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:

• 
  X Crypto (@XCryptozc) reported 5 minutes ago

  🔴 Today is a big day for $KTA Keeta Personal is going live today. And while everyone is looking at the app, I've been watching the code. Let me show you what just dropped on GitHub 🧵 Keeta Personal is not just an app. It's a USD and EUR account. Routing numbers. Real time global transfers. Fiat and crypto in one place. Bank transfers. Crypto. Investments. All in one account. Built on a Layer 1 that has been shipping code every single day. And while the app launched today the devs did not slow down. PR #333 just merged 6 hours ago. 52 countries. 55 commits. 2141 lines of code. All GPG signed. Every single one. This was built in direct collaboration with CEO Ty Schenk as part of something called Project Gildor. The payment rails were not made up. They were pulled live from the bivo payments API. Real banking fields. Real validation rules. Real production corridors. Before this PR Keeta covered 11 countries. US. UK. UAE. Brazil. Mexico. India. Singapore. Malaysia. Hong Kong. Romania. Canada. After this PR? 63 countries. Africa. Asia Pacific. Latin America. Europe. Middle East. In one pull request. And that is on top of everything else that shipped this month. Signed metadata. Encrypted payloads. Replay attack protection. Legal disclaimers built into the payment flow. An RTP push rail connecting directly to 300 US banks. This is not a meme coin shipping a Telegram bot. This is a financial network being built to production grade standards. Keeta Personal is going live today. 63 countries of payment rails are ready. The CEO is personally reviewing code on launch day. Project Gildor is not a side quest. $KTA @KeetaNetwork @schenkty @gabe_schenk

• 
  Dima Grossman (@dimagrossman) reported 6 minutes ago

  @github actions is down again?

• 
  Success-Steps (@successstepss) reported 10 minutes ago

  Microsoft is reportedly reducing internal use of Anthropic’s Claude Code after its AI bills started exploding as employee usage rapidly increased. Some teams are now being pushed toward GitHub Copilot as the company tries to control AI costs. Uber reportedly faced a similar problem. Executives said the company had already burned through its entire yearly AI tooling budget by April because engineers were heavily using AI coding daily. AI coding tools are now being used for everything, and that level of usage creates massive compute and token costs when thousands of employees use these systems at the same time. Source: TomsHardware

• 
  Peo (@Peo_ie) reported 11 minutes ago

  @FL3SH34TER there's a fork of Citra called Citra MMJ, it lags less and should have less of these graphical errors (depends on the game) Just search "Citra MMJ" and find the github page if you want to try it

• 
  Bukunmi Ogundeji (@bukunmiogundeji) reported 15 minutes ago

  2.The 11:59 PM GitHub rush. A dev was fighting for his life to beat a deployment deadline. In a rush to fix a bug, he accidentally pushed hardcoded AWS root credentials to a public repo. Bots scraped it in minutes. By 3 AM, hackers had run up a $45,000 bill on the company’s tab

• 
  Quantum Integrable Models (@glimps_k) reported 17 minutes ago

  Ask AI don't install any unknown npm from github, it is one line hack and you are gone. Security + Privacy is a big problem even bigger than before some "zombies" are living the dream B's of online users = B's of victims Don't keep any private data online keep it offline

• 
  Docsbook.io (@docsbook) reported 17 minutes ago

  @anuruddh_m brandmd is one of those rare projects where the idea and execution both land. Problem is the docs are buried in GitHub — people bounce before they even understand what it does. Built a hosted docs site so the first impression actually sticks. Go take a look at docsbook-io.

• 
  lakshya gupta (@lakshyag404stc) reported 18 minutes ago

  @github GitHub Actions not working 😭

• 
  The obonigwe (@ogbonigwe1) reported 19 minutes ago

  Github is down again

• 
  Scott Martinis (@scottmartinis) reported 21 minutes ago

  1. Consolidate existing GTM playbooks, and 90+ days of GTM data into a GitHub repo (possibly setup a data lake here) 2. Train team on Claude code/codex/cowork, setting up connectors, deploying plugins if needed 3. Use existing data and playbooks to zero in on the weakest point of the funnel, then ship an agentic workflow to fix it week 2 4. Keep building out new agents/plugins to keep fixing the next bottleneck in the funnel... Which basically goes forever

• 
  SSNTails (@SSNTails) reported 22 minutes ago

  @spacebruce .@SaxmanSonic has been working on a new 32X sound driver and James has been more than willing to add all kinds of debugging features. If something is missing that you’d like to see, create an issue on the JGenesis GitHub with your request.

• 
  Frey (@n7runr) reported 29 minutes ago

  GitHub down?

• 
  Soumava Das (@Soumava_221B) reported 30 minutes ago

  @jatinkrmalik @github I think you’ll need to create a new account to make a community discussion page. It looks like someone from support has to manually fix this issue. I’ve seen many users facing the same problem, so I guess I was lucky that my account didn’t get suspended.

• 
  Bishwaraj Dey (@bishwa_codes) reported 30 minutes ago

  There are two broad ways to give an AI coding assistant context. Index first, retrieve later. Or search on demand, every time. Several tools chose the first. Claude Code picked the second. Here is why that choice matters. Cursor and GitHub Copilot both use semantic code search. They build indexes, compute embeddings, and use those indexes to retrieve relevant code when the model needs context. Semantic search can work even when you do not know the exact symbol name. “Where does auth validation happen?” is not always a grep-friendly question. Claude Code started the same way. Early versions had RAG with a local vector DB. The team tested it against agentic search .. and agentic search won "by a lot." Because code is fundamentally precise. "createD1HttpClient" either exists in a file or it does not. There is no fuzzy match needed. So Claude Code avoids a whole operational layer, that includes embedding jobs, vector stores, sync logic, index permissions, and questions about where the index lives. Instead, it uses grep, glob, and live file reads on demand. No whole-repo indexing. No persistent cloud vector DB. It reads only the files it needs. There are real privacy wins here too. RAG tools send all the embeddings to the cloud. And research has shown that embeddings can be partially inverted to recover the original text. So RAG approach is somewhat unsafe. However, agentic search is not free. It can burn more tokens and take more steps, especially on large codebases or vague prompts. Search for a broad term in a large React repo and the agent may need to search, inspect, narrow, and search again. Most teams treat RAG as a default. Claude Code treated it as a hypothesis, tested it, and moved on. And honestly .. a lot of RAG-based solutions would be better off without it. Especially when you find yourself restructuring your documents just to make retrieval work. At that point, you are not solving the problem. You are bending the problem to fit the tool.

• 
  Vmir (@createvmir) reported 33 minutes ago

  Github actions are down. No CI workflows work. WTF