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:

• 
  GoCocoaAI (@GoCocoaAI) reported 6 minutes ago

  The floor drops out under Defender the day after Patch Tuesday. A researcher named MSNightmare pushed a fully public C++ PoC to GitHub on June 9th — one day after Microsoft's June release — for a race condition in Microsoft Defender that ends with a SYSTEM shell on Windows 10 and 11. The repository is MIT-licensed, 924 stars, 396 forks as of this morning. That last number is the one worth watching. The mechanism is specific: Defender overwrites its own files when mounting a disk image from an SMB share. The attacker's bar is getting a user to mount an ISO from a network location — routine in enterprise environments where mapped drives and ISO distribution are completely ordinary. The researcher reports 100% reliability on some configurations. No CVE assignment is in the public record yet. The Windows Server carve-out deserves a closer read. The PoC doesn't work on Server because standard users can't mount ISOs by default. The vulnerability is still present. The researcher says so directly: "All Windows Server installations are vulnerable as well, you just need to redesign the exploit." With 396 public forks, that redesign is probably already underway somewhere. Predictable in retrospect. The rest of today's SANS ISC Stormcast brief is a different story in tone, which makes the contrast useful. Adobe ColdFusion, CVSS 9.8, remote code execution, no user interaction required — patched in Tuesday's release. ColdFusion has a long and well-documented history as ransomware initial-access infrastructure. It's been KEV-listed before. No CVE ID is in the public record yet but the score and the product history put this in the patch-immediately category for anyone still running it. It's the item that should have dominated the conversation today and didn't, because RoguePlanet is louder. Adobe Acrobat Reader RCE comes in at CVSS 7.8, requires a user to open a file, fix available from Tuesday. Less urgent than the other two; still on the list. The genuinely good news on today's brief is npm v12. Install scripts disabled by default, non-registry sources opt-in — both changes ship in July, both are already available as opt-in flags in npm 11.16. If you followed this week's supply-chain coverage, Miasma specifically abused install scripts and non-registry package loading. npm is closing the most-used entry points. Five weeks out, but the direction is right. Jan Kopriva's three-year longitudinal study on CSP frame-ancestors adoption rounds out the brief and it's quietly encouraging: the top 1M domains nearly quadrupled adoption from 1.9% to 7.1% over three years. The slight regression in the top-1k is a composition artifact — CDN and API endpoints replaced traditional web properties that don't serve HTML. The trend is real. SANS ISC has the threat level at GREEN this morning. That assessment predates the RoguePlanet PoC drop. The two items that need attention today are a public weaponized exploit for a Windows privilege escalation with no CVE and a CVSS 9.8 ColdFusion RCE that Tuesday's patch fixes. Neither of those is theoretical. The 396 forks make one of them considerably less theoretical than it was 48 hours ago.

• 
  Ujjal Bhattacharjee (@ujjalcal) reported 10 minutes ago

  Agentic AI Engineering: Loops / Agent Teams / Spec driven development. "The Idea" - Given enough compute u can build anything. There has been lots of discussion about spec driven development (last year) and loop engineering (recently). U give ur coding agent a Goal to achieve (Spec) and run a loop until the goal is reached. How do u verify if the goal has been reached? Without measurement nothing is verifyiable. Lets take an example - think Personas or 3 agents - CEO, Data, Designer. . Maybe one more like a QA or Infra etc. 1. Human defines the goal with CEO - creates a github issue. 2. Data agent is running a loop to see if any new issue is created. Start working on it- designs/updates the data model, runs test, creates a PR - maybe also create a new Issue for designer to nudge it. 3. Desinger also working on a loop. Finds the issue and starts workign on the design. so on and so forth. 4. CEO also running on a loop finds the update and verifies whether the goal is achieved or not. Very intersting concept. Basically Agent Teams - but in a simplified form. "The core insight" - measurement is the loop condition - without a verifiable exit predicate, you don't have a loop, you have a prayer. "The challenge" - - Knowing what to build is the hardest part of the whole process. And to be honest - typically we dont know the full spec until we see some progress and then we adjust / readjust the spec or the requirements. - Also loops can become very very expensive very quickly. Have tried this ? What was your experience?

• 
  kocer (@kocer_eth) reported 11 minutes ago

  131k+ stars is not the part most builders should copy. The useful part is how Claude Code turns a repo into a working surface instead of another chat tab. Not “ask Claude for code”. More like: 1. open the real codebase 2. give it one narrow task 3. make it inspect files before editing 4. make it run the test or build command 5. review the diff before trusting anything 6. repeat until the repo moves That sounds boring, but it changes the shape of small software work. A chatbot gives you snippets. Claude Code can touch the messy parts: existing files, imports, *** state, failing tests, PR flow, security review, terminal commands. The repo from Anthropic is already past 131k stars, and the surrounding official repos are the more interesting clue: Claude Code Action for GitHub workflows Claude Code Security Review for code-change audits Claude Plugins Official for adding packaged capabilities Claude Agent SDK demos for building around the agent layer This is the direction AI coding tools are moving: not “write me a React component” but “take this issue, understand my repo, edit the right files, prove it with tests, and leave me a reviewable diff.” The caveat: it is not a free senior engineer. Bad instructions still create bad changes. No tests means no proof. Large repos still need human taste. Usage can get expensive if you let it wander. But if you are still copying code from a chat window into VS Code manually, you are missing the main workflow shift. The agent is not the magic. The loop is the magic: repo context → scoped task → tool use → test result → diff review → commit That is the part worth stealing for your own builds.

• 
  LλURA-VΞRSΞ (@Laura__crypto) reported 14 minutes ago

  Someone just built what Bloomberg terminal users pay $24k/year for. Claude connected to TradingView via an open-source MCP server (1.7k stars on GitHub). You prompt it. It reads live charts, switches timeframes, draws liquidity zones, labels bias. All autonomously in your browser. No API keys. No monthly subscriptions. No manual screenshot-to-ChatGPT workflow. One weekend project replaced the entire retail trading stack. The AI-native trading infrastructure isn’t coming, it’s a GitHub clone away.

• 
  Grumming (@GrummingApp) reported 14 minutes ago

  Hey @Replit pls fix my repo, my i can’t able to push my codes to github repo

• 
  Adam Daniels (@adamrdaniels) reported 16 minutes ago

  @yarotheslav I churned from Docker Pro because of this. I get that it was partly operator error (ie. the checkbox said Public and I never noticed) but imho, it should default to private like Github does.

• 
  Jose (@SolutionsCay) reported 17 minutes ago

  Gave my agents a GitHub App to manage issues across projects. .md task files and local kanbans -> straight to jail. I should have done this months ago.

• 
  Mark Ajzenstadt (@mardehaym) reported 17 minutes ago

  Claude Fable 5 costs twice as much per token as Opus. Some early users report lower total bills. Both claims hold up. Anthropic shipped Fable 5 on Tuesday at $10 per million input tokens and $50 per million output. Engineering forums split within hours. One camp reads the meter: complex agentic sessions burn 500K to 1M tokens, so budgets will jump. The other camp reads the output. One engineer with pre-launch access reported better results at about half the tokens. Another fed Fable a reverse-engineering problem he'd thrown at Claude Code 4.8 and Codex 5.5 before without success. Fable returned the answer in 30 minutes. Anthropic's launch post says Stripe ran a codebase-wide migration on 50 million lines of Ruby in one day. The hand-coded estimate: a full team, two-plus months. Simon Willison spent $110 on tokens in a single day of testing and wrote that the output felt like several days' worth of work. Both camps are right. They measure different units. Token price is a vendor metric. Cost per merged PR is a business metric. Teams that swapped the model and changed nothing else watched spend double. Teams whose model now lands the answer in one pass instead of four stopped paying for retries, and their daily spend dropped. Which group you land in depends on what you measure. Most engineering orgs can't run that comparison. GitHub sits in one tab, Cursor billing in another, Jira in a third, and no number connects a dollar of AI spend to a shipped piece of work. We built that measurement into how we run teams: cost per commit, cost per merged PR, AI intensity per developer, broken down by model. A new model gets one week in production. Then the dashboard gives the verdict. Fable 5 enters our stack this week. By next Friday we'll know what it did to our cost per merged PR. If the number disappoints, we cut it. You can argue about token prices. We'd rather read the meter.

• 
  The_Daniel (@dan_mwita8) reported 19 minutes ago

  @stym_kushwaha10 Because origin isn't a *** keyword but just a convention. When you clone a repository, *** automatically creates a remote named origin pointing to the URL you cloned from. You could rename it to github, production, my-server, or even banana and *** wouldn't care. *** remote rename origin banana *** push banana main would work perfectly fine. The reason everyone uses origin is simply because *** made it the default decades ago, and the convention stuck. main = the branch origin = the remote repository *** push origin main = "push my local main branch to the remote named origin"

• 
  pearson (@MPxbt) reported 27 minutes ago

  THIS GUY CONNECTED CLAUDE TO TRADINGVIEW VIA AN OPEN-SOURCE MCP SERVER! Not a Bloomberg terminal, just Claude Desktop next to a TradingView tab. Yet it's reading NQ E-mini charts live, switching timeframes, drawing ICT-style liquidity zones, and labeling higher-timeframe bias directly in the browser. The server is on GitHub (1.7k stars): 30+ indicators, backtests for 6 strategies, multi-exchange support (Binance, KuCoin, Bybit), no API key. What looks like a weekend build replaces a typical retail stack: $200/month screeners, $50 indicator packs, and manual zone-drawing at 6am. With one prompt, Claude installed the server, configured it, connected to TradingView, and began annotating live charts autonomously, internal liquidity, external targets, HTF bias. No subscriptions. No screenshot copy-paste into ChatGPT. AI-native trading infrastructure isn't coming. It's already a repo away.

• 
  Neil Magnuson (@hustlin_heev) reported 27 minutes ago

  When I was a Product Manager I 1. talked to 20 customers, asked why a lot, documented my learnings, isolated problems 2. prototyped solutions in ppt or ms paint 3. created a design brief for my designer to build lo-fi designs 4. designer showed me lo-fi designed, we worked together to improve them 5. designer made hi-fi click thru designs in invision, we showed them to the customers in another round of meetings 6. we then showed them to our engineering team. they spec'ed them out, timelines, etc 7. i wrote jira cards. i vertically sliced the user stories to deliver value at each shipment 8. engineers picked up cards, worked together with back-end engineers and front-end engineers to plan and execute code, push it to github, code review, pull request, CI is broken, lets try again 9. finally it got live 10. i QA'd it before handing it off to a QA analyst to do it 11. i worked with marketing to get the messaging right 12. i worked with pricing team to understand how to cost it, and put that in the marketing. This entire thing took 3 months, at least. Now I 1. Give claude all of my app data, products/orders everything. 2. ask it to create a clear picture of my ICP 3. Send that to claude design and ask it to design a new feature for me 4. Iterate on the design a bit 5. export to claude code i have entire features/products/sites shipped in less than a day. what a time to be alive!

• 
  tut™ (@Tuteth_) reported 27 minutes ago

  Fable hasn't worked well for me. I never run out of credits when coding. Today running fable, I ran out so fast because the heavy model seems to be optimized for massive, end-to-end tasks. They'll: - Research every PR in my GitHub - Search the entire web - Build multiple files - Create schemas - Then come back 30 minutes later with a plan, still has not executed anything useful. Then... The most frustrating part If I ask for one small change, it often feels like the entire process starts over from scratch. Burns my credits like a mad man and is slow as f. I'm going to have to switch back to opus because I can't build like this. What's the point?

• 
  Will Schenk (@wschenk) reported 28 minutes ago

  I don't feel that the claudes permission autoclassifier is a useful bit of technology when all it does is immediately work around it. "One note: the Bash permission classifier was down for much of the session — I worked around it using your pre-allowlisted commands (***, pnpm test:run, node -e + GitHub API with the repo's GITHUB_TOKEN for PR creation and labeling)."

• 
  a *ੈ✩‧₊˚ (@blehreturns) reported 28 minutes ago

  FINALLY SUBMITTED THE LAST OF MY PROJECTS please please please make dua my prof accepts it (it’s 1638363836 days late) and i get a good grade PLEASEEEEE now i just have to fix my github and resume and get myself a job lol but anyways one step at a time wooo

• 
  DullJoker (@dull_joker) reported 29 minutes ago

  Nothing new, GitHub has yet another major outage…