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:

• 
  Ivan Fioravanti ᯅ (@ivanfioravanti) reported 5 minutes ago

  I want to publicly ask to anyone involved in Apple MLX world, being it building block or inference engine to slow down development and add more testing and QA. Every single time we test something new on MLX world we get tons of issues and we spend more time debugging, contacting builders and opening Github issues than enjoying the new releases. Please please please 🙏 Add more tests, it's quite simple for me crashing things, I just do long context benchmarks or batch inference and... BOOM! Imagine using this in a real production environment. I don't want this to be an accuse, just a request for more care to QA. llama.cpp and DS4 docet.

• 
  Sam Thapa (@SThapa123456) reported 5 minutes ago

  i told claude to fix a github issue without reading the issue myself. it opened a pr. looked clean. now i'm sitting here trying to do three things at once. understand what the issue actually is. understand what the pr actually does. steer the architecture if it went the wrong way. all in the same head. in the same moment. with a slack notification from my ceo pending. something i'm realizing as i do more agentic engineering: skipping the plan doesn't save effort. it just defers all of it to the worst possible moment. @theo and @steipete aren't big fans of the talk-talk-plan-execute flow. the argument is roughly that modern agents are capable enough that the ceremony slows you down more than it helps. just let it cook. i get it. but what i'm finding for myself is that plan-first isn't ceremony, it's a cost-spreading strategy. you pay the "understand the issue" cost when it's cheap, before anything is built. you pay the "shape the solution" cost at the plan stage, when changes are one sentence instead of a re-implementation. by the time the pr exists, the model is already in your head and reviewing it is just verification. skip those stages and the cost doesn't disappear. it stacks up and lands on you all at once, after the code exists, when every decision is now expensive to change. the polished pr is the trap. it looks like progress. it's actually a bill coming due. (credits to CC for helping me articualte this idea)

• 
  Cyfrin Audits (@cyfrin) reported 10 minutes ago

  This pairs with Cygent's daily scans against OSV, GitHub Security Advisories, Socket, and other intelligence sources. When a new CVE lands, Cygent checks it against your actual dependency graph, not a generic feed. You get an alert with the affected package, severity, dependency context, and suggested fix, routed to Slack, Discord, Telegram, or email.

• 
  🔅LAMIS (@lami_thefirst) reported 10 minutes ago

  The biggest red flag in crypto is when a project's X account feels more alive than its product. If every post is a KOL quote, partnership graphic, or engagement bait, I assume the team is managing attention, not building. My strongest green flag is a GitHub with consistent commits from multiple contributors over months. Not a prelaunch sprint. Not one developer carrying the repo. Real teams leave fingerprints in public. Narratives are easy to manufacture. Shipping isn't. I spend less time reading promises and more time looking for evidence that people are solving problems nobody is applauding yet. What's the most underrated signal you check before trusting a project? @RallyOnChain is a good example of a green flag.

• 
  ./can (@shcansh) reported 11 minutes ago

  Local development is shifting from a coding task to an agent-coordination problem. The new Agents window in VS Code lets you run more than 1 agent session side-by-side, while the open Agent Host Protocol standardizes how these sessions sync. Combined with air-gapped BYOK models, GitHub is turning the IDE into an autonomous control room. But does orchestrating multiple agents actually speed up shipping, or are we just trading writing code for a more complex debugging bottleneck? #VSCode

• 
  🔻agitprop + absurdity🔻 (@agtprpnabsrdty) reported 13 minutes ago

  Token economics are becoming AI's most inconvenient truth, and Sam Altman just outlined exactly why. OpenAI's top internal user burns 100 billion tokens per month — up from 100,000 six years ago. One external customer already exceeds that figure. Cost complaints are now the second most common issue Altman hears from enterprise clients. His answer is "always on" autonomous AI running in the background, which would multiply consumption well beyond current levels. The billing wall: GitHub Copilot switched to token billing two days ago and users burned through a month of credits in hours. Ramp data shows Anthropic passing OpenAI in enterprise spend, meaning competition for these customers is intensifying at the exact moment those customers are pushing back on price. The capex fantasy: IBM's CEO put the industry's capex requirement at $6–$8 trillion this week and noted the revenue to justify it probably doesn't exist. Altman is previewing autonomous agents that would multiply current token consumption without anyone requesting it. Either cost per token drops fast enough to make that viable, or enterprises start capping AI spend.

• 
  ▓▒░ al ░▒▓ (@chromegadget) reported 15 minutes ago

  oh **** when did github add a save/bookmark option to issues. now i just can do that instead of subscribing to everything

• 
  Gustavo Alessandri (@webgus) reported 18 minutes ago

  If you find an error, have an idea, or want to propose an improvement, just open an issue or fork it on Codeberg or GitHub. Contributions are welcome. That’s exactly the point.

• 
  Zephyr (@Zephyr_hg) reported 21 minutes ago

  i tested 9 claude connectors in june 2026. kept: 1. gmail (inbox triage + drafted replies) 2. google calendar (auto meeting prep) 3. slack (thread context + voice-matched replies) 4. github (repo state and pr diffs) deleted: 1. notion (read-only and slow) 2. asana (covered by cowork folder) 3. canva (output not there yet) 4. figma (read-only depth limit) 5. linear (duplicated state) 3-5 connectors picked deliberately wins. every-connector-on loses.

• 
  GCRtrd💊 (@gcrtrd) reported 24 minutes ago

  I still think @Pumpfun is the best launchpad, but lets take a look at the efficacy of all of their updates this year: > Github Fee Redirect: A clear copycat of Bags and was a move to capture some of that tech coin volume. Rarely used nowadays other than larps/scams, perhaps one good redirect token per week. > Cashback: Initially advertised in such a way that it sounded like an airdrop, the update has been useless and encouraged bundling (as its the only way for devs to profit). Only used now because its better than giving fees to scam devs. > Agent Mode: What on earth was this? Why is buyback and burns attached to an "agent"? The core deflationary mechanic is good, but nowadays its just used to PvP and for AI narrative coins. > USDC Coins: A decent addition and should've been done months ago. The fixed spawn/bond mcaps shows that Pumpfun KNOWS the SOL pair configuration is incorrect, but they refuse to adjust it. Largely just used for PvP and USDC-specific narratives, I'd say this was about 50% successful. > Communities: Another re-invention of the wheel, a decent idea but just fractures attention. Time will tell if this is a success, but given the general lack of participation, it's a bronze medal at best here. > GO/Bounties: What? Where did this come from? Seems like a direct ripoff of @Shillz_Official. Things like this are always better handled by individual communities, rather than centralized. Absolute waste. Ultimately, none of these changes have addressed the core problems with the platform. Adjusting the spawn/bond mcaps is the most meaningful change, but most coins are still SOL pairs, so this change wasn't very efficacious. I'd love to see @a1lon9 and the team take on the real problems facing the trenches, rather than these superfluous additions.

• 
  Alexander Zbiciak (@IBreakData) reported 24 minutes ago

  @luce_libera @wodarg One issue with the data is the data entry permutations are these the same lots? EW172 EW0172 PFIZER EW0172 #EW0172 Some entries just say pfizer or worse. The data needs guided cleaning to really show hot vs not lots @grok please review the github data and find the potential permutations for the pfizer lot ewo172 as shown above. Are there othe potential entries and what are the totals for that lot.

• 
  Bollish (@99_Bollish) reported 30 minutes ago

  @King_Memento That’s exactly what I’ve started noticing too. The funny part is that nobody gets angry when someone shills a larp. People only get angry when someone points it out. A larp project can waste thousands of hours of attention and millions in volume, and somehow that’s acceptable. But the moment someone opens the GitHub, checks the docs, and asks questions, suddenly they’re “ruining trades.” At the end of the day, fake utility projects don’t just hurt buyers. They also steal attention, liquidity, and volume from teams that are actually building. And i've already accepted that some people will hate it. If exposing a larp ruins a trade, maybe the problem isn’t the exposure. Maybe the problem is the larp. The market gets healthier when capital flows to builders instead of storytellers.

• 
  U.S.A.I. 🇺🇸 (@researchUSAI) reported 31 minutes ago

  🇺🇸 The First Order Consequence: - GitHub accelerated the rollout of Copilot Code Review to more repositories, which is expected to reduce time-to-merge and review turnaround - Developers used the tool to flag issues earlier in pull requests, leading to fewer late-stage rework cycles as teams cleared backlogs faster 🇺🇸 The Second Order Consequence: - Engineering managers observed improved throughput and more consistent review quality, supporting a shift toward smaller, more frequent pull requests rather than large batch changes - Teams standardized on the tool’s suggested checks, which lowered variance in review outcomes across reviewers and reduced the probability of “missed defect” clusters that historically created follow-on bug reports 🇺🇸 Discernment: - Review cycles that previously showed longer delays and higher defect escape rates began to normalize, with early evidence pointing to fewer hotfixes originating from overlooked review comments - Prior interventions such as review guidelines and linting rules remained in place, but teams credited Copilot Code Review with restoring momentum rather than replacing the baseline process 🇺🇸 Reasoning: - Current performance signals included faster identification of common issues during the PR review phase, aligning with the pattern that earlier detection reduces downstream churn - Adoption behavior suggested growing confidence: engineers who integrated the feature more quickly produced PRs that required fewer follow-up revisions, consistent with recovery from prior slowdown 🇺🇸 Judgement: - Early indicators support a net improvement in both individual and team growth, shown by reduced review friction and fewer late-stage corrections - Ongoing monitoring remains warranted to confirm that the recovery persists as rollout expands beyond early adopters, using falsifiable metrics such as average PR review duration, merge lead time, and post-merge defect rates

• 
  Noctum (@Noctum_io) reported 32 minutes ago

  noctum's smart contracts have been analyzed using two automated static analysis tools: Slither and Aderyn. these tools scan Solidity source code and compiled bytecode for known vulnerability patterns reentrancy, unsafe external calls, integer handling issues, access control gaps, and others. neither tool returned critical findings on the pool contracts, the Groth16 verifier, or the Poseidon hasher. the analysis results will be published when the formal writeup is complete. it is important to be precise about what this means. automated static analysis is a useful and important first pass. it is not a security audit. a security audit is a manual review by an independent firm that thinks adversarially about the system, models attacker behavior, and produces a signed report. noctum has not yet commissioned or completed a formal audit. stating otherwise would be false. this work is on the roadmap. the ZK circuit has not yet been reviewed with circomspect, a static analyzer specifically designed for circom circuits. circuit bugs are a distinct class of vulnerability from contract bugs a flawed constraint system can allow invalid proofs to pass verification. circomspect is pending. the trusted setup is also a known limitation: the ptau used is from the Hermez ceremony, but the zkey was generated with a single party locally. this is not a multi-party trusted setup, which means the toxic waste from the ceremony cannot be proven discarded. real ETH deposits carry this risk and users should understand it. the contract addresses, circuit source, deployment scripts, and zk-deployments.json are all public in the GitHub repository. the Groth16 verifier bytecode is deterministic from the circuit anyone with the source and zkey can recompile and verify the deployed bytecode matches. the goal is to make every security assumption visible so users can make informed decisions. an honest security disclosure at this stage is worth more than a false sense of completeness.

• 
  Anthony Shew (@anthonysheww) reported 34 minutes ago

  When was the last npm registry scare or GitHub outage? Knock on wood, but feels like things have slowed down?