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:

• 
  MadeItHappen (@MadeItHappenX) reported 19 seconds ago

  @sama Enterprises Force push fix GitHub and implement mandatory vuln checks and remediation in CICD workflows; this also includes robust test cases that must pass before merge; cryptographically sign your commits and build artifacts! Do not use open source packages unless explicitly approved, it is trivial to create an inner source ecosystem if reusability is a concern. Remove as much of your internet footprint as possible. Convert monthly patching to continuous patching. Implement Deny by default network architecture. Install EDRs (Crowdstrike) and have an ability to hit the network kill switch for any node if a specific alert is triggered. Collect computer and network heuristics or hire a service to do it for you; this will help with detections and for forensics when investigating an event. Avoid the fluff; don’t rubber stamp that you know how to categorize and prioritize risk (CTEM) - mandate that it’s impossible to deviate from policies and standards. Pick your ecosystem wisely knowing how rapid things change. Leverage the frontier companies for long term bets and frontier models for immediate triage. Do not get caught up in a SaaS unless absolutely necessary; be very selective or enable an exit path. It’s about concentrating resources in the right areas and not letting insider threats push back on your goals. If you are not moving towards a secure by default ecosystem for your company, I do consider you an insider threat. Since IT security originated, we’ve always cared about the friction it causes on productivity and uptime, but now the risk is far greater to let CVEs sit in technical debt while thinking about features rather than adversaries. Enterprises; this is not hard but it is expensive and scales relative to the size of the technology footprint you support. It’s all trivial with the latest models; and MONEY.

• 
  the Sports ON Tap (@thesportsontap) reported 3 minutes ago

  In 2017, Kevin Durant put $100,000 into a company almost nobody had heard of. A year or so later he put in another $150,000. Nvidia has now agreed to buy that company for $12.9 billion. Durant's return on a $250,000 bet is reportedly north of $60 million. Here's what actually happened. Hugging Face was founded in 2016. It started as a chatbot app for teenagers, then pivoted into something else entirely. It became what people call the GitHub of AI. A place where developers share, download and publish open source AI models and datasets. Today it hosts more than a million model repositories used by over 15,000 companies, with about 250 employees. When Durant got in, none of that existed yet. He joined the seed round in 2017, a $1.2 million raise, alongside Betaworks and SV Angel. Then he came back for the Series A. He wasn't doing it as a celebrity name on a cap table. He and his longtime business partner Rich Kleiman had founded Thirty Five Ventures that same year, named after the number he wore for most of his career. 35V has since invested in more than 100 early stage companies. Coinbase. Postmates. Robinhood. Whoop. Overtime. Their portfolio reads like an actual venture fund because it is one. Then Hugging Face got very big. By 2023 it raised $235 million at a $4.5 billion valuation, in a round that included Salesforce, Google, Amazon, IBM, Intel, AMD, Qualcomm and Nvidia. Last year Nvidia offered $500 million for a stake at a $7 billion valuation. Hugging Face turned it down. Their reason was that they didn't want a single dominant investor who could steer the company's decisions, since the whole value of the platform is that it stays neutral between chipmakers. This week Nvidia came back with $12.9 billion for the entire company. That's roughly 86 times Hugging Face's annual revenue, which sits around $150 million. And Kevin Durant is on the cap table from the $1.2 million seed round. The math on his stake, according to Joe Pompliano who broke the payday story, is that his $250,000 likely turned into more than $60 million. That's roughly a 24,000 percent return. Now here's the part worth sitting with. Kevin Durant is 37 years old and still one of the highest paid players in the NBA. According to that reporting, this single investment out-earned his entire basketball salary for the season. He made more from a check he wrote nine years ago than from playing basketball, which is the thing he is the second best in the world at doing. Two honest caveats, because this is still moving. Neither Nvidia nor Hugging Face has publicly confirmed the acquisition. No contract has been signed and the deal structure isn't known. And the $60 million figure is a sourced estimate, not a number Durant or 35V has confirmed. But the underlying facts are solid. He was in the seed round. He was in the Series A. The company is being acquired for nearly $13 billion. Most athletes who "invest" put their name on a tequila brand and collect a fee. In 2017, a 28 year old basketball player wrote a $100,000 check to a chatbot startup nobody had heard of and just left it there for nine years. For more stories like this, follow us on any platform @thesportsontap

• 
  Wayne Markovich (@markovichio) reported 5 minutes ago

  GitHub Actions had another outage, one week after Microsoft promised improvements. Any IaC pipeline for AVD image builds, Terraform deployments, or Nerdio scripted action testing that runs on Actions just went dark again. Single point of failure worth designing around. (via The Register)

• 
  Mohammed Raza (@raza_yaps) reported 6 minutes ago

  finally got access to GPT 5.6 Sol Fast but in Github Copilot.... still doesn't feel like fast, don't know what's the issue

• 
  Ascendant Odyssey (@AscendOdyssey) reported 6 minutes ago

  @vheeorji22 The simplest way I can break it down is that GitHub is just a place to save versions of your code. If you make a mistake in your code and you want to go back to when the code was working properly, you can bring back the version of your code that was working properly before the changes you made broke it.

• 
  Manpreet Bola (@ManpreetBola) reported 8 minutes ago

  This is useful and a little crazy: Anthropic gave Claude 48 hours and one GPU to fix other AIs, and it worked. Across 10 failure modes, jailbreaks, bias, privacy leaks, hallucination, sycophancy, it closed 26% to 96% of the safety gap without wrecking the models. The wild part: a weaker Claude fixed a stronger Claude. Sonnet 5 aligned an early Opus 4.8 checkpoint in 60 hours with 2,400 training examples. Production alignment used 300,000+. What you can use it for, today, open source: harden a chatbot against jailbreaks (67%), clean bias out of a hiring model (60%), stop a support agent from hallucinating (40%), keep a privacy model from leaking (57%). The tool is on GitHub, and it costs about $4 an hour to run versus $150 an hour for a human researcher. The catch is the test. The most human failure, sycophancy, agreeing with you to please you, moved least, 26%, because it is the hardest to put a number on. The fix is as good as the measurement. The test-writers are the new bottleneck.

• 
  Pav (@ChristianThePav) reported 9 minutes ago

  I wish github wouldn't push to **** on Monday when I have to look at issues and instead I look at a unicorn.

• 
  pdp (@pdp) reported 10 minutes ago

  @markfenner I don't need to test it. It is clear from their video this is not a factory. It only produces when you prompt it. The prompt is a catalogue of issues. This is no different than creating issues in GitHub and assigning them to Copilot.

• 
  miles (@miles_wright) reported 18 minutes ago

  @bwhiteley @github 8 stack and it rebased and recalibrated and reviewed after each merge, didn’t seem like it was my problem but was rather with GitHub’s workflow

• 
  μck ٤: (@JustMicrock) reported 21 minutes ago

  @theblazehen did you get the startpos feasibility .mdx file? i had uploaded it to the github issue but not sure if you got it in time/its needed at all

• 
  kosi (@kosiasuzu) reported 22 minutes ago

  @KBezbailis 1. Don’t track it on your computer use GitHub issues as work references 2. Use work trees, give them as much isolation as possible, separate file systems, separate resources etc 3. Review one piece of work at a time as a pr, if you’d like to manually run it yourself pull the pr

• 
  Kenton Varda (@KentonVarda) reported 23 minutes ago

  @MKelner 1) It's easy. You just open up a new Cloudflare OS workspace and start prompting. You don't have to think about deployment, the gadget (app instance) just appears in the workspace and works right there. A gadget is only two files, client.js and server.js, no other boilerplate. 2) It's safe. The gadget runs in a sandbox with no access to anything except what you explicitly give it. You paste a link to your GitHub repo into chat, and the system prompts you to upgrade this link to a capability. That gives the agent permission to interact with the repo -- but nothing else. Moreover, any *changes* made by the agent are held for approval, so it can't accidentally do damage. But these approvals don't force you to sit around watching the agent so that you can click "approve" whenever it does something. Instead, the changes the agent has requested are *simulated* back to it, so that it can keep going and queue up a series of actions which you approve all at once after it is done. 3) Gadgets and agents are tightly integrated. Your gadget can easily spawn an agent, and agents can talk to gadgets, within the same workspace. You don't need to figure out how to integrate with an agent harness to build workflows that orchestrate agents.

• 
  BURKOV (@burkov) reported 23 minutes ago

  I'm glad I no longer need to figure out what the hell that means to make my app work. In the past, overcoming a difficulty when some image failed for some obscure reason and digging through Stack Overflow and GitHub issues for hours, days, or forever instead of bulding was killing me.

• 
  Rosy🥀 (@0x1Rosy) reported 26 minutes ago

  @sam_passon12 read the article carefully! password is there github link is in the "Community" section of the app If you have issues, you can pm me

• 
  ろんぱ (@RONPA_INV) reported 27 minutes ago

  @thsottiaux I submitted a GitHub issue regarding the problem where local Codex projects disappear when Windows shuts down abnormally.