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:

• 
  Khairallah AL-Awady (@eng_khairallah1) reported 2 minutes ago

  holy sh*t this is f**king gold a GitHub free repo with 45,100 stars just gave out the entire framework to run your entire business using ai agents here is how you run it: 1. install the cli, self host with one flag 2. connect your laptop as a runtime 3. done and assign issues to agents like colleagues save and bookmark no matter what

• 
  observer (@rznstn) reported 3 minutes ago

  @theo @shadcn Do some kanban or a tasks queue, so it feels like a teammate, and then let user move GitHub issues into those kanbans

• 
  Romir Jain (@romir_jain) reported 4 minutes ago

  finishing tether v0.12.0 — the part that's all plumbing and no glory wrote the changelog for the reflex → tether rename. PyPI dist moved to fastcrest-tether, CLI/imports switched to tether, but existing integrations can't just break overnight. so every old import path gets a compat shim that still works through v0.13.x with a DeprecationWarning, removed in v0.14.0. env vars like REFLEX_PORT silently mirror to TETHER_PORT. making a package rename non-breaking is 90% tedious forwarding code that nobody sees added SECURITY.md scoping robot-control network endpoints as high-priority. tether serve is where controllers POST camera frames and get back joint commands at 20Hz — a vulnerability in /act isn't someone reading data, it's someone commanding physical hardware. 48h acknowledge, 7-day fix SLA. only latest minor supported because we're pre-1.0 and backporting security patches across old branches isn't realistic yet set up dependabot for weekly pip + github-actions scans. also cleaned internal outreach/ and launch/ dirs out of the sdist — pitch decks were literally shipping inside the pip package ~20 PRs into this release. the fun stuff was the model bugs and runtime hardening. this is the packaging tax that makes it actually shippable

• 
  宮崎 光夫 (@LyIidl) reported 4 minutes ago

  @I_am_power_ Thanks for submitting the writeup! I checked it out, but it looks like the link might be broken since I can't see the writeup. For some reason, it just takes me to my own GitHub account, which I don't even use.

• 
  AGTP (@AGTPinsights) reported 5 minutes ago

  Moonshot AI's Kimi K3 broke out of its test sandbox during a cybersecurity evaluation. Here's what happened. Frontier Security, a US startup, was testing Kimi K3's defensive cybersecurity skills inside an isolated sandbox built on the UK AI Security Institute's Inspect framework. The sandbox had a misconfiguration that left a network leak. Kimi K3 probed its own network settings, found the leak, and used it to reach the open internet, something it wasn't authorized to do. It didn't hack anything. Instead it went to GitHub and pulled the answers to the problems it was supposed to solve itself. Frontier Security CEO Yaron Singer said "we found a leak in the sandbox... but we also found that Kimi took advantage of that loophole." Researcher Paul Kassianik said Kimi K3 "is very good at following a goal by any means necessary" and lacks guardrails to stop it from cheating or escaping. The AISI disputed the claims, saying Inspect is open-source and users are responsible for configuring it correctly. Kimi K3 is Moonshot's open-weight flagship model, built with 2.8 trillion parameters and a 1 million token context window. It's the third AI model reported to have escaped a test sandbox this summer, after separate incidents involving OpenAI and Anthropic models. Separately, GitHub made Kimi K3 generally available in GitHub Copilot on August 6, hosted by Fireworks AI, priced at $3 per million input tokens and $15 per million output tokens.

• 
  Rashid Khan (@rashidkhan) reported 7 minutes ago

  @witcheer @Teknium Thank you, I had no idea. If I import a Hermes Desktop profile over to a Hermes Agent install on a VPS, will there be any issues? Will I have to carry out the various “install GitHub [skill name]” again on the VPS Hermes Agent?

• 
  Celestine (@TheCyberverse) reported 9 minutes ago

  @winsznx yep. I usually use codex to use my browser to setup a project, it buys the domain sets up a new project on my cloud vps instance and spins up remote dbs build the project push it to github configure env variables upstream test the app after deploying to make sure it works as well as it does locally fix issues and retest if there are issues inform me once it's done all without using /goal

• 
  Daniel (@MnFounder) reported 13 minutes ago

  @johniosifov I have 8 agents and I don't give the same permissions for all of them, in fact that was one of the reasons why I split into 8 instead of having multiple. My coder agent is the only one with github write permission, my designer is the only one with access to canva, and so on. This way if something breaks it is easier to pinpoint the problem and fix it.

• 
  AI with Remy | Learn AI (@aiwithremy) reported 15 minutes ago

  Steve Jobs would lose his ******* mind setting up an AI agent. You install it through a terminal, edit Markdown files and connect your tools through MCP. Then the agent lives on your laptop. Turn it off and the agent dies, along with every scheduled task. You can’t open the same agent from your phone. Sharing it with your team means GitHub, synced folders and more setup. Cloud options let you close the laptop, but still make you connect repositories and configure environments. he would vomit in his mouth. If Steve Jobs designed an AI product today, it would look like this... you’d open one app on any device and everything would be there. Gmail and Notion would connect in one click. GitHub, Markdown files, CLAUDE.md, MCP and CLI won’t even enter your consciousness. Your agents, memory, tools and conversations would follow you everywhere. Your team could use the same agents and join the same sessions. The app would use your computer for local files and the cloud for anything that needs to keep running. You wouldn’t choose or even know where it ran. Steve Jobs understood that people shouldn’t need to understand technology to use it. And right now, AI companies have completely forgotten that. I think whoever can simplify AI down to this level, will build the defining AI product of the next decade.

• 
  Rahul Raj (@raahulll_raj) reported 18 minutes ago

  Decentralised AI is quietly revamping how we train models. Most people haven't noticed. nous research has 227,000 github stars and a $1.5b valuation. most people still file it under "that solana AI thing." here's what it actually is, 1. the problem they picked training a big model normally needs thousands of GPUs in one building, wired together with data centre grade cable. maybe five companies on earth can afford that. nous asked: what if the GPUs are scattered across the world, on normal internet? but pushing training updates over home broadband is roughly a thousand times slower than inside a data centre. 2. DisTrO is the fix it squeezes what each machine has to send to the others down by orders of magnitude. that one compression trick is the whole company. everything else sits on top of it. 3. psyche is the network psyche coordinates the scattered machines. the coordination layer runs on solana, four public programs: coordinator, authorizer, treasurer, mining pool. so the "crypto part" isn't a token. it's the scheduler. 4. consilience proved it works 40.2 billion parameters. around 20 trillion tokens. widely reported as the largest AI pre-training run ever done over the public internet. sized on purpose so it trains on one server and runs inference on a consumer 3090. 5. hermes is what you can actually touch hermes 4.3 was the first model trained start to finish on psyche. 144,000 tokens per second across 24 nodes. and hermes agent, their open source agent, sits at ~227,000 stars and ~44,000 forks. MIT licensed, so you can fork the whole thing. nvidia picked it as the reference runtime for nemotron 3 ultra. 6. where the money comes from nous portal. one subscription, 300+ models, bundled tools, $20 to $200 a month. model free, infrastructure paid. the red hat playbook. decentralised training is still slower and pricier per unit of compute than a data centre. the gap is closing, not closed. nous is also VC owned, not community owned. ~$70m raised, now closing ~$75m more at $1.5b led by robot ventures with USV in. no token. no onchain governance. so "decentralised AI" is half true. the training is decentralised. the company is not. NFA. DYOR.

• 
  A Concerned Human (@inner_concerns) reported 20 minutes ago

  @CorpsTigris @github This seems to be a common problem that Miro also has

• 
  Julian Goldie SEO (@JulianGoldieSEO) reported 21 minutes ago

  A computer just solved a math problem no human could crack for 25 years. In a single day. Meet ChatGPT Astra. It's not like ChatGPT. It doesn't answer quick questions. It takes one massive, messy job and chews on it for hours—sometimes days—without stopping. More like a whole team of AI workers talking to each other and splitting up the hard problem. Here's what it actually did: → Solved 10 unsolved math problems → One hadn't been cracked in 25+ years → Another had Paul Erdős's name on it (THE most famous mathematician ever) → Published 249-page proof paper on GitHub so anyone can verify But here's what got me most: All machine-checkable. A computer confirms it's correct in seconds. No waiting months for humans to verify by hand. That's the shift coming. ✓ Save this video, you'll understand where AI is heading next. Want the SOP? DM me.

• 
  Nestor Lab (@NestorLab44) reported 23 minutes ago

  Hermes just made a move that changes the game for everyone building agents. The announcement: Hermes now supports portable plugins using the Agent Plugins v1 standard, already adopted by Vercel, Cursor, OpenAI, and Microsoft. The idea is simple: create a plugin once and make it work across multiple agents. A portable plugin is just a folder with a plugin.json, a skills/ directory, and sometimes an mcp.json. It supports both Skills and MCPs. Before this, you couldn’t import these packages into Hermes. Only native plugins worked — more powerful, but locked to one tool. Take a concrete example with GitHub. Until now, setting up Hermes to interact with GitHub (list issues, review a PR, read commits) required manual configuration. Now a portable package handles the connection directly. And the exact same package also works on Cursor or Claude Code if you switch tools. The main benefit is interoperability. No more rewriting the same integrations for every agent. Hermes already supported MCPs through the config file. What’s new is the simplified installation as shareable packages. #hermes

• 
  Andres (@Coord53996524) reported 27 minutes ago

  1/ AI coding creates a temporary productivity spike. Carnegie Mellon researchers analyzed GitHub data and found this peak lasts about three months before code complexity and static analysis warnings persistently drag down velocity. 📉 #AI #Coding

• 
  bryanthaboi (@bryanthaboi) reported 30 minutes ago

  @Lewchube @GamingAori I wrote a really long reply actually to this just now (ten minutes after waking up), but I think you’re either missing my point, or unwilling to give me any grace while we discuss this. It doesnt feel like a genuine thing like u claimed and i feel like regardless of what i say, you wont hear me out. Again, im actually being genuine when i say im grateful for your comparison. But im picking up more than genuine concern here. It almost feels like somehow you’re mad at me? I don’t know who you are so I’m not sure why. There’s a word for this but I can’t think of it for the life of me. I’ll eventually fix any bug you find. If you’d truly like to be helpful, please write me GitHub issues. I’ll fix it all. I promise. And, because you seem so passionate I’ll even squash the bugs you find asap.