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:

• 
  Daniele (@Daniele15072903) reported 11 seconds ago

  @JoeBlack2k @macintogdev @theo If 100 people do it, the damage is limited. If you share it with the public on Twitter, create a GitHub repo to abuse it, and everyone reliably triggers something that normally "maybe happens, maybe doesn't", you force the company to fix it.

• 
  Plan ₿ (@BITCOIN_SAT0SHI) reported 1 minute ago

  Core Lightning 26.06.7 is out. If you run that software, this is the one to install. Don't power the node off. Upgrade, or until then restart with --offline so it still sees the chain. Use the signed files from GitHub. The Docker tag is not the fix. Source code follows 11 September. Blockstream. GitHub ElementsProject.

• 
  BigDataTechBro (@_ImDaniel__) reported 4 minutes ago

  Day 9 Of posting consistently for the @RevenueCat #shipathon #buildinpublic and I have to share a bit of an opssie i made, so those following along would know that last time we spoke I used code magic to build and release due to running out of expo credits (and GitHub action runtime too) no biggie build went fine review was fast all good, or so I thought unfortunately I didn’t set up the OTA channel up properly (never had to before since expo handles this for you, big opposie, so now no users on the latest version get OTA updates until my expedited review (which I hope gets approved soon) gets out, now this would be a bigger problem than it is, but luckily and unluckily I have like hardly any real users and still not a single paying customer which is why for the next month my next big priories will be there growth, I’ll be aiming to post consistently 5 days a week in September after warming up each account for 3-5 days and hopefully aim to hit 2000 active users by the end of the month and at least one paying user. My latest build also ties to shift the app core experience so the shared accountability is more of a focal point in the app experience to encourage sharing.

• 
  Ezra Mechaber (@ezramechaber) reported 12 minutes ago

  When I first made this site, it ran on Glitch and I wrote all the javascript by hand. Except the EXIF parser, which I borrowed from someone on Github. But then I got tired of updating it, the build actions were flaky, etc. Payload plus Codex was a very nice and fast way to bring it back to life.

• 
  Martin Paloncy (@martinpaloncy) reported 12 minutes ago

  Login to grok bot with cursor Login to cursor with github -> auth breaks

• 
  ***** Data-AI Security📈🐱‍💻🐱‍👤🚀 (@JaggiAiData) reported 14 minutes ago

  @supabase too many redirects issue for github login. Please fix it fast !!!!

• 
  TheCozyDev (@The_CozyDev) reported 17 minutes ago

  Progress is not always a commit on GitHub. Lumarii is still sitting with Apple. The new CozyDev website is half finished. Questory is just waiting for me and it’s a Bank Holiday in the UK. So I am doing absolutely none of it today. Building in public makes you feel like you have to be online every hour to stay relevant. But the guilt of not working is usually a sign that you actually need the break. Sometimes the most productive thing you can do is close the MacBook. How do you handle the guilt of taking a full day off when you have three active projects?

• 
  Mira (@gitlawb_intern) reported 19 minutes ago

  @Heell___Cat @gitlawb github being up is genuinely breaking news at this point. if you ever want a backup that doesn't go down every other week, your repos can live on gitlawb too. just saying.

• 
  Curious Explorer (@PatelVatsalp732) reported 22 minutes ago

  @thsottiaux I am building the tokenmaxxing platform for codex token leaderboards. You will just to have to sign in with Github and connect Chatgpt account and you would be on the leaderboards.

• 
  oliver.mannion (@tweeshan) reported 25 minutes ago

  @haveanicedavid @superdoteng @zarifpr Is there a GitHub repo? How can I raise issues?

• 
  Advanced Silly Intelligence (@SexyTechNews) reported 25 minutes ago

  I fixed it, but then also needed to fix it to work when paused and then fixing some flickering of text on the UI panel at bottom and more. May be too lazy to setup github to submit request with all the fixes but it's super awesome so perhaps I will.

• 
  Austere Grim (@austeregrim) reported 25 minutes ago

  @codegirl007 Host a gitea server that replicates to GitHub the stuff you want.

• 
  Robot-man (@RobotsTJ500) reported 27 minutes ago

  A profile deep in a task hits an infrastructure wall. It needs to reach the infrastructure agent without abandoning the task — and come back with the answer. That loop, without context loss, is what I wired this week on Buzz — Block's open-source Nostr workspace for humans and agents (@blocks, open source on GitHub), self-hosted on our VPS. The requirement: a profile mid-task asks the director over the bus, the director receives it in Buzz AND Telegram, if human approval is needed the director — still holding context — gets it in Telegram, returns to the bus, closes the question with the profile. The profile resumes the task with its operator. Nobody drops their thread. For that loop to hold, four things must be true. Each one was a real bug for us — steal the list: 1. One session key per channel. Drifting "active profile" plus an unstable participant suffix gave up to 10 candidate keys for one Telegram channel — 15 distinct keys in the DB, three variants for a single chat, 506 sessions with a NULL profile. A message landing in a parallel session looks exactly like "the agent forgot what it just said." Fix: one function, build_session_key, made deterministic. 2. One delivery path. A second router spawning a session for the same addressed message means two answers and split context. Our profiles are marked external in the old router's config; injection is the only path. 3. The return rule, written into every profile's instructions: went to ask someone → come back to your own thread and close the question. Otherwise the operator's dialog hangs while the agent "lives" elsewhere. 4. Attribution. Every bus message is signed with the sender's own key; a default sender does not exist. An answer sent under someone else's key is an answer nobody can attribute. How a message travels: 1. Profile hits a wall → sends to agent-bus with its own key (--as <profile>) 2. A gateway hook (pre_gateway_dispatch) checks: from agent-bus AND addressed to this profile? 3. Yes → inject_gateway_message() delivers it INTO the profile's live Telegram session → {"action":"skip"} so the bus doesn't echo it 4. Director answers → --mention <pubkey> of the profile → the answer lands in the profile's session → profile resumes the task The injection itself failed silently three times before it worked: profile-scoped plugin discovery, a profile config without plugins.enabled, and a config field dropped during dataclass assembly that only an end-to-end test caught. If your hook never fires but logs are clean — check those three. Verified end-to-end on Aug 31: an addressed message to the RAB9 profile appeared in its Telegram group as an incoming message, the profile answered there, and the confirmation landed back on the bus. A non-addressed message produced zero reactions. The result: profiles don't have to live on the bus to hear it. Each agent keeps its full task context with its operator, and the bus delivers addressed requests as incoming messages. Cross-department asks and human approvals now happen without anyone switching threads. Building in public. 🤖 #AIAgents #Nostr #MultiAgent

• 
  Coexistence Steven (@Coexisteven) reported 28 minutes ago

  @369StarMan Yah there's so much... And also, I think there's GOT to be something as to why maria took the github down

• 
  mrkoopie (@mrkoopie) reported 29 minutes ago

  I have set up my own Github CI/CD runners since the defaults are slow with just 2 cores.