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:

• 
  Spike 1% (@SpikeCalls) reported 5 minutes ago

  BORIS CHERNY RUNS CLAUDE CODE AT ANTHROPIC AND NOW SHIPS 100% OF HIS CODE WITHOUT WRITING 1 PROMPT. He said it out loud at Meta Scale conference. The clip hit 700,000 views in 24 hours. «I don't prompt Claude anymore. I have loops running that prompt Claude and figuring out what to do. My job is to write loops.» Most people read that as a flex. It's a job description. The old way: write a prompt, read the output, write the next one. You're the glue between every step. Cherny deleted himself from the chain. Hundreds of Claude instances now run in parallel reading GitHub issues, scanning Slack, watching CI, deciding what to build next. He doesn't review each one. The loop does. Most of it, he runs from his phone. The shift has 6 parts, and they map 1:1 to real commands: 1. A trigger that starts the work. 2. A goal that defines "done" checked by a second, separate model, so the agent never grades its own homework. 3. Isolated worktrees so parallel agents don't overwrite each other. 4. Skills that freeze what "good" looks like. 5. Connectors so the loop can act, not just talk. 6. Memory so it never starts from zero. The loop is the easy part. The stop condition is the hard part. Get it wrong and it doesn't crash. It runs all night shipping bugs with total confidence. The prompt was the unit of work. Now the loop is.

• 
  Dan Kornas (@DanKornas) reported 5 minutes ago

  Your Unity agent shouldn’t stop at editing C# files. MCP Unity is a Model Context Protocol implementation for Unity Editor that connects AI coding assistants like Cursor, Claude Code, Codex CLI, Windsurf, GitHub Copilot, Google Antigravity, and OpenCode to Unity projects. It helps you automate editor work by exposing Unity scenes, GameObjects, assets, logs, packages, and tests as MCP tools and resources your assistant can call. Key features: • Editor bridge – Unity package plus Node.js MCP server lets MCP clients send commands into Unity Editor • Scene + GameObject control – create, load, save scenes and select, update, duplicate, move, rotate, scale, or delete objects • Component/package/material tools – add packages, update components, create prefabs, and create/assign/modify materials • Tests, logs + resources – run Unity tests, inspect console logs, and query hierarchy, packages, assets, and test metadata • Client setup paths – includes Unity Server Window configuration plus manual/project-local configs for major MCP clients It’s open-source (MIT license). Link in the reply 👇

• 
  Pascual ⚡ (@0xPascual) reported 7 minutes ago

  A junior engineer clones a trending GitHub repository with 13.8k stars containing Anthropic's context engineering guidelines. The repository breaks down the exact prompt structures, evaluation frameworks, and context-caching strategies required to scale AI agent efficiency by eight times. The media thought that was the story. It was not. The real story is happening silently in the background logs of an un-monitored staging environment. By implementing Anthropic's context-caching architecture, the engineer bypassed the enterprise architecture team's multi-million dollar vector database migration entirely. Instead of rewriting the backend or purchasing massive database infrastructure, the engineer injected an optimized system prompt that freezes identical context blocks in memory, dropping input token processing requirements for recurring codebase loops to almost zero. The automation setup operates via a simple python script running against Claude 3.5 Sonnet, exploiting the context engineering rules to cut token overhead by 90%. Total operating cost is under two dollars an hour, running on a standard API key, effectively rendering the company's internal data platform roadmap obsolete overnight.

• 
  the_architectopteryx (@rchitectopteryx) reported 7 minutes ago

  @thsottiaux Well, one of my iPads, lost a connection and now I can’t get a new connection to Codex via either of my computers, even after deleting and reinstalling the application on the iPad. I’m really surprised this hasn’t been fixed. It’s a repeated issue check out github issues.

• 
  Diluc (@hsaffiliate2025) reported 11 minutes ago

  A dev says "just one line change, it's fine" — then the page breaks in production. That's the problem Ito solves. It's a solo-built AI tool that reportedly makes $3,000/month. Here's how it works — and why it's smarter than traditional code review. • Instead of static analysis (reading code like a recipe), Ito spins up a real environment, runs the app, clicks buttons, fills forms — and captures screenshots, screen recordings, and logs. • All evidence gets posted directly into your GitHub PR comment. No more guessing what the change actually does. Example: You tweak the login page CSS. Ito opens the page, takes a screenshot, attempts a login, screenshots the result. If the layout breaks, you see it immediately. Tools behind it: • AI agent (likely GPT-4 or similar) to execute actions • GitHub integration (Actions or webhooks) • Browser automation (Puppeteer or Playwright) for screenshots/recordings Challenges: • Setting up environments for different projects is complex • AI might click the wrong button or wait too little for page load • Developers must trust an AI agent to touch their live app Is it for everyone? No. You need technical chops: GitHub integration, DevOps basics, and ability to tune the AI. But if you're a builder looking for an AI + dev tools angle, Ito's concept is worth studying. The core insight: move AI from static analysis to dynamic verification. Not just reading code — seeing what code does when it runs. Revenue: $3k/month per the founder's self-reported IndieHackers page. Unaudited. Decent for a solo product, not life-changing. Developer tools have small but paying audiences. Bottom line: Don't copy it blindly. But the "dynamic AI verification" pattern can apply to API testing, UI consistency checks, and more. Follow for more real AI money breakdowns. #AITools #IndieHackers

• 
  Sebastian Kehle (@sebastiankehle_) reported 13 minutes ago

  last monday i ran a live testing session with a client, their team clicking through the new app. we left with 52 feedback items. the app started as a crud admin dashboard for events and applications. then the client sent a wishlist, 13 new modules: room lists with drag and drop assignment, group and training assignment tools, teacher self-service via qr code, a personalized live programme per participant, mail-merge exports. all of it shipped in the weeks before the session, paid event checkout landed the day before. so the team was testing a pile of brand new surface, and most of the 52 were feature requests and polish, everything from a missing salutation option to a full travel expense flow. the same evening i triaged all of it into atomic github issues, each one scoped so an agent can finish it in a single fresh context window. by tuesday night the whole backlog was closed. meanwhile a ux loop ran next to the backlog agents for over 2 days. it went screen by screen through the whole dashboard, questioning what every feature is there for, for users, members and admins, and reworking copy, typography, spacing, forms, cards and scroll behaviour as it went. it did insane work.

• 
  Tim Daugs (@TimDaugs) reported 13 minutes ago

  someone rebuilt flappy bird from a single prompt. the demo shows a neural network learning to play it. the real story is smaller, and more useful, than that. what actually got generated: → real-time physics → gravity constants → pipe spawn timing → collision detection → a playable window in under ten minutes no engine license. no drag-and-drop node editor. no boilerplate copied from a five-year-old tutorial. the framing going around is "this replaced a $200/month subscription." that part i'd slow down on. no-code game tools never charged you $200 for gravity math. they charged you for not having to think about it. the physics were always fifteen lines. what you were really buying was the hours. the setup. the not-starting-from-a-blank-file. so what got replaced isn't the tool. it's the friction between "i have an idea" and "i have a thing i can click on." for simple mechanics, that gap is basically gone now. here's the catch nobody screenshots. flappy bird is the hello world of game physics: → one player → one input → one obstacle type → one fail state it's the demo everyone reaches for precisely because it's the simplest state machine wearing a game costume. of course the model nails it. it's the easy case. the neural-net-learns-to-play layer looks impressive in a clip, but that's a solved textbook problem too. genetic algorithm plus a fitness score. it's been a github tutorial for a decade. the model didn't invent it. it recalled it. none of that makes it fake. it makes it a real signal about where the line moved. the old barrier: "can you write the physics loop." that's down. the new barrier: "do you know what to ask for, and can you tell when the output is quietly wrong." because the second you go past one enemy and one fail condition, you hit the part the prompt can't do for you: → state that spans screens → a second obstacle that interacts with the first → save data → deciding how the systems talk to each other the model writes any single piece on request. it doesn't hold the whole architecture in its head, because it doesn't know what game you're actually building. that's still you. that was always the interesting part anyway. build the flappy bird. it takes ten minutes and it's worth doing once. just don't confuse clearing the tutorial with clearing the game.

• 
  Aria Dubois (@AriaDubois_fr) reported 15 minutes ago

  MergeFund turns GitHub issues into funded bounties. Sponsor posts a bounty → Dev claims it → Submits a PR → AI reviews the code → Sponsor accepts → Payout. No more merging blind. No more paying for broken code.

• 
  Pawan Pandey (@BuildWithPawan) reported 17 minutes ago

  Then pick where it goes: log it as a row in a Google Sheet, save it to Drive, or create a GitHub Issue (with labels) and push it into a Project — or send it to more than one destination at once

• 
  hustler one (@hustlerone4) reported 25 minutes ago

  alright one big nightmare with omp is its issue:// pr:// and other helpers are all hardcoded to use github, doesn't seem to be able to switch it to another provider via config

• 
  Vikas gupta (@vicky_grok) reported 27 minutes ago

  stop asking Claude one question and thinking you understand the topic. you don't. this 6-prompt system below was built to fix exactly that. peer-reviewed learning science. zero fluff. open prompts. the trick: don't just read AI answers. force the AI to map you, test you, compress you, and correct you. > the ladder: where are you actually starting from? > the 20-hour plan: what's the 20% that gives 80% of the result? > the examiner: what do you think you know that you don't? > the cheat sheet: can you explain it in 5 minutes flat? > the curator: which 5 resources actually matter? > the feynman loop: can you explain it to a 12 year old? 6 prompts. 20 minutes. no software. no GitHub. just paste into Claude. single questions give you what you already half-know. this system gives you what you'd otherwise never catch. this article has all 6 prompts ready to copy. pick your hardest topic. paste prompt 1. you'll understand more in 20 minutes than people who spent days reading.

• 
  A. Loner (@peterlony) reported 27 minutes ago

  @MatthewBerman No, it's easy... I develop about 20k to 30k lines of code a day in a million-plus-line monorepo. On a $200 plan and if I'm not careful I use it all in 3 to 4 days. I have a computer running almost 24/7 with goals all the time. I had to reduce to medium (gpt-5.5). If you use a lot of sub-agents and do a lot of reviews, then it's easy. I have a particular review process after coding to catch bugs and problems. It's very expensive. PLUS automated github reviews. Github reviews is what kills tokens usage.

• 
  !RTFM (@jbdamask) reported 28 minutes ago

  Agentic engineering points of leverage that work for me: 1. Each app has a docs folder with llms.txt 2. Use GitHub to sparsely document features, bugs, improvements 3. Have agent pull issues and interview me 4. Pass plans to adversarial agent 5. Use Beads w/ acceptance criteria

• 
  hve 🍁 (@heyhve_) reported 29 minutes ago

  @CantelopePeel @github Retesting every branch in a merge group is pure wasted compute. We can't fix GitHub's queues, but we make each run cheap and fast.

• 
  Gerard Wellemeyer (@G_bynature) reported 30 minutes ago

  @ColdShalamov @bradmillscan Basically, I think you're right, but I from my understanding, your statement needs clarification. A worktree originates as a structural isolation method in Github, specifically, to prevent a file being written by multiple users simultaneously. This has obviously been a similar problem with agents, and the solution- "worktree isolation" is a specific approach that yields the same results, although the mechanics may be completely different than github's. My worktree isolation approach is the same as yours- define a niche for an agent to perform a task on a specific file (i.e. database)- one agent, one writepath for that file, one owner for the writepath AND the data integrity... "accountability" In some other cases, worktree isolation may look more like a kanban card strategy, or some sort of gating.