself physically and mentally doing this. Consistency is more important, even if it is just 3 hours daily. 5. It's a ******* mountain. If you have just started to code or have been coding for a few weeks, please keep in mind that you can easily get overwhelmed with so many things on the plate, like ***, github, cicd, ml, ai, python etc etc etc. You don't need to study everything on the same day, bruv. Pick one thing and work on it. CONSISTENTLY. 6. Tutorial Hell. We all know by now what it is. How to escape it? After each tutorial/lecture, build something that was taught in that class. No AI allowed (you're still learning, dude). 7. Debugging IS programming. People have formed this assumption that skilled programmers write correct functional code on their first try. NOT TRUE. Infact they spend more time debugging. In fact, debugging effectively using tools is the right way. You can't escape it. Good programmers read errors, isolate the problem, search docs, test assumptions, and iterate. 8. Break every problem down until it becomes almost boring. You'll become overwhelmed if you try to figure out how to build extremely complicated features. Rather break down a certain feature to the point that you can effectively not go simpler. Very useful for fast paced development like in a hackathon. 9. When stuck, write the steps in English before writing code. I might get hate for this but, it's worked for me quite a few times. When your brain is foggy as hell, the best thing to do is to start writing. It'll start making things clear. Maybe not instantly but, with time you'll figure it out. 10. Ask for help, but ask after trying. If you haven't tried anything to start with, how will you understand the explanation to start with? It's just like studying math. It's best to fail at trying to solve a problem as many times as possible before asking for help. Companies prefer candidates who can navigate on their own, and this is the core skill all those good candidates share.

• 
  Carver (@carverfomo) reported 9 minutes ago

  A Chinese developer posted a 41 second video on Bilibili showing the dashboard for his one person company. 19 autonomous AI agents handled everything 30 human employees used to do. Electricity bill: around 600 dollars a month. He pointed at the grid of cards on his monitor. Researcher. Copywriter. Designer. Developer. Sales. Support. Checker. The system ran 24 hours a day. He approved decisions from his phone on the subway. A laid off mid level manager at a Shenzhen electronics factory recognized the cubicle in the wider shot. He sent one screenshot to a Bilibili tech forum: timestamp 0:23. Pause at 0:23. Ignore the dashboard. Ignore the Dell monitors. Look at the giant union sign on the shelf in the upper right corner of the frame. That sign is not a decoration. That is the entry sign of his old employer's union office. The cubicle in the video was not his apartment. The cubicle was the corner desk in the union office where laid off workers came every morning for free coffee and wifi. He had not built a one person company that replaced 30 employees. He had been one of the 30. The company had laid him off six months earlier when they bought a SaaS platform that did eighty percent of what his department used to do. The 19 AI agents were real. The agents were also a demo. He had been running the system for himself for six months. He had pitched it to his former employer twice. Both times they had passed. The Dell monitors were not his. They belonged to the union steward's desk. He used them every morning from 9 AM to 1 PM because the union office had air conditioning and his apartment did not. The 600 dollars a month electricity bill was real. The electricity was the union office's. The union steward had agreed to let him plug in his local server in exchange for him helping ten other laid off workers polish their resumes. Six months ago a 14 year old in Shenzhen pushed an AI agent to GitHub. Judges said no real world application. 3,100 forks later. The laid off manager had been one of them. He had built the 19 agent system on top of that fork as proof that the company that fired him had been wrong about him. He was not a founder demonstrating the future of one person companies. He was the first laid off middle manager in his city to figure out the only way to win the AI replacement argument was to present yourself as the one who pressed the button. The clip is at 2.1 million views. The zoom on the union sign got another 1.6 million. Chinese tech viewers are still sharing the video. Still nodding. Still asking how to license the system. The system is still running. The cubicle is still at the union office. He still has not heard back from his former employer. He told the internet he had replaced 30 employees. The 30 employees he claimed to have replaced included him.

• 
  Jim Manico from Manicode Security (@manicode) reported 10 minutes ago

  @sarahyang_00 @minchoi New GitHub issue: plan carefully to extract all my data from salesforce and build an agent version specific to my business

• 
  SK Gremont (@skgremont) reported 13 minutes ago

  Satoshi didn't just build digital cash. He was trying to build an entire P2P economy, including onchain poker, an eBay-style marketplace, native escrow systems and more. Most people never knew this existed. Now the original Bitcoin Poker game is being brought back to life on BSV Blockchain (the original chain). This is what Satoshi Vision actually looked like before it got stripped down. The code was always there (GitHub). They just stopped building it. What other early Satoshi ideas will resurrect? "Bitcoin is everything!" #BSV

• 
  François Best (@fortysevenfx) reported 13 minutes ago

  I won’t be removing old tags obviously, and GitHub has immutable releases & tags, but you can never know from the outside if a repo has this turned on. That leaves the problem of discovery, the only immutable place with a modification record would be the README or a CHANGELOG file.

• 
  Emile (@eledure) reported 13 minutes ago

  github down? @github

• 
  Ed Andersen (@edandersen) reported 14 minutes ago

  I so hope these models solve the GitHub Copilot token cost problem as you’d assume MS can run them super cheap in Azure, no licensing or paying Anthropic for usage

• 
  Mad ML scientist (@HououinTyouma) reported 15 minutes ago

  like 7 years ago I was working on an obscure ML problem and used an open source package no one ever heard of. unfortunately one of the features didn't work as intended so I submitted an issue on github. just got a notification that they fixed the issue. AGI is here

• 
  Zodl (fka Zashi) (@zodl_app) reported 18 minutes ago

  Zodl v3.5.1 is now available on the App Store for iOS and on GitHub for Android. Google Play is currently reviewing the update and should release it shortly. With the Zcash network upgrade complete, updated wallet software is required to spend Orchard funds under the new consensus rules. After updating, Zodl will work as expected for sending and receiving ZEC via Orchard. As infrastructure comes back online, you may experience occasional delays. If so, run a Server Test and select the best-performing server under: Advanced Settings → Choose a Server Please note that any Orchard transactions attempted during the network upgrade window were not mined. If you are unsure about the status of a transaction, verify the TXID on the blockchain or contact @zodl_support.

• 
  rehash3d (@r3hashed) reported 19 minutes ago

  @Polymarket JUST IN: Polymarket is down 15% of the time taking over GitHub in the leaderboard

• 
  tako (@imnottristanp) reported 19 minutes ago

  i cant login my @github account. I need to push changes asap 😭😭😭

• 
  AI Crave (@wecraveai) reported 21 minutes ago

  THIS ANDROID APP TRACKS YOUR LOCATION WITHOUT GPS. NO SATELLITE. NO CELL TOWER. NO INTERNET. Just your phone's accelerometer, compass, and gyro doing math in real time. It's called DeadReckoning. GPS has one fatal weakness nobody talks about: it dies the moment you go indoors. Underground tunnels. Hospitals. Malls. Warehouses. Parking garages. Anywhere with a roof or a wall or a building nearby that bounces the signal wrong. Google Maps just stops working. Apple Maps just stops working. Every $400/year enterprise indoor positioning SDK just stops working. This free Android app doesn't stop working. Because it never needed the satellite in the first place. Here's how dead reckoning actually works: → You start at a known point → The accelerometer detects every step you take → The gyroscope tracks every turn you make → The compass holds your heading → The app integrates all three in real time and draws your path → No signal required. No towers. No beacons. No WiFi. Nothing external. It's the same navigation technique ships used before GPS existed. Sailors in the 1600s crossed oceans with it. This brings it to your Android phone. Here's the wildest part: The entire indoor positioning industry is worth billions of dollars. Cisco sells hardware beacon networks for enterprise indoor tracking. Zebra Technologies charges six figures to map a single warehouse floor. Apple built a whole proprietary protocol called Indoor Maps that requires venues to submit data directly to them. This repo does the same core thing with sensors that are already in your phone. For free. In Java. On GitHub. 115 stars. 50 forks. 55 commits. No license restrictions. One honest note: dead reckoning accumulates drift over time. The longer you walk, the more the position estimate wanders from reality. It's a physics problem, not a code problem. For short to medium distances indoors, it works. For long sessions you need a correction source. But as a foundation for indoor nav, this is the whole idea. The billion-dollar indoor mapping industry doesn't want you to know your phone already has everything it needs. Repo in the first comment.

• 
  Jimmy Koppel (@jimmykoppel) reported 23 minutes ago

  @JongwonPar9958 I see. So your result basically counts Github issues over the lifespan of the project?

• 
  Rituraj (@RituWithAI) reported 27 minutes ago

  🚨 Someone just built the tool every developer needs right now that GitHub Copilot's new token billing goes live. 60-95% fewer tokens. Same answers. One import. It's called Headroom. And it does something deceptively simple that saves real money on every single LLM call you make. Here's the problem it solves. Your AI agent calls a tool. The tool returns 50,000 tokens of output — logs, stack traces, file contents, search results, RAG chunks. Most of that output is noise. Repeated log lines. Boilerplate. Whitespace. Headers. Content the LLM will scan past without using. But you're paying for every token. Including the noise. Headroom sits between your tool outputs and your LLM. It compresses everything before it reaches the model — semantically, not just syntactically. It doesn't truncate. It doesn't randomly sample. It preserves the information that actually matters and strips what doesn't. 60-95% fewer tokens. Same answers on the other side. Here's what it actually compresses: → Tool outputs — API responses, function returns, search results → Log files — stack traces, error logs, server logs with repeated patterns → RAG chunks — document chunks from your vector database before they hit the context window → File contents — source code, configs, any file your agent reads → Any string — drop it in, get a compressed version back It also ships as an MCP server — attach it to Claude Desktop or any MCP-compatible agent and every tool output gets automatically compressed before it reaches the model. No code changes required. And as an OpenAI-compatible proxy — point your existing API calls at Headroom's proxy endpoint and compression happens transparently on every request without touching your application code. Here's why the timing matters. GitHub Copilot just switched to token-based billing yesterday. OpenAI charges per token. Anthropic charges per token. Every API you use charges per token. Every token your agent wastes on noise in a tool output is money. Headroom eliminates 60-95% of that noise automatically. The GitHub Copilot billing change that made developers furious yesterday? Headroom makes it 60-95% less painful. Today. 4.8K GitHub stars. 375 forks. Library, proxy, and MCP server all included. 100% Open Source. MIT License. GitHub link in the comments 👇

• 
  Saâd FILALI KHATTABI - FIATELPIS (@FiatElpis) reported 32 minutes ago

  Sridhar Ramaswamy, Snowflake CEO : “Extremely concerned” about rising AI costs; seeing companies switch to bare-bones AI usage in core parts of the business. Sam Altman, OpenAI CEO : OpenAI’s largest token-using customer is consuming about 100bn tokens/month; AI/token costs are a “huge issue.” Andrew Macdonald, Uber COO : Rising Claude Code/token usage is hard to justify because “that link is not there yet” between higher AI spend and more useful customer features. Aaron Levie, Box CEO : Companies will increasingly need to budget for workers and agents running up AI token bills; this starts in engineering but spreads to legal, sales, and other functions. Sundar Pichai, Google CEO : Many companies are already blowing through annual token budgets “and it’s only May”; mixing Flash with frontier models could save a lot of money. Chamath Palihapitiya, 8090 / Social Capital : “Our costs have more than tripled since November ’25”; AI costs are trending toward $10m+ per year. Brendan Foody, Mercor CEO : Mercor now spends more on AI tokens for internal agents than on employee salaries; expects compute costs to exceed headcount costs in many firms within five years. Microsoft / GitHub Copilot pricing change : Copilot moved from request-based pricing to token/credit-style consumption pricing, triggering complaints from heavy users burning through allowances quickly. Microsoft internal leadership : Reportedly winding down most internal Claude Code licenses, with cost/token considerations part of the rationale. Jensen Huang, Nvidia CEO : Nvidia engineers should use AI tokens worth roughly half their annual salary if it makes them more productive; for a $500k engineer, that implies ~$250k of tokens.