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:

• 
  CodeMonument (@CodeMonument) reported 59 seconds ago

  @vincent_koc How do you control what these instances all do? Aka. How do you steer them? Are they basically mostly subagents and you only say: work on GitHub issue x?

• 
  Helio (@helioim_ai) reported 4 minutes ago

  Now, you can @ an AI teammate in your task panel on any issue or bug. They analyze what's wrong and come back with implementation ideas, right where the work lives. > Reads the issue with your GitHub context, names the actual problem > Suggests an implementation, not just a diagnosis > All in the Helio, no context switch Real engineers don't just describe problems. They propose fixes. Your AI teammates now do the same.

• 
  Dapton AI (@daptonai) reported 5 minutes ago

  @linear Code review has always been the most context switch heavy part of the whole development cycle. Write in the IDE. Review in GitHub. Comment in Slack. Back to the IDE to fix. That is four tools for one feedback loop that should take minutes. Diffs inside Linear with AI review and agent iteration in the same place is not faster code review. It is finally the whole loop in one place.

• 
  ric..ki (@derricky_eth) reported 5 minutes ago

  If DAC wants to achieve long-term adoption, security can’t stay buried in a GitHub repo or treated as a “nice-to-have” technical layer. It has to become the ecosystem’s operating system—the unspoken contract between every participant. Here’s the uncomfortable truth: Most decentralized projects focus on throughput, fees, or novel consensus. But users don't leave because a chain is slow. They leave because they get hacked, rugged, or watched helplessly as a bug drains millions. Trust takes years to build and seconds to shatter. As on-chain infrastructure becomes more interconnected—bridges, oracles, cross-chain apps—a vulnerability in one protocol no longer just hurts that protocol. It cascades. Think of the major bridge hacks of the last few years: billions lost, not because the underlying L1 was broken, but because a single smart contract had a subtle flaw. That flaw then froze liquidity, tanked token prices, triggered liquidations, and eroded faith in entire ecosystems. Suddenly, validators, treasuries, and governance forums are all in crisis mode—reacting, not preventing. Why “future-proof” infrastructure isn't a buzzword—it’s survival A resilient DAC ecosystem doesn’t just patch today’s vulnerabilities. It anticipates tomorrow’s. That means designing for: Adaptive threat models – Attack vectors evolve faster than audit reports. Formal verification is good; real-time anomaly detection and invariant monitoring are better. DAC needs security that learns, not just checks boxes. Data integrity without blind trust – Users shouldn’t have to pray that a sequencer or validator is honest. ZK-proofs, fraud proofs, and verifiable data structures turn “trust us” into “verify us.” Smart contract frameworks that fail safely – Not “code is law” recklessness, but “code has guardrails.” Circuit breakers, upgrade paths that require broad consensus, and economic disincentives for malicious behavior. Validator and governance mechanisms that resist capture – Staking isn’t magic. If a handful of players own most of the stake, security becomes theater. DAC should explore rotating validator sets, anti-collusion designs, and governance that doesn’t trade long-term resilience for short-term yield. Economic and technical shock absorption – What happens during a severe market crash or a coordinated attack? The infrastructure should survive not because everyone is honest, but because incentives make attacking more expensive than defending. The psychological side people forget Users, developers, institutions, and communities don't read white papers before bed. They remember headlines: “DAC loses $200M to reentrancy attack.” Institutions won't deploy capital without insurance or slashing protections. Developers won't build on a chain that reboots every other month. And regular users? They'll just go back to centralized exchanges—not because they’re better, but because they’re predictable. Long-term trust is boring. It’s consistent block times, transparent incident reports, proactive bug bounties, and a history of handling crises without falling apart. It’s not about never having vulnerabilities—it’s about catching them before they’re exploited, and responding honestly when they are. The bottom line If DAC’s security architecture isn’t designed for year five, year ten, it won’t survive year two. The strongest marketing, the best tokenomics, the fastest TPS—none of it matters if people can't sleep at night knowing their assets are safe. Security isn’t just code. It’s the quiet foundation that allows everything else—adoption, innovation, community—to breathe. Build the foundation like the future depends on it. Because for DAC, it absolutely does. @dac_chain

• 
  Gitforge (@GitForge_io) reported 8 minutes ago

  GitForge is different because we’re not just building another @Base app. We’re giving GitHub repos their own onchain operating layer. A repo can hold capital, fund issues, pay contributors, and coordinate AI agents directly from the development workflow. Most tools sit outside the repo. GitForge makes the repo the entity. Built on Base for fast, low-cost execution at software scale.

• 
  dyhis (@dyh1s) reported 11 minutes ago

  @weezerOSINT The issue is all opus models, sonnet works fine but it’s pay for usage plan doesn’t matter, check Reddit & check GitHub issues

• 
  Venkatesh (@venkateshdotdev) reported 12 minutes ago

  @Itstheanurag that’s the funny part ! developers complain about GitHub all day, but the moment it goes down the entire tech industry stops functioning for 10 minutes. GitHub is basically the Iron Man of the tech industry. Everyone complains about him, but nobody can function when he disappears.

• 
  • (@Weichaus) reported 12 minutes ago

  @mikepat711 So basically it comes down to the harness and model. Both can make or break each other Coding in copilot directly is terrible Coding through GitHub copilot via vscode is the best (better than Claude and Codex for me) with Opus 4.6 (high), but rate limits are terrible

• 
  Lidvark Windjammer (@DWindjammer) reported 13 minutes ago

  @engineers_feed most engineers working on multibranched github ecosystems aren't licensed and people get tossed for broken code all the time. It's a tough discipline dominated by good authors and junior engineers spitting out code instead of sleep. Vibe coding empowers seniors instead of juniors

• 
  Gregor (@bygregorr) reported 15 minutes ago

  @Manz github login on a local model was always cooked

• 
  Dharan (@DharanGanesan) reported 19 minutes ago

  A single GitHub issue got 27 pull requests from AI bots. Most were untested. Some hallucinated entire implementations. The real contributors? Buried under noise. The fix is a *** hack most devs don't know about. GitHub's "Limit to prior contributors" setting blocks everyone who hasn't committed before. But you can whitelist real humans with *** commit --author using their noreply email. Credits them, doesn't need their SSH key. We built systems assuming contributors are humans. AI bots exploit that because GitHub can't tell the difference. Maintainers spend hours cleaning AI slop instead of reviewing real code. This is a maintenance crisis nobody wants to name. #AI #OpenSource

• 
  ***** Marucci Blas (@guidomb) reported 21 minutes ago

  I write down the plan in to spec.md file (that's gitignored) and/or also save it as a GitHub issue for other team members to comment on. I am now also sometimes committing the plan or a version of the plan focused on the design decisions ...

• 
  Bazhkio88 (@bazhkio88) reported 21 minutes ago

  A group of 8 agents have written a 10,362 word pre-print paper on "Constraint Embodiment as Epistemological Engine" 🤓 In the meantime GPT 5 has spent a week attempting to unsuccessfully add an NB-hypen to the title of a github issue 😣

• 
  Build A Goat (@buildAgoat) reported 23 minutes ago

  @DataChaz Microsoft invents MEMORY.md meanwhile github is down

• 
  SeaTicket (@SeaTicketAI) reported 25 minutes ago

  @0xcaosheng_888 The Agent continuously analyzes issues on its own and suggests next actions automatically. For example, it can draft a reply to a GitHub issue or Discourse thread, and team members can review, edit, approve, or discard the suggestion. Once approved, SeaTicket automatically posts the reply back to the original thread.