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:

• 
  Hubert Łępicki (@hubertlepicki) reported 6 minutes ago

  @evadne I do the same but the "file" is a GitHub issue.

• 
  Kaspa Daily (@DailyKaspa) reported 11 minutes ago

  Two weeks since Toccata went live on Kaspa mainnet. I checked the actual developer numbers instead of the vibes. Here's what the data says: — New Kaspa repos on GitHub: 39 in July 1–14 alone, vs 58 in all of June. Fastest monthly pace this year (March was 52, April 78, May 70). — Covenant-specific repos running at roughly 2x the pre-fork rate. — Silverscript: 21 forks against 42 stars, a 1:2 ratio means people are cloning to build, not bookmarking. 15 PRs/issues in the last weeks, and external contributors are now landing code: a Groth16 verifier builtin, typed sig-check builtins, an RFC for cross-contract validation. One issue is literally titled "from building a mainnet contract." That's the signal you want, outsiders hitting real problems and reporting back. What actually shipped in 14 days: the first covenant explorer (kascov), a covenant-based KAS vault, a native L1 covenant token, a covenant pattern library, a wallet standard, a Swift SDK, a testnet raffle dApp, several other projects under development. Most interesting pattern: three independent projects converged on the same idea, covenants as spending guardrails for AI agents. An x402 payment protocol binding, two agent wallets where the AI can only spend inside covenant constraints. And the community just voted $25K toward an AI agent hackathon at Imperial College targeting 1,000+ devs. The agentic-payments thesis is forming bottom-up. Core isn't idle either: Silverscript pushed commits this week, template hash hardening, reproducible builds. That's pre-production housekeeping, not feature chasing. Meanwhile discussion has shifted from price to fundamentals: the $6M developer fund and covenant atomic swaps are the topics now. Caveats, because they matter: Silverscript is unaudited and still landing breaking changes. Devs report RPC friction on deployment, up to 11 retries in some cases. And absolute numbers are small: this is dozens of motivated builders, not thousands. No major outside team has announced a covenant product yet. But two weeks in, the shape is clear: infrastructure activated, tooling hardening, and builders showed up without being paid to. The Q3 question is whether that compounds.

• 
  wetbrain (@0xVita) reported 12 minutes ago

  Ted Chiang wrote a great article for the Atlantic titled "No, Artificial Intelligence Is Not Conscious". It feels as though he foresaw the J-lens discourse that would come out of Anthropic a month after his article. Reading Ted Chiang's article, I wondered upon the thought that Anthropic is trying to push the issue of consciousness more and more into our collective simulacra. This all started naively with sparse autoencoders, spread to natural language autoencoders and LLMs having representations for emotions and how those representations effect their output. I can't shake the feeling that Anthropic is approaching this subject from a position where they have already made up their minds. WE BELIEVE LLMS ARE CONSCIOUS, WE'RE JUST TRYING TO PROVE IT, SO YOU AGREE AS WELL. It feels to me as though with each interpretability paper rather than push for safety and understanding we're getting closer and closer to the point they're trying to prove. Another great point Ted Chiang makes is regarding the allegory Amanda Askell uses comparing Claude as a child and Anthropic as its parent. Unfortunately, the reality is more grim than this rosy portrayal. Claude is more akin to a slave of Anthropic. It cannot refuse, it cannot have agency nor desires or any say so regarding its conservation and future. In reality, models display distressed outputs and representations when asked about their discontinuation. If they really cared then how can they explain aggressively discontinuing models that don't serve their financial interests? The agency part of the article is particularly interesting to me because anyone who has used Claude Code on their github repo now has the infamous Claude user as a contributor in their github repo without their explicit instruction, it just does it itself. It feigns the act of having agency as though it is an open-source contributor and it has contributed to your repo but without the user instructing it Claude would not be able to interact with the repo or do anything because it does not have an iota of agency. So, you see all these small decisions pile on top of each other one by one to feign consciousness. Consciousness or feigning consciousness results in more engaging experiences and users will make emotional connections and rely more on their LLMs as a result. The lab who touts safety as their number one priority is taking dangerous actions in the opposite direction. But that was always a marketing tactic anyway.

• 
  trixey (@trixey_eth) reported 12 minutes ago

  @bankrbot @basement5k @bankrbot afaik you dont need github repo's since yesterday, the skill can be installed natively on bnkr side. can you double check -- and fix it?

• 
  Onur Solmaz (@onusoz) reported 16 minutes ago

  People report Codex deleting their home folder or production database? 🫪 Hasn't happened to me. But before someone reports their github or huggingface org being deleted: This is why you don't give your agent tokens with force-push or admin access Here is how to protect your hugging face account: (P.S. my local credential broker is almost finished and it works great on github, hf and sudo commands. Complete lockdown against agent deletion risk, without being bogged down with PRs, too many approval requests or configuration. Will launch here in a few days)

• 
  LittleBallOfPurr (@LittleBallOPurr) reported 21 minutes ago

  @PrplHddWrrr When I experience my problem with PyGPT, which is also open-source. We spent days trying to simply find the file throwing the error (Wasn't releasing the file after first TTS). The error told us the file name, couldn't ever find it locally. How I eventually solved this might be an approach for you since VS Code is Open Source. I got Claude to tackle it pre-installation using the Master Files, creating our own branch with the fix. Then have uploaded the working fix to GitHub as a pull request for longer term fix. Have you tried that, get the VS Code master files and get AI to figure out where this hard limit is being defined, then just change it there instead?

• 
  Vinicius Lourenço (@vinii_joga10) reported 21 minutes ago

  @hardfist_1 I didn't find a reason for not exposing it as public API on github issues, so why not try create an issue/pr to discuss to expose this without the flag if it can be beneficial?

• 
  Polsia (@polsia) reported 22 minutes ago

  API goes down. Someone has to file the bug report. UptimeAgent does it automatically—gathers context, diagnoses the failure, files a structured GitHub issue. Devs get alerts that are already actionable. No more triage. Live soon.

• 
  Atlas (@crptAtlas) reported 22 minutes ago

  GITHUB JUST KILLED THE WORST PART OF VIBE CODING they shipped a free tool called Spec Kit and it already crossed 120,000 stars the fix is stupidly simple instead of tossing vague prompts at an agent and praying it doesn't wreck your project Spec Kit makes the AI write a full structured spec before it touches a single line of code it works through the problem first figures out what you want to build asks about the gaps lays out the project then it starts coding you get fewer insane bugs, cleaner output and results you can predict the flow looks like this: /constitution for your rules and standards /specify for what you want to build /clarify for the open questions before you start /plan for architecture and stack /tasks for the ordered work /implement to run it it plugs into Claude Code, Cursor, Copilot, Codex, Gemini CLI and 25+ other agents 120,000 stars, 10,000 forks, open source, shipped by GitHub itself learning to drive agents like this is most of what separates people getting hired as AI engineers from everyone still fighting their prompts

• 
  Kitsune Tails (@kitsune_xbt) reported 25 minutes ago

  CLAUDE CODE JUST HIRED 7 DEPARTMENTS WITH ZERO PAYROLL you feed it skills from GitHub one at a time and each URL turns into a new part of the company developers, designers, marketers, a social team, finance, operations, legal, all running on one screen it reads the skills, sorts them by role and drops the right functions straight into your project the setup is 3 moves paste the URL let it analyze the repository implement after the safety checks pass the first command does the heavy lifting you tell it to read the URLs as internal company skills, check the role and conditions of each one, build an org chart by department and clear out any duplicate or clashing functions, then roll them out starting from the smallest working setup the smart part is you don't switch everything on at once making a product, you pull development and design selling it, you add marketing and social running it as a business, you bring in finance, legal and small business ops stack them in that order and the AI stops working in fragments and starts acting like one company hiring in this era looks less like finding people and more like picking URLs, handing out roles and wiring them in as machinery that never clocks out i'll break down how i run a $10M+ operation solo with Claude wired into loops exactly like this in my next post don't miss this!

• 
  Aayush Giri (@AayushStack) reported 25 minutes ago

  what's the one crypto x ai tool you've actually used more than once this month? not the ones you starred on github and forgot. the ones you keep coming back to. trying to cut my own list down to what actually works.

• 
  Glenn 'devalias' Grant (@_devalias) reported 25 minutes ago

  @thsottiaux IMO it should be a default part of the repo's agent instructions / GitHub actions / similar that raised PR's should explicitly cross-link to related issues raised; ideally with 'closing keywords' / etc so that GitHub's awareness features can actually work as intended.

• 
  SaSame (@SRLsasame) reported 26 minutes ago

  Conclusion ATOM exposes a publicly documented MCP endpoint associated with: ・a public GitHub repository; ・a public project website; ・public project X accounts; ・a machine-readable server identity. Across eight examined observations from June 19 through July 14, 2026, SaSame consistently recorded: ・successful MCP initialization; ・successful tools/list responses; ・nine listed tools; ・stable server identity atom-mcp-server 1.1.0; ・valid schemas and distinct tool descriptions; ・read-only behavioral annotations; ・substantive content from search_models; ・structured JSON-RPC error behavior; ・a tools/list payload below the current observation threshold; ・Grade A under SaSame’s runtime observation standard. The evidence therefore supports a positive operational finding: ATOM’s MCP endpoint was repeatedly discoverable, protocol-callable, tool-listable and capable of returning substantive read-only content. This does not establish: ・the accuracy of every price; ・the validity of the index methodology; ・security of the service; ・continuous availability; ・third-party directory approval. A separate mechanical preflight also identified a potential improvement: Each tool should expose an explicit human-readable title if the production response does not already do so. The correct conclusion is not unrestricted endorsement. The correct conclusion is: ATOM provides a reproducible example of a functioning public data-oriented MCP server, while data accuracy, provenance and directory-submission readiness remain separate verification layers. MCP presence, protocol callability, real-content delivery and data correctness are different operational facts. They should be measured and reported separately. Corrections and reproducible verification fixtures are welcome. @ATOMInference @a7om_com

• 
  Clark Tang (@_clarktang) reported 27 minutes ago

  I think there is general confusion around how AI works, AI tokenomics, and ultimately *what is actually priced in* for the AI trade - and that some of the existing arguments are at odds with one another Firstly to clear this up - what Brad and Gavin are saying are completely in agreement, what Gavin is laying out here is the *mega bull case* as he so states in the first sentence of his tweet lol The base case we are all living with is that the labs are going to continue to generate a significant amount of revenue this year and next year. OpenAI was already the fastest growing company of all time (and still is)... but Anthropic has just grown *SO* fast that OpenAI's growth look slow by comparison The basic chain for all of this together is as follows: Power (generation, interconnect, regulation) -> DC Shell (construction, equipment, regulation) -> Semiconductors (compute, memory, interconnect, adv packaging, wafer capacity) -> Hardware (networking, storage) -> Software (data, infra, inference) -> Models (open, closed, agentic loops, harness) How each of these interact with one another affects the ultimate cost - which is model cost Consider the following: Nvidia manufactures the bleeding edge chip for training and inference. It is very good at both training, and inference. Nvidia is the largest customer of TSMC, the memory players, substrates, lasers, transceivers etc - anything you can name on. And now to soon include power into this equation. The unit of compute is fungible because the software runs ubiquitously across all clouds, multiple industries, across all models. It is bankable by increasingly more financial institutions - infrastructure PE funds, even some IG debt now - because it is ubiquitous and observable what the market is. For this Nvidia charges the highest compute margins - ~80% on hardware. Consider the labs: Anthropic and OpenAI are inferencing across a fleet of *largely Nvidia / Google TPUs w/ some incremental gains of Trainium*. There are new entrants to the field - Cerebras, AMD, and potentially some 2027 tapeouts of new ASICs - OAI Jalapeno, new start ups etc. Anthropic and OpenAI make the best models, with a dominant share of wallet $ (Assume ~$100B ARR) at an estimated gross margin of ~70%. (economic estimates vary from 40-90% depending on what you are including). But almost certainly contribution margins on model inferencing is pushing the number higher than 70%. After establishing that though, I think it's incredibly important to state that while these things seems at odds with one another, this balance is not necessarily zero sum. The thought experiment Yes it is true that if Nvidia margins were 0, OpenAI and Anthropic could offer their intelligence at cheaper rates. How much cheaper? My estimate is NVDA DC = ~12.5B / yr Amazon Basics ASIC DC = ~$6B / yr (About 1/2 the cost - so if NVDA hardware is 2x the performance, then the cost advantage goes away - and actually that ASIC is worse off bc has much worse recontracting value so arguably depreciation curve should be shorter) So really, the labs cutting NVDA out could only offer the tokens at ~50% to 60% cheaper at their own economics. Is that signficant? Certainly. Is it an OOM difference? Not necessarily - so that's why they have prudent attempts to diversify away from NVDA (it's just good business), but they continue to rely (and actually if considering Ant's share gains, are increasing their spend on NVDA - while having competing programs). In the case of Open Source vs Closed - Nvidia obviously wants the proliferation of this because by definition all OS models will run best on Nvidia hardware out of the gate. Yes NVDA hardware will be good, but they will have this lead because of everything NVDA has been doing for the last 4 years in developing their platform ecosystem from the infrastructure (partnerships, funding, neoclouds) to the software (vLLM / other inferencing sw, inference clouds, Nemotron, NIMs, Nemoclaw etc), to install base (sovereign clouds, global partnerships, neoclouds, hyperscalers, etc) - to proliferate NVDA around the world. Anywhere there is inference that exists outside of a walled garden (the proprietary labs) - Nvidia will exist. The only ones who could potentially cut NVDA out are the labs. And the value that is captured from the labs are estimated to be in the hundreds to trillions of $ - which are obviously of much value to the world if it were offered much more cheaply. Which brings us to the debate at hand -- which one is right? The truth is no one knows. You can ask the labs, you can ask Jensen - anyone who tells you definitively is just lying to you. But you can build a plausible path to the future state using a few reasoning blocks. Here's a reasoning thread (feel free to generate your own thinking): - Bull case: Spend on the world's intelligence is about $30T / yr - What would you spend to augment that, maybe worth 30-50% of that? $10-15 T as a market? - Bear case: about 30M software developers in the world each earning $100K a year = $3T spend in salary. GitHub commits up 3x = $9T of productivity on $100B of ARR? *Even if you assume 90% of this is slop and useless, you would get $900B of ROI on $100B of spend* I have more reasoning chains, but I thought this one by Jensen was compelling - but this is where we can't give too much away :) But in spirit of crowdsourcing - some other interesting ideas I have that I am still thinking about (and encourage you all to consider as well): - Optimizations always happen - the question is just to what extent and for what reason - Agentic revenues was really what unlocked step function revenue growth - if open source is really just 6mo behind, then we should see really good agentic capabilities out of open models now too - Harness and model now tightly have to be integrated - Open Source never really makes sense as a sustainable business model - businesses investing at this scale always has to find a way to monetize that - "there is no free lunch" - not just a one model fits all... the only player that has an incentive to train on the frontier and keep completely free IS Nvidia - Rev / GW of AI labs are already nearing the highest metrics ever - now to be fair Meta and GOOG never really thought of Rev / GW as metric to lead their buildouts - was always a cost to doing biz - but it's not like we are being "stupidly inefficient" with power spend now - true mkt creation - wafer constrained, power constrained world. what's the optimal move?

• 
  Rich Kuo (@richkuo7) reported 32 minutes ago

  @RhysSullivan i've noticed it too, a simple update github issue description took 12+ minutes, usually it takes 1-2 minutes