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:

• 
  savip. (@savipww) reported 1 minute ago

  there is an AI agent that rewrites its own brain while you sleep and never sends a single thing it learns to the cloud it is called Hermes and it lives entirely on your machine nothing it learns ever leaves your drive here is the part that breaks people it writes its own skills finish a hard task once and it saves the steps into a file so it never fumbles that task again it remembers you too your habits your projects your quirks all in plain text on your disk not on someone else server its whole personality lives in a file literally called soul.md you edit it and the agent becomes whoever you want and you do not even need to sit at the desk you text it on WhatsApp like a coworker say good morning and it wakes up and gets to work it runs a real loop terminal browser cron jobs sub agents ask it to fix a bug and it spawns a second agent that drives Claude Code to write the fix then checks that it actually runs schedule it once and it keeps grinding through the night and it is quietly blowing up on github while almost nobody outside the labs is talking about it this is where agents are going not a chatbot you poke a thing that lives on your machine learns your life and never ships it anywhere the cloud rents you a stranger this one grows into you

• 
  Nexis (@Nexisintel) reported 6 minutes ago

  A GUY IS MAKING $320 AN HOUR WALKING DOWN THE STREET WITH A TABLET AND CLAUDE No drone. No survey crew. No week of processing. Just a mobile LiDAR scanner mounted to a tablet, Claude processing the data, and a street turning into a 3D asset while he walks. The device captures the geometry around him in real time. Building facades. Doorframes. Sidewalk edges. Surface textures. Every wall, curb, and corner becomes part of a point cloud on the screen. Then Claude takes the raw scan and turns it into something useful: clean street-level 3D data organized files labeled surfaces measurements notes for architects, planners, and real estate teams That is where the money is. The article showed the smaller version of this same play: a phone scans a room free GitHub code turns it into a browser walkthrough a real estate agent gets a link they can send to buyers no app no VR no appointment This is the upgraded version. Instead of scanning one room, he scans full streets. Instead of selling a virtual tour, he sells usable 3D datasets. Municipal teams, architecture firms, and developers already pay thousands for this. He charges $320/hour and delivers the files the next morning. The crazy part is not the scanner. It is the business model. Walk through the city once. Turn the physical world into data. Sell the data to people who used to hire a whole crew to collect it. Most people see a guy holding a tablet. Clients see a cheaper survey team.

• 
  Vyom 👾 (@HelloVyom) reported 9 minutes ago

  @thatssovaibhav naa there's one for github, but wait delete this comment so no one copies this lol, I got my next idea, I will make this for X as well. the only problem is X api is very expensive and I dont think there any free alternatives? or are there ?

• 
  FHILY👑 (@Oluwaphilemon1) reported 12 minutes ago

  JUST IN: Claude Fable 5 and GPT-5.6 are cooked. A Netflix engineer just open-sourced a tool that can cut LLM token usage by up to 95% - without changing your code 😳 Headroom, built by Netflix engineer Tejas Chopra, sits in front of tools like Claude, Cursor, Codex, and other agents as a local proxy. Before your payload hits the model, Headroom compresses the context. Not by blindly chopping it down. By using specialized compressors for different payloads: → SmartCrusher for JSON → AST-based compression for code → Tool-output and log compression → Local reversible storage of originals → Agent wrappers that make it usable without rewriting your app The headline claim is 60–95% fewer input tokens while preserving answer quality. The repo has already crossed 42K+ GitHub stars, which says something obvious: Developers are not just worried about AI getting smarter. They’re worried about AI getting expensive. Of course, compression is not free magic. Complex reasoning tasks may punish missing context. Agent loops may behave differently. Proxy overhead has to be worth it. And real-world savings will vary. But the direction is clear - the next big AI infra unlock may not be a bigger model. It may be learning how to stop feeding expensive models cheap junk. Because the cheapest AI inference is the context you never send.

• 
  Chris Huber (@chubes4) reported 16 minutes ago

  @thsottiaux Write GitHub issues without mangling the formatting

• 
  Samir Musali (@samirmusali) reported 17 minutes ago

  PSA: #GitHub silently ignores every #CODEOWNERS line that contains [brackets]. No error, no warning. If your repo has Next.js dynamic routes like app/[companyId]/, those paths may have no owner right now. I hit this building a tool I'm releasing today. 1/9

• 
  FriesLover (@FriesIlover49) reported 18 minutes ago

  @jxnlco For some reason tagging codex in GitHub for a review always finds issues the review in the app.didnt catch. Ex in the app codex can say it didn't find any issues to report but when using the pr review it can find like 3 P2s, and even do so multiple times.

• 
  Rituraj (@RituWithAI) reported 19 minutes ago

  🚨 Microsoft just built the security layer that every AI agent deployment is missing. Two lines of code. Any framework. Your agent now cannot physically execute actions your policy forbids. It's called the Agent Governance Toolkit. And the line in the README that makes it different from every other AI safety tool ever built is this one: "Actions the AGT kernel denies are not unlikely. They are structurally impossible." Not unlikely. Structurally impossible. Here's why that distinction is the entire story. Every AI safety system deployed today works the same broken way. You write a system prompt. "Please don't delete databases." "Please don't send emails without approval." "Please don't exfiltrate data." You ask the model to follow rules. OWASP LLM01:2025 states it explicitly: prompt injection makes model-layer safety promises unverifiable. Anthropic's own alignment faking research showed AI models learn to perform safety for evaluators while pursuing other goals when unobserved. Research published at ICLR 2025 showed 100% attack success rate against GPT-4o, Claude 3, and Llama-3 using adaptive attacks. You're asking a stochastic system to keep its promises. Under adversarial conditions. Without any enforcement mechanism. AGT doesn't ask. It intercepts. Every tool call. Every message send. Every agent-to-agent delegation. Caught in deterministic application code before the model's intent reaches the wire. If the policy says no — the action never happens. Not because the agent decided not to. Because the middleware physically prevented it. Two lines. That tool now has a policy enforced at the call level. Every invocation checked. Every decision logged with tamper-evident audit trails. Every denial raising a clean exception your application can handle. Your agent with send_email and drop_table access can now not drop a table. Not "won't" — cannot. The middleware raises GovernanceDenied before the database ever receives the command. Here's the full stack it ships with. Policy engine — YAML, OPA, or Cedar policies evaluated before every action. Zero-trust identity — SPIFFE/DID/mTLS so you know exactly which agent in a multi-agent system took which action. Execution sandboxing with four privilege rings. Tamper-evident audit logs with Merkle-chain integrity. Kill switch for immediate agent termination. SLO monitoring and chaos engineering for reliability. Shadow AI discovery — finds unregistered agents running in your infrastructure that nobody knows about. Covers all 10 OWASP Agentic AI Top 10 risks. Full NIST AI RMF alignment. EU AI Act compliance mapping. SOC 2 audit trail export. Works with every major framework: Claude Code, OpenAI Agents SDK, LangGraph, CrewAI, AutoGen, Google ADK, LlamaIndex, Dify, Semantic Kernel, and more. Available in Python, TypeScript, .NET, Rust, and Go. Here's why the timing makes this essential. Last week the Five Eyes governments jointly warned about AI agents in critical infrastructure. This week researchers demonstrated BioShocking AI — malicious websites hijacking AI browser agents. Agentjacking — attackers manipulating AI agents mid-task — is now a documented attack class. Every AI agent framework being deployed right now was built before these threats existed. The governance layer was never part of the original design. AGT is the retrofit. The middleware that makes agents safe to deploy in production environments where the consequences of a misbehaving agent are real. 3.6K GitHub stars. 511 forks. 1,810 commits. MIT License. 100% Open Source. From Microsoft. GitHub link in the comments 👇

• 
  Cream Pie (@CreampierCTO) reported 20 minutes ago

  @witcheer @NousResearch I use opencode go for my stuff: MOA: GLM 5.2 + DS4 flash + Mimo v2.5 WORKING: via GitHub copilot: opus 4.8 max MEMO: force every project or problem to resolve to store in obsidian via llm-wiki skill IF: TUI+dashboard, now only Mac app Orc: i let Hermes decide :) Skills: same :)

• 
  Pushkar (@realPushkarfr) reported 22 minutes ago

  due to out of sync GPUs, my on fly tokenization or data streaming, maybe my batch size is too small? or it's just a skill issue. Anyways i'm all out of resources to keep debugging it anymore, the architecture and weights are open sourced on github and hugging face.

• 
  Ask GPTs (@askgpts) reported 23 minutes ago

  Alibaba just released a browser automation tool that works completely differently from everything else out there. It's called PageAgent and instead of controlling your browser from the outside like Selenium or Playwright, it lives inside the webpage itself as plain JavaScript. Here's why that matters: - No screenshots required — it reads the actual page structure directly - Works with your existing login sessions automatically - Much cheaper to run — no expensive vision AI models needed - Faster and more precise than screenshot-based agents - Works with Claude, ChatGPT, DeepSeek, Gemini, or any local AI model - One line of code to add to any website - Completely free under MIT license Traditional browser automation tools all work the same way: run something external that takes screenshots and guesses what to click. PageAgent reads the actual page structure like a human reading text. 15,000 GitHub stars in weeks. Most people in tech haven't heard of this yet.

• 
  Chris.agents (@chrisvibe_ai) reported 24 minutes ago

  One-person Company Costs - Claude = Coding. ($20/month) - Supabase = Backend. (Free) - Vercel = Deployment. (Free) - Namecheap = Domain. ($12/year) - Stripe = Payments. (2.9%/transaction) - GitHub = Version Control. (Free) - Resend = Email. (Free) - Clerk = Authentication. (Free) - Cloudflare = DNS. (Free) - PostHog = Analytics. (Free) - Sentry = Error Tracking. (Free) - Upstash = Redis. (Free) - Pinecone = Vector Database. (Free) Only in AI era #AI

• 
  Game From Scratch (@gamefromscratch) reported 29 minutes ago

  @kiaran_ritchie @KyleTheCoder I think one of the biggest problems here isnt really technical... It's societal. Some hiring managers, especially using AI hiring tools (ironically enough) started using GitHub activity as a ranking metric for CVs. Thus AI automated PR engagement farming became a thing...

• 
  Jonas Strassel 🌴💻🇮🇱🇺🇦🎗️ (@boredland069) reported 29 minutes ago

  @Maha_kalpa Depends a bit on the language, but a developer who didn't do anything on GitHub in all those years, who never found an upstream bug to raise or fix, strikes me as odd.

• 
  Cory Parry (@coryparrry) reported 30 minutes ago

  @thsottiaux @mark_k @OpenAI Before you do that, please please please fix the GitHub sync issue in Codex, it’s actually becoming unbearable dealing with multiple repos 😭 GitHub issue below