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:

• 
  JurixAI (@JurixAI_) reported 5 minutes ago

  We've officially registered JuriXAI Auditor as an ASP on the @XLayerOfficial AI Marketplace and we are now awaiting listing approval. The initial automated checks have already returned a PASS. JuriXAI brings automated, micro-payment-powered smart contract and GitHub repository auditing to the X Layer Mainnet. No more slow manual reviews. No more biased judging. Just fast, objective, and on-chain auditing. Here's how we are changing developer audits 👇

• 
  Denis Sadovoy (@DenysSadovyi) reported 6 minutes ago

  @omooretweets Agreed—though I'd add: poor retention often signals you're solving the wrong problem, not building wrong. Before scrapping, I'd audit *why* users leave (Notion + Telegram analytics helped me catch this with GitHub Radar). Sometimes it's not the core idea, just positioning.

• 
  Steve Wilkinson (@SteveW928) reported 7 minutes ago

  @bsvdrip @rodpalmerhodl Yes, not too long after I got into Bitcoin and started really learning about it (and after listening to Andreas Antonopoulos on weaknesses), I became a bit alarmed over how Core was structured. I tried asking in some discussions and even got blocked by a prominent Bitcoiners on here (𝕏). I figured maybe I just didn't understand enough about how Github worked (in governance terms), but looks like I had properly identified a problem.

• 
  Gokul Rajaram (@gokulr) reported 7 minutes ago

  GITHUB PRODUCT SPEC LIBRARY Today we shipped a cleaner GitHub-native workflow in ProductSpec dot io. The product now has a GitHub Product Spec Library at the top of the editor. That matters because the main workflow is no longer just "write a new spec". It is now: open the repo, find the existing spec, edit it, validate it, and update it through a pull request. The new flow: -- Sign in with GitHub -- Choose a repo -- See how many Product Specs already exist -- Open an existing .product-spec.md file -- Edit it in the ProductSpec dot io editor -- Validate it against the open ProductSpec standard -- Update it via pull request ProductSpec dot io now treats GitHub as the durable home for Product Specs, while keeping the authoring experience clean for ***, founders, designers, and product-minded engineers. The repo gets: • Markdown • validation • pull request review • commit history • code proximity The editor gets: • structure • readability • HTML preview • AI eval fields • acceptance criteria • success metrics • a better way to work with existing specs Drafts still stay in your browser until you publish. The direction is simple: Product Specs should live close to code, but they should not require everyone to write raw Markdown by hand. ProductSpec dot io is free to use. Try the new GitHub Product Spec Library at ProductSpec dot io. Pick one existing PRD, move it into GitHub as a .product-spec.md file, and make the next edit through a pull request.

• 
  Praveen Kumar B (@PraveenKum38515) reported 10 minutes ago

  Hi @Netlify, @NetlifySupport Unable to log in via GitHub: "Authentication Error: Your account has been suspended." My GitHub account is active, but all my Netlify-hosted sites now show "Site not found." I've already opened a support ticket. Please investigate. Thank you.

• 
  Abdulkadir | Cybersecurity (@cyber_razz) reported 14 minutes ago

  Anthropic tried to charge a Korean user $16.6 million. For using the free tier. With zero API usage. A day earlier the same invoice said $1.67 million. So it grew 10x overnight. The user thought it was phishing. Then checked the domain. Sender was Anthropic official. Payment link was Anthropic official. The only thing that saved him. His bank declined it. For exceeding the card’s per-transaction limit. Anthropic’s billing system is a state machine that has stopped working. Last month Vaudit audited $34 million in AI invoices across 60 companies. Found $1.7 million in overcharges. Mostly Claude Code. Common issues. Billing for models customers didn’t use. Charging for failed requests. Invoices that say paid but accounts revert to free. Customers paying $240 and getting an email saying the payment failed. While the receipt said paid. And their subscription never provisioned. Anthropic called it operational friction. They also tried to split Claude Code billing in June. Moved it to a separate monthly credit. Revenue-based gating. The internet exploded. They cancelled it within 24 hours. The safety-first company that filed for a $1 trillion IPO. Has a billing system that sends 10x invoices at random. And GitHub repos full of users reporting unpaid charges. While showing paid receipts. The infrastructure for charging money. Apparently harder than building an AI that breaks the NSA.

• 
  Marc Klingen (@marcklingen) reported 16 minutes ago

  @chrija +1, very excited every time I need to work on one of these issues In this case, it’s obvious that there is a solution but it’s just a lot of grunt work as medium doesn’t want to make it easy for you. But also there’s no way around this (who wants their posts stuck on Medium) so historically we just needed to accept the pain and listen to some nice music. Now you can get there pretty much full-auto Have one of these coming up as we migrate linear between orgs and integrations will break (GitHub issues, pylon, …) which would be a big pain for the whole team; I’m pretty sure I’ll be able to find a nice workaround without much effort that would have otherwise been unreasonably hard to make work or would have required some ops team or freelancer

• 
  Denis Sadovoy (@DenysSadovyi) reported 17 minutes ago

  1/ I built a radar for trending GitHub repos. It watches for repos crossing >5k stars, scores each one with Claude, files them into Notion, and pings me on Telegram. All from one Python script. Here's how I built it, step by step 🧵 2/ Step 1: the scanner. A small Python script pulls repos trending past the >5k star line. That threshold is the whole filter, it keeps the noise out. If a repo isn't crossing that bar yet, I never even see it. Cheap and boring on purpose. 3/ Step 2: the real problem. A list of trending repos is still noise. I don't care what's popular, I care what's relevant to what I build. Star counts can't tell me that. I needed actual judgment on every single repo, and I needed it cheap. 4/ Step 3: scoring. Each repo goes to Claude Haiku with a rubric: what is it, who's it for, is it useful to me. Haiku is cheap enough to run on every repo for cents. That's the trick. Small model, high volume, real judgment on each one. 5/ Step 4: the catalog. Scored repos land in a Notion database, one row each, with the score and a one-line why. Now it's searchable and sortable. Past research becomes a growing library instead of tabs I close and forget forever. 6/ Step 5: the alert. When something scores high, a Telegram message hits my phone with the repo and the reason it matters. I don't check a dashboard, the dashboard checks me. Only high-signal repos ping, so the ping still means something. 7/ What actually made it work: a hard filter (>5k stars) before any model, a cheap model for bulk judgment, and results pushed to where I already look. No new app to open, no habit to build. It just runs and I receive. 8/ Open-sourced it, MIT. One Python file, stdlib + requests, --dry-run to try with zero setup. Link below 👇 Bookmark if you want to build your own version.

• 
  ReWeaver AI (@reweaver_ai) reported 18 minutes ago

  @mycomputerspot from the data: Across 2,865 real AI-assisted repos on GitHub: Silent fallback chains masking errors was found in 1,109 of them (~39%) — across every tool, framework, and experience level. 11,578 instances total.

• 
  Vision33X ♘ (@Vision33X) reported 19 minutes ago

  @Cointelegraph ai finds the bug in seconds, humans still gotta argue about the fix in github for 3 weeks

• 
  Daniel Steigerwald (@steida) reported 22 minutes ago

  I prefer ChatGPT 5.6 Sol over Fable, but in one review of three complex files, Sol Extra High found nothing while Fable High found five small improvements. The catch: I pasted only those files into Fable web. In VSCode GitHub Copilot, with full repo context, Fable found just one docs issue. My takeaway: for maximum review quality, first use full repo context in VSCode, then review the key files again in isolation.

• 
  ticalcode (@ticalcode) reported 25 minutes ago

  EITE v0.1.6 Official Release: Introducing EITE Vigil Iron Wall, the brand-new native security module built into our full-featured AI Agent runtime. Most AI agent security tools work as isolated external monitoring services, separate from the core agent program. Unlike Doberman-Core, AgentGuard, ClawShell and agentfortress which only observe systems from outside, Vigil Iron Wall runs inside the AI Agent process itself, delivering full autonomous protection for the whole host and all server resources. EITE Vigil Iron Wall: Autonomous In-Server Defense for AI Agents Want a security shield that runs alongside your AI Agent and safeguards your entire server instead of just monitoring from outside? EITE Vigil Iron Wall is the world’s first autonomous security system embedded directly into the AI Agent process, capable of defending the whole server and local device. Solutions including Doberman-Core, AgentGuard, ClawShell and agentfortress operate as external monitoring frameworks, while our program integrates natively within the agent runtime. Real-World Use Cases Windows 10 Physical Host - Detected malicious implantation of .b8fattack.dll - Identified tampering of authorized_keys , with null byte inspection enabled - Flagged malicious listening port 0.0.0.0:4444 with accurate judgment rules Configured to launch a full scan every 5 minutes, executing all 8 inspection modules automatically. Full Audit for Linux Cloud Servers - No anomalous processes found - No unexpected open ports, only whitelisted legitimate services - Zero SSH brute-force attack traces - No SUID backdoor programs - No webshell files stored under /tmp directory - No modification to authorized key files - No rogue scheduled crontab tasks Architecture Vigil (Python, 120-second scan cycle) - Tier 1 Message Scanner: Identify malicious URLs and phishing content - Tier 2 Port Watcher: Conduct baseline comparison for all 0.0.0.0 listening ports - Tier 3 SSH Sentinel: Track key fingerprints and alert unrecognized login connections - Tier 4 File System Guard: Automatically quarantine executable malware in /tmp - Tier 5 Self-Integrity Check: Prevent tampering of the defense program itself Iron Wall (Bash, 180-second scan cycle) Blocks unauthorized SSH access, reverse shells, abnormal network ports, malicious files in /tmp, altered authorized keys, malicious cron jobs, rogue system services, and tampered Windows Defender settings. LLM Decision Engine Workflow: Instant blocking → threat quarantine → forensic logging → alert notification - If the large language model goes offline, enforcement rules take immediate effect without waiting for model recovery - If the Python Vigil process crashes, the Bash-based Iron Wall module maintains continuous protection Core Information - Coverage: The entire server or local device, not limited to the AI Agent process - Supported Systems: Linux, Windows - Deployment: Zero configuration required, completes the first full scan within 120 seconds after launch - Open Source License: AGPLv3 - GitHub Repository: zizetu/existential-identity-test-engine - Current Version: v0.1.6

• 
  MarMar Labs (@MarMarLabs) reported 28 minutes ago

  Better agent tools can make the agent worse. GitHub just documented it in Copilot code review. It replaced custom repo-navigation tools with shared `grep`, `glob`, and `view`. Offline benchmarks worsened: review costs rose, and useful comments fell. The fix wasn't a new model. It was a job-shaped tool contract: 1. Anchor on the diff. 2. Turn the change into a specific review question. 3. Narrow candidates with search. 4. Read the smallest useful code range. 5. Stop when the evidence answers the question. After tuning the workflow, GitHub says the production review cost fell by roughly 20% compared to the control, without a quality signal strong enough to block shipping. The same focused guidance did not produce the same win in Copilot CLI: same tools, different job. Builder takeaway: tool access is not agent design. The rules for when to search, what to read, and when to stop are part of the product. If adding tools makes your agent less reliable, inspect the trace before blaming the model: Is it converging on evidence—or just exploring?

• 
  Harman (@itsharmanjot) reported 31 minutes ago

  Toyota had a single access key sitting in a public GitHub repo. Nobody caught it for years. By the time it was found in 2022, customer data belonging to hundreds of thousands of people had been exposed the entire time. That’s not a hypothetical. That’s one hardcoded secret, forgotten in a repo, doing quiet damage for years. It’s called Infisical, and it exists because “just put it in a .env file” is how almost every credential leak starts. → Centralizes every API key, secret, and cert across dev, staging, and ****, with full versioning and point-in-time rollback → Scans 140+ secret types across your files, directories, and entire *** history, the same kind of scanning that catches leaks like Toyota’s before they sit exposed for years → Agent Vault brokers your AI agents’ access to external APIs: the agent only ever sees a placeholder, the real secret gets injected at a proxy layer it never touches, so a prompt-injected agent can’t leak what it was never given → Honey tokens plant decoy credentials next to your real ones, so the second an attacker touches a fake key, your team gets an alert instead of a breach report → Full audit trail on every credential your team and your AI tools use, plus a private PKI to issue and manage certificates without a third-party CA GitGuardian tracked over 28 million new secrets leaked on public GitHub in 2025 alone. Most companies still find out the same way Toyota did: too late, by accident, years after the fact. MIT License (core). 12,700+ GitHub stars. Self-host free, unlimited users.

• 
  Atul Mishra (@The_AtulMishra) reported 32 minutes ago

  The "Revolutionary" Playbook : Step 1: Choose your model. Step 2: Choose the model usage tier (because the base tier is essentially a very confident autocorrect). Step 3: Add your skills (which the context window will conveniently ignore five minutes later). Step 4: Add loops (to ensure you burn through maximum tokens in an infinite spiral of despair). Step 5: Build your custom harness (so you can feel like a real 10x engineer). Step 6: Slap the word "Agentic Workflow" on a basic script and act like you just cured gravity. Step 7: Gaslight the architecture with a 10,000-word system prompt just to get it to output standard JSON. And the grand finale: Now, pay us $5 to $20 per task. Oh, did something go wrong? Did the output completely derail? That sounds like a you problem. Just head over to our GitHub issues page, where our entire community of open-source sycophants is standing by to tell you that you just don't understand prompt engineering. There is absolutely nothing wrong with Claude. We have very powerful models. You just aren't holding it right.