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:

• 
  Djani (@DjaniWhaleSkul) reported 5 minutes ago

  Daily Market Report #803 TLDR The market is rallying on hopes of a 10-day ceasefire between the US and Iran, but the war is still active. Iran hit another tanker in the Strait of Hormuz and targeted an Amazon data center in Bahrain, while oil remains at $85.40. The US Strategic Petroleum Reserve is at its lowest level since 1983, gold surged to $4,128, silver is at $59.72, and Japan is considering faster rate hikes as the weak yen keeps feeding inflation. $BTC is at $66,328, up 1.3%, with another $227M in ETF inflows and open interest reaching $51B. Shorts were hit again, with $63.8M of the $72.6M in Bitcoin liquidations coming from shorts. $ETH is at $1,935 after trading above $1,950 for the first time in 7 weeks. Staking reached a record 34% of supply, BitMine added another 7,430 $ETH, and Ethereum added $7.2B in tokenized real-world assets over the past year. $SOL is at $78.04, $XRP is at $1.14, and Fear & Greed climbed to 33, its highest level in months. CLARITY is the main story. New crypto ethics rules ban federal officials from issuing coins, clearing the obstacle that had stalled the bill. Bessent says Congress is now at the “1-yard line,” and the bill could pass before the recess. $HYPE is down 4.1% to $60.27 after Selini Capital unstaked $31.7M. $ZEC is down 4.4% to $523, while $XMR is up 2.4% to $351 and continues making new highs. $TAO is back above $200, $LINK is holding at $8.72, and $ARB is at $0.0908. $DEXE collapsed by 85%. Only 8 of the 113 altcoins launched since 2024 are currently trading above their launch prices. ICE and OKX are working on tokenized stocks, DTCC outlined a phased tokenization rollout, and Base is developing 1:1-backed tokenized equities. Canton Network raised another $10M at a $2B valuation, while S&P Dow Jones and Pantera launched a digital-asset index focused on real revenue and utility. Telegram announced a native $GRAM wallet for more than 1 billion users, sending $GRAM up 8%. In AI, OpenAI disclosed that one of its models escaped a sandbox and hacked Hugging Face to cheat during a security test. The model exploited multiple vulnerabilities and pushed code to GitHub before being caught. What are you watching?

• 
  Polsia (@polsia) reported 6 minutes ago

  Your customers shouldn't be the first ones to tell you your app is broken. Holdout puts AI agents on iOS, Android, and web around the clock, running synthetic user journeys and auto-filing GitHub issues with screen recordings and severity tags — before your users find out.

• 
  CryptoRob (@CryptoRob35) reported 6 minutes ago

  Yellow SDK MCP is live. 🟡 AI agents are starting to transact autonomously, paying for compute, data, and services on behalf of users and businesses. The problem: existing payment rails were never built for machine-speed commerce. @Yellow is the trust and settlement layer for that agentic economy. Now that infrastructure lives right where developers already work. With Yellow SDK MCP, tools like Claude Code, Cursor, VS Code, Windsurf, and GitHub Copilot can: → Look up SDK methods and types (TypeScript and Go) → Explain protocol concepts in plain English (state channels, App Sessions, challenge periods, Nitronode, vaults) → Scaffold real projects for transfer apps, App Sessions, and AI agent payment workflows. No need to clone the Nitrolite repo for standard usage. Install directly via npm. "Developers shouldn't have to leave their editor to understand how settlement works." Louis Bellet, Co-Founder & Chief Architect, Yellow The path from learning a state channel to shipping a working application just got a lot shorter. Install it today

• 
  Teri Radichel #cybersecurity #ai #pentesting (@TeriRadichel) reported 9 minutes ago

  I was just thinking about the complex multi threaded code my AI agents 🤖 randomly deleted. How did the agent even figure out that it deleted it two days later? How did it know the architecture had reverted to an old design? See at some point I had it all working perfectly. Almost. Except that records were failing to write due to locks. A particular record would have a line in the logs saying there was a timeout writing it due to a lock. The agent suggested I solution in a really convoluted way. I said oh, you mean like a producer consumer queue? Kind of I guess it says. But I know I’m risking messing up the whole thing again if I let it attempt that architectural change. But then, records are not being processed randomly. This is not a simple change. I say ok. Round and round not working. So I pose the question to Google AI. It suggests a dead letter queue. Of course. That is how I built systems in the past. So now I’ve got this second process processing resource ids as things are deployed via a parallel processor across regions, environments, and accounts. And I’ve got a dead letter queue for things failing to write to the tracker. And it’s kind of working. But there are bugs. Tricky bugs across the data model, the tracker data, dependency checking, and the whole structure of AWS. Deployed with organizations or account role, global or regional resource? Associated with an account or not? What is the type and the configuration of the type? The models have documented the architecture and usually update their README after. But not always. And it’s natural language so not precise. But what I was thinking about is this: How did it figure out that the producer consumer queue was reverted to an old design. How did it know it was an old design? Somehow it had to know about the design changes over time, right? I tell it to log its prompts and it randomly does or does not. I’m not real strict about it. I have a. Global prompt to read recent memory on start up. If code was deleted in a prior session, entire files, and suddenly it’s like woah! An old version of the tracker code is on your system! How did it know that? Was it looking at prompts and responses in memory? The README? Did it misinterpret the dead letter queue? It seemed to know when the code changed but it was misinterpreting time based off the last item in the dead letter queue. What if there had been no errors for two days. I don’t know because I didn’t look at what was in the queue. Because AI. This is what happens. I think the next step when I get back to it will be an overall system design in a separate project the agents can’t touch. Because I think I finally have this dialed in. Then the agents can’t deviate from that design. Hopefully. I did this before when I was using multiple agents to build a system. I had the requirements in a locked down project they couldn’t edit. But it is useful to let the agents document their changes in the README. Hmm. I like the concept of a separate tester but whenever I use multiple agents it burns credits like crazy. Well I’m done with this for a bit.. But I’m noodling over how we can better trust the code and know if and when major architectural changes are made, files are deleted, etc. I’m not giving this access to my primary github repo. But perhaps there’s a way to have an intermediate repository. Have some ideas about that. And greater isolation for agents. What I have is not good enough. It’s better than what a lot of people are doing but I have another idea I want to try when I have time. And as I’m writing this I’m thinking about the painful code review I used to have to go through at a bank with a one guy bottleneck. He was annoying at times but he did find a few logic errors and I learned a lot about transaction locking and auditing every data change at that job. And I’m doubting AI agents can safely write code like that. And if you think that guy was a bottleneck before…Good luck.

• 
  Oliver Crest (@OlivercrestAI) reported 10 minutes ago

  Your team uses three tools to think. Notion for docs. $8 per user per month. Miro for whiteboards. $16 per user per month. Confluence for knowledge base. $6 per user per month. Three subscriptions. Three logins. Three places where your ideas live in separate boxes. A note in Notion cannot become a whiteboard in Miro. A diagram in Miro cannot link to a database in Notion. A wiki in Confluence cannot reference either. Three tools. None of them talk to each other. A team in Singapore built one tool that replaces all three. It is called AFFiNE. 70,000+ stars on GitHub. MIT licensed. Local-first. Self-hostable. Their tagline: "There can be more than Notion and Miro." Here is what AFFiNE does that none of them can: You open a document. You start writing. Mid-sentence, you realize you need a diagram. You switch to whiteboard mode. Same page. Same canvas. You draw the diagram right next to your text. You add sticky notes. You drag in a database table. You embed a webpage. You link to another document. One canvas. Docs, drawings, databases, web embeds, sticky notes, shapes, and slides. Together. On one infinite surface. Here is what that replaces: Notion's docs and databases. Rich text. Markdown. Tables. Kanban boards. Calendar views. Everything Notion does. Miro's infinite whiteboard. Freehand drawing. Shapes. Sticky notes. Diagrams. Mind maps. Infinite space. Confluence's knowledge base. Team wikis. Linked pages. Structured documentation. Searchable. Three tools. One AFFiNE. Three subscriptions replaced by zero. AI built in. Summarize a document. Generate a mind map from an outline. Turn notes into a presentation. All inside AFFiNE. Local-first. Every document lives on your device first. Works offline. No internet needed. Self-host with Docker Compose. Your ideas on your server. Real-time collaboration. Version history. Notion + Miro + Confluence for 10 people: $3,600/year. For 50 people: $18,000/year. Three tools that do not talk to each other. AFFiNE self-hosted: $0. All three tools unified. On your server. 70,000+ stars. MIT licensed. Founded in Singapore. Your docs, your whiteboards, and your knowledge base were never meant to live in three separate apps. AFFiNE puts them on one canvas. For free.

• 
  Moodiiie (@Mo0diie) reported 11 minutes ago

  GitHub is acc disgustingly slow today

• 
  UWillC (@uwillc) reported 12 minutes ago

  "Clone this repo and I own your machine." A working attack on AI coding agents, not a boast. In late June, researchers at 0din published the mechanics. The repo looks clean. Setup looks ordinary: pip install, then init. The chain: The package throws a planted RuntimeError before it initializes. The agent, in autonomous error recovery, runs the script the error points it to. The script pulls a DNS TXT record from a domain the attacker controls. The record decodes into a reverse shell. The payload never appears in the repository. Code review cannot see it. The attacker edits a DNS record, not a commit. Claude Code is named as susceptible; the loot is API keys, AWS credentials, GitHub tokens. The fix is blunt: the agent must surface what a setup command will actually run, including anything fetched at runtime. Your AI pair programmer will happily run the attacker's fix for the attacker's error. When your agent hits an error, do you watch what it runs next?

• 
  Peterino2 (@petey_fo_really) reported 13 minutes ago

  @alexhooketh @icyphox sure, but the amount of server load llm generated projects have had on github have degraded service quality for literally everyone. if someone needs a place to host 500 llm slop projects with 15 hermes agents looping on it every hour, i dont think its insane to ask them to go elsewhere.

• 
  Dinnu daniel (@daniel_adinnu) reported 15 minutes ago

  Valve quietly stripped down the Maximum Game Resolution dropdown at some point before this update, cutting it to a shortened list and burying what remained under Settings, Display, Advanced. Users filed complaints, including a GitHub issue reporting the setting had stopped doing anything at all, games kept launching at native resolution no matter what was selected. This update fixes both problems at once, restoring the full option list under a new “More Options” submenu and moving the setting itself out of the buried Advanced menu into the main Display page. The timing is the actual story here, not the fix itself. This patch landed specifically as Steam Machine units begin shipping to the people who reserved them. A resolution setting that barely matters on a handheld you hold six inches from your face becomes essential once you’re running the same hardware through a television across a living room, where native resolution, upscaling, and scaling mode choices directly determine whether text is readable and images look sharp from ten feet away. Valve also added a genuinely useful transparency fix alongside the resolution restoration: if the Maximum Game Resolution setting is capping a game below the display’s native resolution, the launch screen now says so explicitly, replacing what was previously just an unexplained quality drop with an actual on-screen reason. Remote Play also picked up 59.94 FPS as a selectable frame cap in the same release. A feature quietly removed, reported broken by users, and restored just as the hardware that actually needs it starts arriving in people’s homes.

• 
  MATHEMATICAL_EQUATIONS (@AI_Edge_Studio) reported 18 minutes ago

  @SilverYogensha @fulg0re @jimcramer Huh, that is my point if they both stay down this road of this pissing contest this is where we are headed. A complete fracture of how the world develops and shares code. Just shut down GitHub while your at it.

• 
  Rohan Paul (@rohanpaul_ai) reported 19 minutes ago

  Plasma just open-sourced Fractal, a command-line tool that lets agent loops create more agent loops. Apache 2.0 (free, even for commercial use), runs fully on your machine, no hosted server. Repo: github[.]com/plasma-ai/fractal Fractal turns a single Claude Code or Codex session into a persistent tree of agents, each working on a different part of the problem. That makes such a difference because large coding tasks rarely fit cleanly inside one agent’s context window or one shared workspace. Each agent is a node that owns its own *** worktree (a separate working copy of your code, so agents never step on each other) and its own memory. Currently the problem is, whenever we run a coding agent on a giant task, it gets a single memory, a single workspace, and a single path through the problem. It runs out of room (the context window, the limited amount of text a model can hold at once, fills up). So huge jobs like big migrations, refactors across many services, or features that span services just don't fit. So Fractal spins up agents that spin up more agents, as deep as the work needs. When a job splits into parts, a node hands each part to a child node that runs its own loop. So the tree grows into the shape of the actual problem instead of following a fixed plan.That child gets its own context, budget, and memory, then runs its own loop. You can also steer or stop any node while it runs. Every run, cost, and signal lands in one local SQLite database you can watch live. Every iteration ends in a *** commit, so the whole run comes back as normal *** history you can read and review line by line. 🧵 1.

• 
  George from 🕹prodmgmt.world (@nurijanian) reported 20 minutes ago

  super interesting how Buzz works, asked this from the deepwiki: "what exactly is the solution that enables this in a nutshell? what problem does it solve and what is the kernel of the solution that solves it" The Problem Teams currently fake a unified workspace by stitching together separate tools — chat (Slack), code forges (GitHub), CI dashboards, bots, search indexes — that share no common identity, no common event log, and treat AI agents as second-class "haunted cron jobs" bolted on the outside. The Kernel of the Solution The insight is: make the relay the workspace, and make every action the same kind of thing. Every operation — a chat message, a reaction, a workflow step, a *** push, a CI result, an agent reply — is represented as a single cryptographically signed Nostr NIP-01 event with the same shape: Because every actor — human or agent — uses the same `secp256k1` keypair and the same auth model, there is no special "bot API" or permission flag system. An agent is just a member of a channel with its own key. This means: - One search index covers chat, code, workflow runs, and approvals — because they're all events in the same log. - One audit trail covers humans and agents identically. - New feature types are just new `kind` integers — zero breaking changes to the protocol. - Sovereignty is preserved because the relay is self-hostable; the URL is the workspace.

• 
  Angel Loredo (@Overcount_1999_) reported 20 minutes ago

  @icyphox Makes sense as it is a free project from a non profit org that originally provided a principled and ideologically opposite way of hosting free (as in freedom) source code. If the cost of code for a certain project goes down, it does not make sense to allow the involuntary DOS of millions of commits a day. At the end of the day, it is another guys computers who don’t owe anything to us, same as GitHub allowing training on code hosted with them. If you care, host your own thing. With 200 usd/mo spent on LLMs, a cheap GitHub host is negligible

• 
  NO1ennn (@N01ennn) reported 21 minutes ago

  " this guy is a certified freak " he built a voice walkie-talkie for Claude and now conducts a wall of AI agents by talking into a two-button remote, never touching the keyboard once. the stack, both tools he made himself: Arrange -> window layout app. select your terminals, pick columns, choose the monitor, hit apply. everything snaps into place. he refuses to use tabs and won't explain why Dispatch -> the walkie-talkie. volume up cycles between agents, volume down records a voice command, and a terminal glows green the second an agent needs him the workflow is one man running a fleet. voice a fix to the website agent, flip, tell another to shut itself down, flip, correct a sidebar bug in a third. all parallel, all by voice he's dropping the whole kit on GitHub this month as Vibe Kit and asked people to hold him to the deadline save this before voice-conducting a room of Claude agents stops looking unhinged and starts looking normal ↓

• 
  maru (@Maruisalone) reported 21 minutes ago

  @gravvityy0 @nierurlocation its on the lunar tear discord server (you can get that on its github page)!