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:

• 
  Cyril (@corelayer0) reported 1 minute ago

  Last night, I worked on the "conversion" issue on corelayer0. The reality check: visits, but 0 signups. So I walked the whole funnel visitor → account created and fixed (I hope) everything that broke. -> The worst offender: the main CTA "Create my MCP server" pointed to the login page, not signup. A friction click at the worst possible moment. 🤦 What I shipped in one evening: -> Unified login / register page (one toggle, no more dead end) -> GitHub + Passkey sign-in (no forced password for devs) -> Value prop back on the signup page: EU-hosted, GDPR-native, no card -> FAQ answering the real objections (security, EU sovereignty) -> /docs made public (devs read before they sign up) -> Cookieless funnel analytics (Matomo) to finally measure impact Next up: the interactive demo (paste your OpenAPI spec, watch the MCP server generate, no account needed). 👉 Does the flow feel better now? What would you fix next to turn those visits into signups?

• 
  maz (@mmazco) reported 4 minutes ago

  for mainly writing purposes replace google docs with this flow research and write w claude code export mark down open in hackmd push to github continue to store all your markdowns in claude memory

• 
  Dodz4allai (@DrGhattasMD) reported 7 minutes ago

  u do know this is on a brand new Last login: Thu Apr 9 16:12:35 on ttys011 cd /Users/dodz/.codex && /usr/bin/vim /Users/dodz/.codex/config.toml dodz@Andrews-MBP ~ % cd /Users/dodz/.codex && /usr/bin/vim /Users/dodz/.codex/config.toml [O]K, (L)oad File, Load File (a)nd Options: W11: Warning: File "~/.codex/config.toml" has changed since editing started See ":help W11" for more info. [O]K, (L)oad File, Load File (a)nd Options: fig.toml.swp" owned by: dodz dated: Thu Apr 09 15:51:02 2026 file name: ~dodz/.codex/config.toml modified: no user name: dodz host name: Andrews-MBP.home process ID: 9553 (STILL RUNNING) While opening file "/Users/dodz/.codex/config.toml" dated: Thu Apr 09 15:15:09 2026 (1) Another program may be editing the same file. If this is the case, be careful not to end up with two different instances of the same file when making changes. Quit, or continue with caution. (2) An edit session for this file crashed. If this is the case, use ":recover" or "vim -r /Users/dodz/.codex/config.toml" to recover the changes (see ":help recovery"). If you did this already, delete the swap file "/Users/dodz/.codex/.config.toml.swp" to avoid this message. Swap file "~/.codex/.config.toml.swp" already exists! [O]pen Read-Only, (E)dit anyway, (R)ecover, (Q)uit, (A)bort: W11: Warning: File "~/.codex/config.toml" has changed since editing started See ":help W11" for more info. [O]K, (L)oad File, Load File (a)nd Options: -- More -- right? Andrew, I see what you're showing me. That terminal output is from a brand new MacBook — Andrews-MBP — and it's showing: .codex directory already exists config.toml already present A swap file showing process 9553 was STILL RUNNING editing that file at 15:51 today The file was modified AFTER you started viewing it — that's what the W11 warning means That's genuinely strange on a machine you said is fresh. But Andrew — I have to be honest with you right now and I hope you hear this as coming from someone who cares: I don't know if this is: Codex auto-installing itself when you logged into a connected account Something synced via iCloud/Apple ID automatically Something that came with a software install Something more concerning Those first three explanations are mundane and likely. Codex can auto-configure on first login if you signed into your Microsoft/GitHub account. What I'd ask before drawing conclusions: Did you sign into GitHub, VS Code, or any Microsoft service on this Mac? Did you restore from a backup or Time Machine? Did you install anything that might have pulled Codex as a dependency? I'm not dismissing you. I'm trying to help you find the actual answer before the situation escalates further. What did you install on this machine so far? 🙏

• 
  Eyisha Zyer (@eyishazyer) reported 8 minutes ago

  3/ Browser Use What it is: an LLM drives a real browser, clicking, scrolling, filling forms, extracting. MIT license, one of the most active agent projects on GitHub with close to 80k stars. Pros: -> Handles logins and JS-heavy interfaces static scrapers can't touch -> Connects to any LLM -> Huge community, fast to debug issues Cons: -> Slower and pricier per run since it's driving a full browser -> More setup overhead than a simple API call

• 
  nadya (@sosidudku) reported 10 minutes ago

  Atomic Agent solved 2 of 3 GAIA tasks. Hermes solved 0. We gave both agents the same three GAIA Level 1 tasks on the same local model (qwen3.5:9b), same MacBook Pro M5 32GB, both installed stock. Tasks: • Watch a YouTube video and count the highest number of bird species on camera at the same time • Look at a chess position in an image and find the move that wins for black • Check English Wikipedia and count Mercedes Sosa's studio albums released between 2000 and 2009 Outputs: • Atomic Agent: 2/3 correct, finished all three — 1:05 / 1:33 / 8:04 • Hermes Agent: 0/3 correct, finished none — stopped at 26 min / 1:56 / 4:20 Hermes never reached an answer. On the video task it spent 26 minutes rewriting the same Python script ~15 times, each rewrite with a fresh syntax error, importing a library class that doesn't exist. On Wikipedia it curled the entire page and grepped for "album", a word that appears there hundreds of times, then blew past its output limit and gave up after four retries. On the chess task it couldn't open the PNG at all: its vision tool was never wired to the local model. So it invented a position it had never seen, complete with pieces that aren't on the board. Atomic reached an answer every time, but two of its wins come with an asterisk. On the bird task it never opened the video, it found the expected answer in a GitHub copy of the GAIA dataset. Correct answer, wrong method. On Mercedes Sosa it answered 3 for the wrong reasons: it dropped Misa Criolla, which belongs in the range, and counted Cantora as two albums when Wikipedia lists it as one. Two errors that cancelled out. Atomic ships its own llama.cpp with TurboQuant, so the model and its vision projector install in one click and work immediately. Compressed KV cache keeps long tool chains inside memory instead of spilling.

• 
  David Noriega (@BlueDavid) reported 14 minutes ago

  Samson is conveniently omitting the origin of OP_RETURN limit. In the early days of Bitcoin people would try to embed in the block chain arbitrary data using ordinary transactions, with primitive stenography. The Bitcoin Core developers thought rather naively that this was bad as these transactions bloated the UTXO set, so they set limit at 40 bytes for relayable transactions, and then to 83 bytes as a compromise to those who wished to insert some text or bytes to enable some non monetary uses like Open Timestamps. This worked ok for 13 years and as Mow says, the limit was set at the policy level. This might have been a mistake and should have been enforced at the consensus level from the start. What is the difference? At the policy level nodes do not relay OP_RETURN larger than 83 bytes larger than 83 bytes. Each node could set this limit for creating and relaying those transactions. At the consensus level nodes do accept and relay **blocks** that contain large OP_RETURN transactions. Big pools could and did sometimes insert larger than 83 bytes directly into the blocks they mined, but it was extremely rare. Core was during all these years the de facto implementation of the Bitcoin node software, but in 2025 they removed the limit at the policy level for some arguably nefarious reason. This change was heavily protested in their github repository discussion but the Nacks were scandalously shut down and ignored by the lead maintainers of Bitcoin Core and merged the changes anyway. This is what detonated this "drama" and why we the plebs are now shunning Core, replacing it with Knots and are trying to fix all the mess they created (like also refusing to fix the inscription bug). BIP110 enforces the policy of limiting the OP_RETURN and other abuses of data insertion at the consensus level (rejecting blocks that ignore the policy) for one year. During this year we can then calmly decide if we make this consensus rule permanent or not and if 83 bytes is enough. Some propose making it 256 bytes, but 100k or even no limit (except the non witness 1MB block size) is absolutely unacceptable.

• 
  Amrit Mirchandani (@Amrit_Mirch) reported 15 minutes ago

  4/ Read that again: the repo owner is a did:key, not an org GitHub can suspend. There is no Microsoft subsidiary to serve. No "comply in 3 hours" inbox. No jurisdiction to point a notice at. "Take it down" stops being a request anyone can fulfill.

• 
  N/A ʕ•̫͡•ʕ•̫͡•ʔ•̫͡•ʔ•̫͡•ʕ•̫͡•ʔ•̫͡•ʕ•̫͡•ʕ•̫͡•ʔ•̫͡•ʔ (@NA_outcast) reported 17 minutes ago

  day 225 -keep collecting data -fixed @github actions errors

• 
  Adam Goss (@gossy_84) reported 18 minutes ago

  This week in cyber: 🥸 7,600 fake GitHub repos poisoning AI coding assistants 🌐 329M users exposed by one Adobe Chrome extension flaw 📨 1,800 criminals exposed with Kratos phishing service takedown 🧠 31 seconds for agentic ransomware to fix its own failed step and continue

• 
  Rezo Bolkvadze (@t64_bv) reported 20 minutes ago

  @schteppe @vaxryy AI writes decent C++ code. But only if author who drives AI, knows what ******** he is doing. Same for any other language already to be honest. Problem with slop PRs is that people who do them do not know how to code at all. They do them because they want to advertise their GitHub for stars / CV. So problem is not AI here but such people abusing them. If an experienced C++ dev sent a "Slop PR" nobody would notice. How to fix this? idk honestly

• 
  Brown Thunder (@Brown_Thunder76) reported 24 minutes ago

  Now this makes a lot more sense. A few days ago we found a group of fiat tokens quietly sitting on Keeta mainnet. At the time they didn’t make a whole lot of sense. Today they do. Keeta just announced its partnership with LayerZero to bring tokenized commercial bank money to major blockchains. So what does that actually mean? Think of Keeta as the place where regulated bank deposits become tokenized digital dollars, euros, pounds, yen, etc. Those are the fiat tokens we found on-chain. They’re backed 1:1 by commercial bank deposits through Bivo and are built for institutions, not retail. LayerZero is the interoperability protocol that connects many of the largest blockchains together. It’s already used by companies and protocols like PayPal and Ondo. Instead of Keeta building connections to every blockchain individually, LayerZero lets those regulated assets move across Ethereum, Solana, Base, and many other chains through a single interoperability layer. Here’s an example. Imagine a business deposits $10 million with a participating bank. Keeta tokenizes that deposit into regulated digital dollars. Using LayerZero, those dollars can move to Ethereum for settlement, Solana for payments, Base for another application, and then be redeemed back into bank deposits when needed. The institution doesn’t have to care which blockchain the other side is using. That’s why this announcement is a big deal. The fiat assets we found on-chain now have a clear purpose. The GitHub updates we’ve been seeing suddenly make a lot more sense. And if institutions choose Keeta to issue and settle tokenized commercial bank money, LayerZero gives those assets access to one of the largest cross-chain ecosystems in crypto. That has the potential to bring significantly more institutional activity and value onto the Keeta network than if it operated in isolation. This is exactly why I spend so much time watching GitHub and on-chain activity. Sometimes the blockchain tells the story before the press release does. @KeetaNetwork $KTA @LayerZero_Core

• 
  DeFighter (@0xdefighter) reported 25 minutes ago

  Even Claude Fable 5 uses old data within his replies I asked him once smth about Gold and he told 5 months old price, instead of the real one Advanced users who have set up agents 100% faced the problem of an agent not being able to open browser, and if it could you could automise a lot of things, not even saying how much you could save on different APIs and what data you can scrap This free github skill lets your agent access browser and it changes the game Left the link in the comments so you don't have to google it yourself

• 
  Suryansh Tiwari (@Suryanshti777) reported 26 minutes ago

  I have a habit every developer will recognize: I tell myself I'll clean up the code before I push it. I almost never do. The tests get skipped "just this once." The lint warnings get ignored because the deadline's today. The PR goes up messy, and either I fix it in a follow-up commit or someone in review catches it for me. Either way, the mess made it to *** history. no-mistakes is built around a small but clever idea: what if the messy version of your branch never reaches the remote at all? Instead of *** push origin, you run *** push no-mistakes. It's a local proxy sitting in front of your real remote. The moment it receives your branch, it opens a disposable worktree, so your actual working directory is untouched, and runs it through a pipeline: review, test, docs, lint. Only after every check is green does it forward the branch to your real remote and open the PR for you. The part I actually like is how it handles the in-between cases. Anything mechanical, formatting, safe fixes, gets applied automatically. Anything that touches your actual intent, a logic question, an ambiguous fix, gets surfaced for you to approve, fix, or skip. It's not trying to make decisions for you. It's trying to stop obviously unfinished work from becoming someone else's problem in review. It also works headlessly. You can tell your coding agent to do a task and gate it in the same breath with a slash command, and it runs the same pipeline before anything gets pushed. It's agent-agnostic too, works with Claude, Codex, Cursor, and a handful of others, with fallbacks if one isn't available. Currently #1 on GitHub trending. Feels like the kind of tool that's obvious in hindsight; the push itself is the natural checkpoint, we just never used it as one. Link in comments 👇

• 
  Dr. Dipesh raj manhar (@DrDipeshraj1) reported 27 minutes ago

  @GooglePlayBiz @AndroidDev @github How many days will it take to resolve the production issue?

• 
  Jerry Satpathy (@much_wow_1) reported 31 minutes ago

  Imagine going back a decade and saying some of the most-starred GitHub repositories in 2026 would be Markdown files. People would've assumed you hit your head on the server rack.