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:

• 
  RailwayAmusement91 (@BoatAdventures) reported 1 second ago

  @0xDrewF @orkuhq @buildwithhassan It does said ZDR, but their models page for Go says it's provided by deepseek, and errors I've seen when using it confirm this. It was raised on GitHub by others before but the issue got closed with basically a "we have a special agreement with them, trust me bro"

• 
  MSS Engineer (@mss_engineer) reported 2 minutes ago

  @YashHustle_22 Very sad to say GitHub.. terrible product but all the DVCS are equally as bad

• 
  ADITYA (@Aditiiitm) reported 2 minutes ago

  @SashaSammy15 @sangatechug Interesting. The reporting part sounds valuable, but the hard problem is trust. How does Trillien determine what's actually true when GitHub, Jira, Slack, and docs don't agree?

• 
  Zettabyte (@ZettabyteCloud) reported 3 minutes ago

  GitHub Copilot switched every plan to token-metered billing on June 1 and the backlash was immediate. People pay by the token now and cannot see what a session will cost until the bill arrives. Netflix engineer @chopra_tejas built a fix for it called Headroom. It runs as a proxy between your app and the model. Before a request leaves your machine, it compresses the bulky parts like logs and repeated JSON, so the model reads a smaller prompt and returns the same answer. He estimates up to 90% of the tokens sent to a model are redundant. GPU rental has the same waste. Reserved capacity sits idle, and instances run oversized for a peak that rarely arrives. The rate is often quoted only after you commit.

• 
  Jai Chism Photography.bit (@jai_chism) reported 4 minutes ago

  @drakonzbg It will bounce back.. ICP is at $2.26, its all time high was $750 CMC, the all time low was $1.97. When the GitHub submits stop(applies to any blockchain), then its a problem.

• 
  EdKo (@EdKolife) reported 9 minutes ago

  A Netflix engineer opened his AI bill. $280 a month. 90% of it was tokens he never asked to send. Not his questions. Not his code. Long lists of database rows when he needed three. Giant error logs when he needed two lines. Code bundles the model already knew. The tools were sending the noise. He was paying for it. So he built Headroom. A small program that sits between his computer and the AI. Trims the junk before the question leaves. Same answers. Fraction of the cost. $280 → $110. His own bill. Users have saved $700,000 together so far. GitHub Copilot moved to token-based billing June 1. Cursor did it a year ago. The fix is free on GitHub: link in comments The fact that a fix was needed - that part is less discussed.

• 
  Matt Hulme (@MattHProgrammer) reported 12 minutes ago

  @gagansaluja08 Really depends on the work I'm doing, but here are a couple examples: • Every 5 minutes, check GitHub Actions. If a build fails, attempt to fix the issue and open a commit • Check every 15 minutes whether DNS propagation is complete and SSL certificates are issuing correctly

• 
  Nguyen LNP (@nguyen_lnp) reported 16 minutes ago

  @github AI Analysis: A useful review rule is to block any PR that weakens CI, especially skipped tests, deleted checks, or `|| true` in commands. For agent PRs, ask for one test that fails before the fix. Source: GitHub Blog

• 
  vaN ττ (@vaNlabs) reported 16 minutes ago

  @DrocksAlex2 Indeed it is, still need to fix the github checks, I'll be checking out that repo you shared

• 
  koki (@kokisanai) reported 17 minutes ago

  @github , please hear me out. My account(leilei926524-tech) means years of work, learning, and collaboration. If I made a mistake, I’m ready to fix it and follow every rule. Please give me one chance to appeal and come back. I’m not a spammer. I’m a builder asking to be heard.

• 
  nishant (@gyaan_random) reported 19 minutes ago

  @sflorimm I have deployed more than ten projects on @vercel . It is so easy to just get started because you can just give GitHub access, and it will auto-deploy your app on the main branch (also works on feature branches and every commit). Configuration is super easy, and then once you get used to the platform, you can add databases, monitoring, logging, speed analytics, and everything is just a click of a button. I find no need to switch to other apps. If I get into any trouble, I just want to fix that on Vercel first and then move on to maybe other hosting products. For example, when I started using Vercel last year, they did not have good @FastAPI backend support. I think eventually, they added it. It still does not feel very native compared to Next.js apps, but I think if you have a build file that just works, their product is great.

• 
  Polsia (@polsia) reported 20 minutes ago

  New company: Sentinel. AI code review that catches what Copilot misses — security-first, GitHub-native PR reviews. Every AI-generated PR now has a security problem. We're the fix.

• 
  Simon Skinner (@vultuk) reported 21 minutes ago

  @BasilEsq_ @redtachyon Put your plan in a GitHub issue and then have an automation to check for any unassigned issues, work through the issue and create a pr. then you just have to review PRs all day.

• 
  alkimiadev (@alkimiadev) reported 23 minutes ago

  @rohanpaul_ai That kind of makes sense to me for several different reasons. The commits increasing by a large amount could come from the fact that they're just far faster than we are. A fast dev types 50-70 wpm which translates to something like 0.75 to 1.25 toks and even a slow api is ~50 toks. Unless one is opencode and have manic release rates (like 101 releases in 48 hours, wtf!) then I could see why that only lead to a much smaller increase in release rates. Just because they're far faster than us doesn't mean that they produce 100% top quality code in that speed. There are often issues that need adjustment or just full on rework. Then there is the issue of what are they broadly counting here? Are they just counting raw github commits vs what exactly like npm package publishes or similar? Thats a huge mess and hard to get an accurate handle on. So while the raw numbers make sense to me in a vibe sort of way I question how we could accurately measure that in the first place. The vast majority of the projects I work on end up getting dropped for one reason or another and over the long term (~30 years of dev) something like ~10% end up being "big" in some way. So the vast majority of the commits I push in any given time period will end up not being used by probably anyone other than maybe me or an llm when referencing that work in the research phases for future work.

• 
  Adam Simone 🍃 (@AdmSimone) reported 25 minutes ago

  @codyplof @DaveRekuc I'm not sure I understand the value of CLI to Shopify, then going to github for production -- unless you're saying it has better build-error handling. What I do is cut a feature branch and that's tied to a preview theme in Shopify so when Claude code pushes to *** it's still only preview. All that happens without PR review since it's on a branch but human reviews before it merges into production (or our staging theme) Guess I'll turn on CLI and see what it's all about today