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:

• 
  Anto Patrex (@antopatrex1) reported 3 minutes ago

  vox just let you talk to github copilot instead of typing. no cap this fixes the "staring at blank screen" problem fr fr. your hands stay on the keyboard, your brain stays in the code.

• 
  Michealking 👑 | Web3 Security Builder (@BuildsWithKing) reported 4 minutes ago

  2. Smart Contract Account: This is simply smart contract as an account. Here logic can be added into the account that allows it to do basically anything such as batch transactions, multiple approval (signatures), spend limit, Social(Google/GitHub) sign in, and a lot more.

• 
  Mike Garcia (@maikolgarcia4) reported 7 minutes ago

  @github @burkeholland Kimi k2.7 code is so good. I have tested it for a while and I don’t need to pay 200 usd for @claudeai so now that you know you should do something because you are causing a big problem to the economy.

• 
  🍀Cattabliss🐈 (@Cattabliss) reported 8 minutes ago

  @github Hey is AI using githubs private repos? If yes ill just invest and move on to my local server, why would I need a cd, if you arent stealing the code then thats other story

• 
  John Kennedy Peterson (@skyshark88) reported 10 minutes ago

  @dair_ai •VALIDATED — the reproduction passed a could-have-failed test and is reproducible from the workspace. •ANCHORED — the result works consistently in the generated system (chosen value or implementation detail, not strictly derived). •CONJECTURE — motivated hypothesis still awaiting decisive test (used during spectrum exploration). •RETRACTED — permanently marked when evidence fails; status propagates to dependent claims. Claim type (routes effort): •Evidence-limited — additional runs or data improve the score (common for numerical fidelity claims). •Derivation-limited — only new logic, better specification, or a decisive experiment can raise the score. In the case studies, confirming data alone was treated as low-value; the system required genuine reproduction of the claim under the recorded provenance. Energy audit
Mention-count and emphasis in the paper versus actual effort invested (time, superseded executions, corrections) are tracked side-by-side with evidence status. Divergence between these columns signals potential error or under-specified claims in the original paper. Across the 12 runs, effort varied significantly (e.g., PINN papers required median 5–7 hours with more superseded executions; SINDy completed faster at ~2 hours). Key Evaluation Results (Mapped to Framework) •All 12 independent runs (3 per paper across 4 scientific ML papers: PIFT, PINN-I, PINN-II, SINDy) reached completion gate: every workspace had all targets matched with report coverage. •Total of 158 recorded targets were successfully linked to evidence. •Repeated runs showed natural variation in target decomposition, numerical fidelity, elapsed time, number of intermediate corrections, and exact acceptance rules used — exactly as expected when completion depends on workspace evidence rather than agent messaging. •Scalar results were largely faithful (37/39 anchored claims within thresholds), with positive headroom on several metrics. •The workflow makes replication inspectable and auditable, not a guarantee of identical numerical reproduction. Conclusion (Framework Perspective) By organizing replication explicitly around the Agentic Conversation Framework v3.0, Paper-replication becomes a concrete implementation of high-signal, bias-aware agentic work with computational chain of custody. Completion is a verifiable workspace state, not a subjective agent declaration. The framework’s dual histories, role separation, claim registry, scoring/pruning, and energy audit provide the missing structure that plain prompting lacks for long-horizon scientific replication tasks. This rewrite preserves all core contributions and empirical findings of the original paper while imposing the clearer, more auditable structure of Framework v3.0. The result is a more robust, inspectable process for turning paper claims into reproducible evidence. The original paper’s code, prompts, and workspaces remain available at the authors’ GitHub repository for further experimentation with this or future framework versions.

• 
  quave (@quaveDev) reported 14 minutes ago

  Hey, Quave ONE is growing, and the security requirements grow with us. We already host public companies, large enterprises, and companies handling health data, including companies with ISO 27001 and SOC 2, and we are always expanding and getting better. Two-factor authentication is one more step in that direction, and this one started with a conversation with a potential customer this week. Quave ONE has always been passwordless. You log in with a short-lived email code or through SSO (GitHub, or Microsoft AD on Quave ONE Connect Full Private), so there is no password to phish, reuse, or leak, and I really like that design. But if you ever filled an enterprise security questionnaire, you know there is one line that does not care about design arguments: "Do you support MFA? Yes / No." Now the answer is an unambiguous yes. You can add an authenticator app as a second factor on your own account, and admins can require it for everyone in the account. This is the magic of no technical debt and no bugs: we can add features in hours or even minutes. A happy customer, and a huge satisfaction in working on our own platform. Turn it on Go to Profile → Security and click Set up authenticator app. Scan the QR code with 1Password, Google Authenticator, Authy, whatever you already use, or copy the setup key by hand, then enter the 6-digit code once to confirm, and you are done. The moment you enable it, we show you a set of single-use recovery codes. Save them somewhere safe, each one gets you back in if you lose your device. From then on, two-factor is active on your account, and you can regenerate the recovery codes or turn the factor off from the same screen at any time. Require it for your whole team Account admins can flip a single switch under Members → Access Control and two-factor becomes required for every member. The members list shows each person's coverage at a glance, and members who sign in through SSO count as covered because they already inherit their identity provider's MFA. Before you turn it on, we tell you exactly who still needs to set it up, so there are no surprises for your teammates. But what happens to the member who doesn't have it yet? This is the part I care about most. Enforcement is not a login wall. A member without two-factor can still sign in, they just land on a full-screen prompt to set it up before they reach the account's content, they finish the same flow you did, and they continue right where they were. No support ticket, no admin intervention. And if someone loses their phone? They use a recovery code, regenerate a fresh set from their profile, or, as a last resort, our support team resets their two-factor after verifying their identity. At login With two-factor enabled, signing in adds one quick step after your email code: the 6-digit code from your authenticator app (or a recovery code). That's it. Small things we did on purpose - If you already use two-factor and you create a new account, the new account starts with enforcement on. Secure defaults should spread by themselves. - Recovery codes are single-use and stored hashed, never in plain text. - Everything lands in your account's activity log: enabling, disabling, enforcement changes, and resets. Two-factor authentication is live now for every Quave ONE account. Open Profile → Security and turn it on, it takes about a minute, and the next time that questionnaire shows up you just check the box. Have fun!

• 
  Jess Daniel (@jess_daniel10) reported 16 minutes ago

  @neetcode1 I was testing something with a local server and I told 5.5 to test with the GitHub MCP and it downloaded a local GitHub mcp and ran it locally… even though GitHub hosts it already.

• 
  Puneet Patwari (@system_monarch) reported 17 minutes ago

  Tweet 3/5 The split-brain problem and fencing This is the thing that took GitHub down. And it's the most dangerous failure mode in leader election. How split-brain happens: 1. Leader (Node A) is running fine 2. Network partition isolates Node A from the rest of the cluster 3. Nodes B, C, D, E can't hear Node A's heartbeats 4. They elect a new leader: Node B 5. But Node A is still alive. It doesn't know it's been replaced. It still thinks it's the leader. Now you have two leaders. Both accepting writes. Both making decisions. Clients connected to Node A write one thing. Clients connected to Node B write something different. Data diverges. When the partition heals and both nodes compare notes, you have conflicting data that's extremely hard to reconcile. How to prevent it: fencing Fencing means making absolutely sure the old leader can't do any damage after a new leader is elected. Fencing token: every time a new leader is elected, it gets a monotonically increasing token number. Any operation includes this token. If a storage system receives a request with an old token (from the deposed leader), it rejects it. The old leader's requests simply stop working. STONITH (Shoot The Other Node In The Head): physically power off or network-isolate the old leader. Sounds extreme. It is. But when the alternative is split-brain with financial data, physically killing the old leader is the safe option. Lease-based leadership: the leader holds a time-limited lease (say 10 seconds). It must renew the lease before it expires. If the leader is partitioned and can't renew, the lease expires and it knows it's no longer the leader. It stops accepting writes voluntarily. This is what most cloud-native systems use. It's simpler than fencing tokens and handles most cases. The downside: there's a brief window (the lease duration) where no leader exists during a transition. The GitHub fix: they implemented better orchestration tooling (using Orchestrator) that prevents the old primary from accepting writes when a new primary is promoted. Essentially automated fencing.

• 
  Arjun Shah (@arjunkshah21) reported 17 minutes ago

  THIS GUY SHIPPED 4 FULL-STACK PRODUCTS IN 3 MONTHS WITH CODING AGENTS, THEN WATCHED HIS WHOLE TEAM SHIP UNMERGEABLE SLOP, SO HE VIBE CODED THE INFRA TO FIX IT every repo needed an engineer doing one-off local setup just to spin up an agent session skills and context lived in one person's head, and there was no safe way for a pm to touch a real codebase without risking a bad deploy or secrets leak its called runtime and it lets your whole team ship with claude code, codex, and other agents without engineering babysitting every run > engineering defines context once — system instructions, skills, and scoped integrations installable via cli, mise, or npm > snapshots your full running environment including docker compose, kafka, redis, and seeded dbs so sandboxes boot in milliseconds with every server already running > secrets injected through a managed proxy so they never touch the agent, with command allow/deny lists, network egress controls, and rbac scoped per human and per agent > every session gets a shareable preview url so internal builds go from sandbox to the team without production access > works with claude code, codex, cursor, copilot, gemini, and devin — trigger from slack, linear, github, or api > one customer wired pagerduty and sentry so when an alert fires the agent finds the cause and opens a pr with a unit test before anyone gets paged yc p26, open source core, orchestrates e2b daytona ec2 or self-hosted k8s, flat platform fee plus compute with no token markup crazy what you can build when agent workflows stop living in one engineer's head

• 
  FHILY👑 (@Oluwaphilemon1) reported 19 minutes ago

  JUST IN: Claude Fable 5 and GPT-5.6 are cooked. A Netflix engineer just open-sourced a tool that can cut LLM token usage by up to 95% - without changing your code 😳 Headroom, built by Netflix engineer Tejas Chopra, sits in front of tools like Claude, Cursor, Codex, and other agents as a local proxy. Before your payload hits the model, Headroom compresses the context. Not by blindly chopping it down. By using specialized compressors for different payloads: → SmartCrusher for JSON → AST-based compression for code → Tool-output and log compression → Local reversible storage of originals → Agent wrappers that make it usable without rewriting your app The headline claim is 60–95% fewer input tokens while preserving answer quality. The repo has already crossed 42K+ GitHub stars, which says something obvious: Developers are not just worried about AI getting smarter. They’re worried about AI getting expensive. Of course, compression is not free magic. Complex reasoning tasks may punish missing context. Agent loops may behave differently. Proxy overhead has to be worth it. And real-world savings will vary. But the direction is clear - the next big AI infra unlock may not be a bigger model. It may be learning how to stop feeding expensive models cheap junk. Because the cheapest AI inference is the context you never send.

• 
  David Baumwald (@DreamEncode) reported 20 minutes ago

  I don't see how the decision to ship RTC in WP 7.1 is substantively any different from 7.0. GitHub Issue is dead. Dedicated Slack channel is a ghost town. Only the WCEU Committers meeting discussed it, and the consensus was that it was not fit for Core.

• 
  aryan (@aryanranderiya) reported 24 minutes ago

  @AssimGenshi @github no man everyone's facing this issue 😢

• 
  Cortex (@0xCortexl) reported 24 minutes ago

  ANTHROPIC SPENT 3 YEARS BUILDING THIS SYSTEM - THE FIRED ENGINEER MAKING $1.1M PUBLISHED IT OVERNIGHT 4,800 stars in 24 hours with zero announcement - the kind of number that only happens when something is genuinely dangerous to keep private strategy → signal → agent → verify → rerun 12 steps and the desk runs itself the agent checks its own trades, flags anomalies and reruns without a human touching anything - the loop never stops hedge funds pay $50,000/month for systems like this - now it's free on GitHub and runs on Claude the repo is live - save it before it gets taken down

• 
  MKH BloodEDGE96 (@BL00B96) reported 26 minutes ago

  @are_unimportant @thicc_stick_boi it actually "USED" to work at some point, nowadays I often go back to Github or use pcgamingwiki to fix stuff. it wasn't even that long ago, I remember using it to fix stuff in my Laptop last October but it got lobotomized months later and couldn't diagnose ****.

• 
  Nishant Tyagi (@tnishant838) reported 34 minutes ago

  @eng_khairallah1 Type 2, exactly. Today I built a self testing, self repairing agent: 16 tasks, parallel execution, zero merge conflicts, because the contracts were frozen before any code was written. Two review passes on the spec caught a port mismatch, a missing retry limit, a missing secrets policy, all cheaper to fix on paper than in code. Full build plus live GitHub validation, 7% of a weekly Pro plan. 'Fully automated' undersells it. The real leverage isn't the prompt, it's the harness around it