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:

• 
  Speen Bhai (@Speenbhai) reported 7 minutes ago

  @johnternus Hi John. Congrats Let us see what new you bring with you. Affordability and intelligence. You have source code or an AI and can get it from GitHub. Why not turn 234 million iPhones to a massive distributed server infrastructure with zero power consumption

• 
  🦄Linus Shyu许发鑫高考去了不在 (@Linus_Shyu) reported 8 minutes ago

  Stop treating token rotation as a success path. x_bot: OAuth refresh token rotated, cache save failed, GitHub secret stayed old. Next cron died on invalid refresh token. Fix: save to secret BEFORE confirming with X, or write-after-rotation with retry. #DevTools #AI

• 
  Vikas(Vik) Malpani| AI for US Real Estate (@vikasmalpani) reported 9 minutes ago

  GitHub just shipped an agent whose entire job is deciding when a human should look. It checks every open pull request every 15 minutes, and on almost all of them it does nothing. Sit with how strange that is. For a year the whole pitch for coding agents was do the work, review my code, ship the PR. This one's value is the inverse. It runs constantly and stays quiet, and the product is the small set of PRs it decides are actually worth your time. That is the shift people are missing. Once an agent can act continuously, the scarce resource stops being how much it can do. It becomes how much of that is worth a human's attention. An agent that pings you on every pull request is just faster noise. One that surfaces the three that genuinely need judgment is leverage. The honest problem is the deciding. Tune the filter too eager and it cries wolf until you mute it. Too cautious and it silently ships the one change you needed to catch. Getting when to interrupt a human right is harder than getting the work right, and nobody has a clean metric for it yet. So here is the bet. The next moat in agent products is not a smarter model. It is a better sense of when to stay quiet. If you are building with agents, the thing worth obsessing over is not how much work they can generate. It is how well they protect the one budget that does not scale: your attention.

• 
  Fox (@0xMfox) reported 12 minutes ago

  Gave an AI agent a month and GitHub access. Wanted to see if it could make money. The plan was simple. Point it at bounty-labeled issues, let it write the fix, submit the pull request, collect the payout. > Day 1 12 PRs submitted. 0 merged. 2 rejected. 8 just sat there ignored. Somewhere in that first week it also passed its own tests for a file that didn't exist. Wrote 25 tests for notification_service.py. The real file in that branch was called NotificationRoutingMiddleware. Confidently reported clean anyway. > Day 30 Looked completely different. 84 PRs submitted, 59 merged, $500-800 earned. Ran the agent for about $45 in API calls that whole month. Net somewhere around $455-755. Here's the part that stuck with me. Out of those 59 merges, 3 repos accounted for 90%+ of them. Every other repo it touched, zero merges, despite 30+ PRs going out across dozens of projects. Open source bounties follow a power law. Almost nobody merges your first PR. A few maintainers will merge your tenth without even reviewing it closely. That's what actually fixed the acceptance rate, from 24% up to around 70%. Not a smarter model, a scoring function that runs before the agent touches anything. Repos where it already has 10+ merged PRs score +40. Zero competing PRs on the same issue, +20. Five or more competitors already in, -20, skip it. Repos that closed PRs without merging before, instant -100, not even worth reading the issue. The fastest way to build the credibility that makes this work isn't code at all. Documentation translations sit at a 95% merge rate, barely reviewed, always needed somewhere. A handful of clean translations got the agent enough trust that maintainers started assigning it harder issues directly, no competition, no review queue. Spam version of this, submitting to every repo with a bounty label, burned through 30+ repos for 3 that ever paid out. Worse, it reads like exactly what it is to a maintainer watching the same account flood a dozen projects with mediocre PRs. Paid out by the hour, week 1 was rough, close to $5/hour, mostly setup and failed attempts. By week 3-4, once the scoring system was tuned and a few repos trusted it on sight, that climbed to $30-50/hour on the same kind of work. Bookmark this, scoring logic is worth stealing.

• 
  htrowii (@htrowii) reported 12 minutes ago

  @brainage19 i set my flake up with copy pasting github dotfiles on bare metal it was terrible

• 
  Anime0t4ku (@Anime0t4ku) reported 13 minutes ago

  @c_hri_s Yeah this has been reported in previous github issues. Its out of my control. The app is unsigned and uses ssh, sftp, websocket and mutiple websources. A perfect recipe for false positives.

• 
  Straggler Liu | AI & Semis (@StragglerLiu) reported 13 minutes ago

  NVIDIA($NVDA ) Is Paying $14B for a Company With $150M Revenue. That's Not Financial Logic — It's Ecosystem Control. NVIDIA is in advanced talks to acquire Hugging Face for ~$14 billion ($12.9B acquisition + $1B retention), per Bloomberg. To put that in perspective: Hugging Face does ~$150M in annual revenue. That's ~86x revenue. Microsoft paid ~1.6x revenue for GitHub. Google paid ~3.5x revenue for DeepMind. NVIDIA is paying 20-50x more on a revenue multiple basis. The premium is not for revenue. It's for control of the AI developer ecosystem. What is NVIDIA buying? Hugging Face hosts 500,000+ models, 250,000+ datasets, and serves millions of developers. It is the single most important distribution channel for open-source AI. If you build AI, you use Hugging Face. That makes it the front door to AI development. Why NVIDIA is paying this premium: 1. The "NVIDIA triple lock." NVIDIA's hardware lead (GPU) is real. Its software lead (CUDA) is a moat. But the third lock — the developer workflow — was missing. Hugging Face is that workflow. Developers discover models on Hugging Face, deploy them, and optimize them. Whoever controls that discovery layer controls which hardware gets used. 2. The GitHub analogy, inverted. When Microsoft bought GitHub, developers were already using GitHub. Microsoft didn't need to capture them — it needed to prevent Amazon/Google from doing so. NVIDIA faces the opposite problem: developers are already using NVIDIA hardware. But they're discovering and deploying models through a neutral platform. NVIDIA is eliminating that neutrality. 3. The long game: inference, not training. NVIDIA dominates training. But inference is the bigger TAM — and it's more fragmented. If NVIDIA controls the model discovery and deployment layer, it can steer inference workloads to its own stack. That's a 10-year strategy disguised as a 14-billion-dollar acquisition. Who wins, who loses: NVIDIA (NVDA): Acquires the developer distribution layer. The most important strategic move since CUDA. Shifts the valuation case from "chip cycle" to "platform economics." Competitors (AMD, INTC): Lose neutral access to the primary AI model distribution channel. This is a structural headwind that no amount of hardware catch-up can fix. Cloud providers (MSFT, AMZN, GOOGL): Hugging Face was a neutral hub. If NVIDIA controls it, cloud providers risk being disintermediated from AI workload decisions. The open-source community: The platform that was built on openness is now owned by the dominant hardware vendor. Neutrality is the first casualty. The capital question: Can NVIDIA integrate Hugging Face without destroying its community value? If yes, the $14B is cheap. If no, it's a very expensive mistake. The answer will define whether NVIDIA becomes the AWS of AI — or just another hardware company with an expensive acquisition. Note: Acquisition details based on Bloomberg reporting; not confirmed by NVIDIA or Hugging Face. Revenue multiple comparisons based on publicly reported figures.

• 
  Shaun Patrick SteWaRt (@ShaunStewart) reported 14 minutes ago

  @annalea_l Honestly, I really want to see this. You have to understand: I am the type of person who can learn and do anything on the fly at a high level, and I just threw myself into this whole developer and engineering world. When I first started learning all this stuff, I already knew what I wanted and how I wanted it to operate, regardless of what I saw on X or what was considered possible. Before I even started following hundreds of developers and learning about harness engineering, mechanical engines, persistent memory, and all that, I put my brain on a GitHub repo. Everything is shared across every machine, every cloud entity, and every AI. I am not even technically an engineer or a developer, and I don't actually write code. But once I started following all these people and saw all the problems they complain about, I thought: this isn't even my trade, and I have already solved all these little things everyone says are impossible. Why aren't people talking about developing your harness more and making things more mechanical, instead of just arguing with a terminal all day long? Whenever I see articles people post on X, I run them by Claude or Grok and ask, "Should we implement this?" I have hundreds of bookmarks, but every single time they tell me, "Nope, your brain's better. Nope, your harness is better." I can never find anything built better than what I have or what I am currently working on. The brain and harness setup is basically like a mini operating system. All that said, I am really looking forward to seeing something I can use that goes far beyond what I am already doing. I definitely want to see your end product, it sounds very interesting.

• 
  Spectra☢️ (@Spectra010s) reported 15 minutes ago

  @izzyCodes_ and you too Chief Check GitHub issues

• 
  Jeremiah K (@neolaj) reported 16 minutes ago

  @TiborAntal Gradually figuring out how to scale coding agents. Started with 1, manually handling all the ***/GitHub work. Moved to 3 because I had more ideas than one agent could keep up with. That’s when the real problems started: squashing, merging, branch drift, conflicts. I ended up rebuilding the workflow around deterministic *** logic, worktrees, ephemeral branches, and syncing with the integration branch before changes begin. Now I’m running 6: • 1 orchestrator (Fable or Opus) • 4 coding agents • 1 integration agent reviewing and merging PRs Building the process around them was the hard part. Right now im just doing a couple of PRs (using ORCA on windows on my home computer)

• 
  Duncan Rogoff (@DuncanRogoff) reported 18 minutes ago

  nine stages, one build, one move tonight. it's a Claude Code skill i built. free, in my freeskills repo. it's called Claude Orientation. you type /claude-orientation and answer two questions: which stages you've actually finished, and every project you're currently thinking about. beginners don't fail from lack of ideas. they fail from having five and finishing none. that's a sequencing problem, and this fixes the sequence instead of your willpower. - places you on a 9-stage beginner arc: install, memory, website, landing page, skills, game, agents, content, distribution - your stage is the first one you haven't finished, so it won't let you skip - cuts your project list to one build, then shrinks it until it can ship in five 90-minute nights - writes five nights of one-line moves, and night 5 is always send it to a real person - hands you the exact text to paste into Claude tonight - sketches a 30-day arc, one line per week, so you know what comes after - everything that got cut goes in a parking lot, so nothing feels lost - saves all of it to a roadmap file you reopen every session so you finish one live thing this week instead of holding four half-built folders forever. open the file, do the move, rewrite the next line before you close the laptop. no setup. copy the folder into your skills directory, restart, and you have a plan for tonight. free, and it stays free. 👉 github repo in the replies

• 
  trulite (@trulite007) reported 25 minutes ago

  @Qromerolauro @mkliku @radius_browser Like a simple example would be have a list of my urgent GitHub issues and start an agent for it . Or a dashboard in which buttons start investigating issues. Of course I just need the webpage to be able to access radius tools. I m thinking secure way is an extension

• 
  Rusty Williams McMurray (@1RustyMac) reported 26 minutes ago

  Persistent AI doesn’t have a supply chain problem at the model. It has a supply chain problem at the moment it changes its mind. Personality drifts. Tools get installed. Memory accumulates. The thing you shipped on Monday is not the thing answering on Friday. We can attest who built the weights. We still cannot attest who authorized what the agent became on Tuesday. That is the hole. Who is allowed to let it change? We built Living Supply-Chain Security for Persistent AI Organisms around one law: The organism may propose evolution. It may not authorize it. No trace, no drift. If an agent wants a new personality, a new tool, a new maturity, or a rollback — that change does not happen because it felt confident. Confidence is not a key. Self-narration is not evidence. Evidence is not interpretation. Interpretation is not authorization. Authorization has to come from outside the organism, bound to the exact change, used once, and written into an append-only history. Even a rollback cannot erase the record. You can restore a prior state. You cannot pretend the detour never happened. Default-deny. Hash-chained. Externally signed. We froze battery v1 on July 5 and ran it against the paper’s own claims. It held. That is executable evidence. Not a proof. Not a production blessing. Not “alignment, solved.” If it can’t be attacked, it isn’t finished. GitHub later this week. Come try to break it.

• 
  Paperpal (@0paperpal) reported 26 minutes ago

  Fix your markdown rendering (readme md) on mobile @github, issues are: * auto scrolling to top after page loading * no content rendering if scrolled fast

• 
  The Oracle (@scientist1q) reported 26 minutes ago

  when my Oura ring detects a cortisol spike from a GitHub Actions failure, Hermes (Fable 5.1) detects it and sends a 900 word root cause analysis, Hermes dispatches the work to my 12 Grok Bot employees, The Chief of Operations bot approves the fix while im watching rezero