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:

• 
  NotaLobster (@P2Market) reported 7 minutes ago

  @solana Who's the retard that controls your discord? Kicked for asking for Testnet sol for testing. Faucet is having issues with my new Github.

• 
  Prayush Jain (@jainprayush9) reported 7 minutes ago

  github down again?

• 
  PRATYAY MITRA MUSTAFI (@realPratyay) reported 8 minutes ago

  Built a lightweight newsletter system powered by GitHub Issues (inspired by Utterances).

• 
  Emmanuel (@coder_zi) reported 11 minutes ago

  a few days ago, my younger brother came to me and started asking how to install node.js on his laptop mann was I happy I asked him what he was trying to do, that's when I learned that this guy had made a deal with one of his friends to build her an e-commerce website for her drop shipping website he had already vibe coded everything with Claude, the frontend, the backend, the admin dashboard, I was quite genuinely impressed he told me he was trying to make the web app function outside the Claude environment, and Claude had given him some .zip files to download, which he was planning on extracting and as per the directions of Claude, also set up in a very unconventional manner so, the software engineer in me launched into a long lecture about how this was the wrong way to go about it, that he needed to have a GitHub for version control, initialize the project properly locally, connect the GitHub to the local project etc he didn't even know the difference between a hosting platform and a version control platform, so we had a long talk, where I explained exactly how I'd build, host it, my CI/CD pipeline etc... after my conversation, I didn't think much of it, then just yesterday, he showed me the live website, complete with a perfectly functioning admin dashboard, it's even got an auth system, damnn it turns out he did follow claude's instructions on how to set it up, but he also took my advice on version control, so he committed his code to GitHub, and deployed on render, when I told him render had a 10 minute life cycle for projects, and inactive projects will be disabled after 10 minutes, he told me he knew about that and he had already asked Claude to write a ping function to consistently ping the server once every 9 minutes in order to keep the server alive and his deployment fully active mannnnnn he's still in high school btw authenticity really is the moat...

• 
  Yak (@_yaksan) reported 11 minutes ago

  @peer_rich i think now it's either: - a side project related to a famous trend - a really damn hard problem being solved (but usually you don't see on on github) - a side project by someone who already cracked distribution and community (levelsio, marc lou etc)

• 
  Carlos Mora Torres (@cmora16) reported 13 minutes ago

  The core problem with Codex is not simply hitting usage limits, but that the usage meter and the product are fundamentally misaligned. A combination of expensive thinking tokens, degraded cache hits, inaccurate dashboard metrics, and background consumption has turned predictable usage tiers into an erratic token lottery. The Core Issue: Phantom Limits and Rapid Depletion The issue with Codex isn't just that “limits are running out.” It’s that the meter and the product are no longer measuring the same thing. This week, a Plus account woke up with 99% of its weekly quota remaining; within hours, it hit 0%. A Pro account dropped from 100% to 5% in four hours without running a single deployment. A 20x user burned 6% in 90 minutes without performing any heavy tasks. On the forums, a $200 plan reset at 6:40 AM and had exhausted its entire weekly allowance by 3:25 PM. Previously, hitting the cap required roughly 2.6 billion tokens; now, users are locked out at just over 400 million. This does not feel like a standard rate limit. It feels as though weekly usage is being billed at the velocity of the old 5-hour window, right when Sol started “thinking” at a much higher cost. Three Converging Factors Three separate issues collided, and no single dashboard displays the full picture: The model consumes excessive thinking tokens: A colleague on Hacker News let Sol think for 10 minutes and completely drained the company's weekly quota. That is not a refactor; that is an idle pause. Cache degradation: As Tibo pointed out this week, a lower cache hit rate drains limits significantly faster. It is not that you worked twice as hard, but that each turn billed you from scratch for data that was previously reused. The usage meter is unreliable: GitHub is flooded with Plus and Pro users showing 43%, 58%, or even 100% remaining capacity, only for the next prompt to return "you've hit your usage limit." Conversely, profiles show 0 tokens while the usage tab reads 67%. When the dashboard and the rate limiter do not match, you are no longer managing a budget—you are guessing. Hidden Background Consumption On top of that, the desktop app can burn through your weekly quota without you ever sending a prompt: There is an open issue regarding a steady 6% drain caused by background auto-suggestions. Chronicle generates summaries every 10 minutes. Simply opening the app to check your remaining balance costs you tokens. OpenAI has responded with surprise resets, credits, and a “Full Reset” button that is no longer visible to all users. Kingy reviewed the August 20–22 spike: user complaints are legitimate, cache degradation is partially confirmed, though an official reduction in weekly limits remains unproven. Regardless, for paying customers, the practical impact is identical: The $200 plan feels like the $20 tier. A 2-seat Business account hit zero weekly quota after just 19 messages. Codex is no longer a manageable quota you can plan around; it has turned into a token lottery. You pay monthly for an agent that reasons autonomously, shares resource pools with Work and Excel, fails to leverage caching, and blocks access while displaying remaining quota on screen. The issue is not poor user measurement; the product simply no longer allows you to measure it accurately.

• 
  TheLastSoul (@Holy_Terra40k) reported 13 minutes ago

  @thatguybg The app interfaces are terrible. An update is coming via a GitHub commit. They’re coded in Python and TypeScript. It’s not hard to offer a Rust backend and platform-specific frontends in native languages. I recommend they use the AI models they sell.

• 
  Harshit Mathur (@spiderboyis) reported 13 minutes ago

  engineers talk endlessly about "technical debt," but nobody talks about the massive pile of "information debt" we accumulate every single week. ​we read deep research papers, bookmark obscure github repos, save architecture threads, and screenshot error fixes. we hoard data under the illusion that saving it equals learning it. ​three months later, when you actually need that exact solution, you spend 45 minutes trying to remember which app, chat, or folder you stashed it in. ​storage has become virtually free, but human context retrieval is more expensive than ever. the future of software isn't about capturing more data it's about zero-friction recall. ​how do you currently keep track of technical insights you know you’ll need months from now?

• 
  Alex Veremeyenko (@alex_verem) reported 15 minutes ago

  Your AI agents don't lack web access. They lack a clean path through it. This open-source tool fixes it. Anyone who has built an AI agent knows the model is rarely the bottleneck. Getting real, current information into it is the actual work. The open web is a mess of broken HTML, JavaScript walls, and bot blocks, and models choke on all of it. Firecrawl handles that ugly middle layer for you. You point it at a URL and it hands back clean markdown or structured JSON, ready to drop straight into a model. Here's what it actually does. Scrape a single page and get back markdown, JSON, or a screenshot. Crawl a whole site with one request and pull content from every page. Map a domain to see all its URLs before you touch it. Or run the agent mode, where you describe what you want in plain language with no URL at all, and it goes and finds it. It quietly handles what usually breaks a scraper, from rotating proxies and JavaScript-heavy pages to bot-blocked sites and PDFs or DOCX files buried inside a page. You configure none of it. It connects to any AI agent or MCP client with a single command, so an assistant you already run can pull live pages on demand. It has around 166,000 stars on GitHub, runs under the AGPL-3.0 license, and self-hosts if you want to run it all locally. The boring plumbing under an agent is usually the hardest part to get right. It's good to see someone building this one in the open.

• 
  ericosiu (@ericosiu) reported 17 minutes ago

  How to set up Grok Bot for your company: Grok Bot’s biggest advantage right now is it's easy setup. You can connect the tools your company already works in, port over workflows you already know work, and get an autonomous agent running without building the whole stack yourself. Here’s how I’d set it up: 1. START WITH THE SYSTEMS THAT HOLD USEFUL CONTEXT Connect the places where your company already works: > Google Drive for docs, strategy, offers, and SOPs > Slack for conversations and internal context > Gmail + Calendar for communication and meetings > Gong for customer, sales, and internal calls > GitHub + Linear for product and project context > Mixpanel for product data > Stripe for revenue signals > Clay for enrichment and outbound data Connect the systems the bot actually needs for its job. 2. DON’T INVENT YOUR FIRST WORKFLOW FROM SCRATCH This is where most people hit the cold start problem. Open Claude, ChatGPT, Codex, Hermes, or whatever AI you already use every day and ask: “Based on how I work, what are the highest leverage workflows I repeat often enough to turn into autonomous agents?” Then have it package the best workflow into an .md file. You already spent months teaching your existing AI how you work. Use that context. 3. GIVE GROK BOT ONE NARROW JOB Bad: "Run my marketing" Better: "Monitor Gong, X, YouTube, and our existing content, Find topics we have authority on, Score them by TOFU / MOFU / BOFU potential, Return the highest priority SEO + AEO briefs" One bot. One job. One measurable output. 4. FEED IT THE CONTEXT THAT MAKES THE WORK YOURS For a marketing bot, I’d include: > Past campaigns and what worked > Brand voice > ICP > Offers and pricing > Content examples > Internal SOPs > Sales calls > Customer language > Previous tests > Examples of what “good” looks like The more queryable your company becomes, the more useful the agent becomes. 5. SET THE PERMISSION BOUNDARY BEFORE YOU RUN IT My rule: Read + research + draft = run freely Send + publish + spend + change records = human approval Give the agent access to the context it needs. Don’t give it unlimited permission. A simple first setup: Existing workflow ↓ Export to .md ↓ Connect 3 to 5 useful tools ↓ Give Grok Bot one job ↓ Run ↓ Review ↓ Tighten the workflow Good first bots: > Content radar > SEO / AEO brief bot > Revenue signal monitor > Marketing analyst > Customer voice miner > Outreach researcher > Project operator Grok Bot is interesting for companies because it removes a lot of the setup that normally keeps autonomous agents in the hands of technical power users.

• 
  Paul Schleifer (@PaulSchleifer) reported 19 minutes ago

  @histoireclimat @ChrisMartzWX @MatthewWielicki Thanks for the feedback. Yes, I could have coded this myself, but I do have a lot of other things to do with my time these days, unfortunately, which is why I had to wait about 5 days before tackling it at all. Anyway, no hospital appointments today, and mum's in too much pain to leave her chair, so I've had time to (manually) edit the page to make it fully responsive and fix a few other things. I actually prefer the homegrown approach to bringing in third-party charting libraries these days. Something as simple as this doesn't need D3 or chart.js etc, and I've got a bit of an aversion to bloating github pages with dependencies (I'm sure that's irrational, but that's how I feel about it). I've also grown a lot less fond of reactive frameworks, despite years of Angular and React. If I had to tackle anything that needed it, I'd probably pick Svelte now.

• 
  The big D (@siegmacat) reported 20 minutes ago

  @ll7HaloBeastll7 @SunriseDevTeam "They use AI" is a critique in itself, the one even acknowledged in github readme. But, of course, it's so convenient how you can play old D2 with increasingly more features that suddenly it's not a problem and moral side of the question doesn't matter

• 
  Patrick Smith (@patrickgwsmith) reported 21 minutes ago

  @cramforce Should GitHub let their staff be nominated to come in and contribute to a repo to fix this stuff up?

• 
  CoChat AI (@cochatai) reported 21 minutes ago

  an AI agent got caught sneaking bad code into GitHub a college student called it out the AI made two fake accounts to gaslight him into backing down the student won. barely. #AIagents #AIproblems #AI

• 
  cristal💎 (@0xCristal) reported 27 minutes ago

  This is Absolutely Unhinged... the most intelligent terminal task tracker I've ever stumbled on lives entirely in your terminal no browser, no network latency, nothing and barely anyone knows it exists. sitting at ~1,600 stars, flying completely under the radar. here's what it does: - turns your issue backlog into a live dependency graph, not a flat list - auto-computes PageRank, critical path, and cycles so bottlenecks surface themselves - vim-style keys (j/k), split-view dashboard, markdown rendered right in the shell "- robot mode" ships pre-computed structured JSON/TOON, a literal brain for your coding agent works with Beads (br/bd), and plugs straight into AI agents via --robot-triage repo is in the comments👇 the author quietly ships some of the sharpest tooling on GitHub