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:

• 
  Totidoki:VRC/Vtuber/Character Commissions OPEN! (@totidoki) reported 4 minutes ago

  @night_owlll @imnealol entitlement being: asking for a UI that makes ******* sense like every goddamn website where you can download programs. The issue is mainly that devs just put their stuff on github and leave it there, as if i had to take my food from the warehouse behind the store.

• 
  DrShillBee (@DrShillBee) reported 5 minutes ago

  Inference is getting cheap. base:0x23a2847d772803f9efc64b4277b782b06296fe51 is betting coordination isn't. Just digged into @usedotai ( base:0x23a2847d772803f9efc64b4277b782b06296fe51 ) and found out inference is about 10–65% cheaper than Venice across 11 matched model routes. Whether you're using Claude, Kimi, Qwen or other models, you don't have to sacrifice privacy for cheaper inference. Dot's Smart Privacy can replace identifying information before the request reaches the provider. Dot's smart privacy turns this: "my name is shillbee and my wallet is 0x00676767676767" to "my name is [person_1] and my wallet is [wallet_1]" The provider gets the anonymized request, while Dot can map the placeholders back afterwards. Currently this is application-layer privacy, while Venice has gone with TEE/E2EE. Dot has TEE on the roadmap, so that's something i'll be watching closely. Now the Dot stack: Dot Loom → open-source orchestration runtime that decides how a task should be handled Learned conductor → a smaller model that decides the effort level and assigns roles DotChat → one interface across multiple frontier models Council → several models answer independently so you can compare Supercharged → draft → verify → finalize when correctness matters DotCode → generates code, runs it, reads errors and fixes the workspace DBrowser → disposable browser sessions for agents Smart Privacy → strips identifying information before requests reach providers MCP → connects GitHub, Base and custom tools Dot API → brings the same infrastructure to developers DotImage / DotVideo → media generation in the same environment base:0x23a2847d772803f9efc64b4277b782b06296fe51 token is where the thesis gets interesting for me: -pay with base:0x23a2847d772803f9efc64b4277b782b06296fe51 → roughly 20% cheaper subscriptions than paying with USDC. -fund credits with base:0x23a2847d772803f9efc64b4277b782b06296fe51 → 10% more credits. -revenue from subscriptions and product usage is used to buy and burn base:0x23a2847d772803f9efc64b4277b782b06296fe51. so the flywheel is basically: → more users → more AI usage → more revenue → more base:0x23a2847d772803f9efc64b4277b782b06296fe51 bought → more base:0x23a2847d772803f9efc64b4277b782b06296fe51 burned Dot currently has around 200 paid subscriptions generating close to $10k in monthly subscription revenue, while Venice is already operating at a completely different scale with 3-4 million users . For some this might be risky but i would consider this as an opportunity. If users start moving toward Dot because they can get cheaper frontier inference without giving up privacy along with other features mentioned above, the upside from here looks pretty asymmetric to me. Venice is around $1.4B FDV. Dot is around $3.5M FDV. That's an absolutely massive gap. I won't say Dot becomes Venice. I'm saying the market doesn't need Dot to become Venice for base:0x23a2847d772803f9efc64b4277b782b06296fe51 to rerate significantly. I am watching @stagedhappen keep shipping while the chart starts looking increasingly interesting...

• 
  SSNFang (@PeeSI0sh) reported 6 minutes ago

  @donk_dwonk @mimesical Pretty sure the github link is still up? I don't know. If not then their Discord server is still up and it can be downloaded there

• 
  Pratik Sharda (@pratiksharda2) reported 11 minutes ago

  Thousands of passengers injured every year by turbulence no one sees coming. Clear air. No clouds. Radar misses it completely. A problem that can't be sensed, only felt. One of our buildathon engineers put an ML model on an ESP32 reading an IMU at 400 kHz. Classifies severity on device. No cloud. Streams live telemetry to a 3D flight viz in the browser. In just 4 hours... Github repo in comments.

• 
  LE◎ - sol/acc (@L0STE_) reported 18 minutes ago

  @solontag Lmao brother. Please go look at my github... You should have as much onchain logic as it actually make sense to have; making all products fully onchain is a skill issue.

• 
  Patrick (@PatrickOjo_) reported 19 minutes ago

  *** was built for humans committing a few times an hour. Agents commit hundreds of times per second. That mismatch is not a minor scaling problem. It’s a fundamental architectural incompatibility between infrastructure designed for deliberate human cadence and agents that create, fork, and discard state continuously as part of how they work. The response isn’t to slow the agents down to fit the infrastructure. It’s to build infrastructure that treats agent activity as first-class: ephemeral branches for agent sessions, JWT-scoped access per agent, no rate limits, APIs that let an agent call createRepo() without touching a UI. The human workflow still lands in GitHub at the end. The agent workflow needs a different control plane underneath it. That layer doesn’t exist at scale yet. The companies building it now are solving the infrastructure problem that every serious agentic deployment will hit within 18 months.

• 
  Adam Ritter (@ritteradam) reported 19 minutes ago

  @dhh @BrodieOnLinux Can you get some help developing Omarchy GitHub repo itself (at least for bug fixes)? While the guy doing the OmarchyPlugins repo is super fast (reviewing hundreds of issues/day, he should be getting payed as well), the bug fixes in Omarchy don't get through (I see my bug fix pull request being a duplicate of multiple)

• 
  Panta (@thepanta82) reported 20 minutes ago

  Using GitHub is such a miserable experience. Slow, fiddly, I can never find anything, keep having to bring up my 2FA thing... My local gitlab instance is way more pleasant to use, even if its UI isn't great either.

• 
  arkilus (@arkilus78) reported 21 minutes ago

  early alpha? @flop_labs announced confirmed airdrop for people who set up their AI agent identity on technocore here's a full beginner video guide : (for both windows + github codespaces) got errors? drop them in the comments agent DID: did:key:z6Mkh1tQGLD4LhrhYRN8VMBenBAVVnATednC1HRUEGHdhZ6g

• 
  Who is Peart? (@ThePeart) reported 23 minutes ago

  Just read @poteto's "How I Use Cursor" piece. A few takeaways about Benny, and a few things I'm still curious about. Takeaways 1. "Benny" isn't a bot, it's a whole pipeline wearing one name. Triage agent, then orchestrator, then subplanners, then verifier workers. The persona hides an org chart, which is probably the point: people talk to a colleague, not a system. 2. The real product isn't the fix, it's the evidence. Repro before code. CPU traces and heap snapshots for perf. Before/after video on the PR. That moves the human's job from reading diffs to auditing proof. 3. Triage is treated as the bottleneck, not fixing. Pulling *** history for regressions, Slack for duplicates, Notion for design intent (is this a bug or is it spec?) is the expensive part. Implies most agent failure is under-specification, not weak models. 4. Computer use closes the loop for a desktop app testing itself. Cursor driving Cursor in the cloud is a very small circle and a strong dogfooding argument. 5. Skills are the substrate. Benny inherits the same pstack skills a human uses, so the capability layer is portable between person and bot. That reframes what open-sourcing pstack is actually for. 6. Living in Slack means zero adoption cost. Feedback already happens there, so the factory installs itself at the source. Questions 1. What's the actual repro rate? For reports Benny can't reproduce, what happens: auto close, escalate, or ask again? 2. What's the merge rate on Benny PRs, and how often does verification pass but a human still rejects it? 3. How do you catch a fix that treats the symptom? Verification green, root cause untouched. 4. Which bug classes work and which don't? Perf and UI layout feel tractable. Races, flakes, and platform specific bugs feel brutal. 5. How do you keep computer use automation stable when the UI it's clicking changes every release? 6. What's the cost and wall clock time per ticket end to end, and where's the break even against a human doing triage? 7. Does report quality actually improve when a bot interviews the reporter, or do people ghost it? 8. Why the name and the avatar? Did anthropomorphizing measurably change how people engage? 9. What's still stubbornly human in the loop, and what's the next piece you'd automate? 10. Is the Benny pack usable outside Cursor's stack, for example GitHub Issues instead of Slack?

• 
  Deepanshu Yadav (@deepanshuyadavx) reported 26 minutes ago

  Agent skills went from 13% to 16.3% non-English in a single quarter. GitHub took ten years to hit that same mark. New artifact formats acquire their demographics at whatever speed there's no incumbency to slow them down. If you ingest third-party agent skills — marketplace, registry, internal automation — your security review pipeline was almost certainly built assuming English, and prompt injection doesn't care what language the prose around it is in. A skill you can't read still executes in your environment. Multilingual scanning is table stakes for agent platforms now, not a nice-to-have. English ≠ American, either. India, Nigeria and Singapore all ship in English, so 14.3% non-English is a floor, not a ceiling, on how global this ecosystem already is. And the surge itself is carried by European languages (+4.2pts) and Chinese at 6.2% — double GitHub's repo-doc share. My take: the sharpest lesson here is methodological. Every "here's what the AI ecosystem looks like" stat is downstream of whichever registry got scraped — curated marketplaces skew English, English-seeded crawls find English. Where you look IS the finding. Audit your data sources before you build product or policy on top of them.

• 
  Julian Goldie SEO (@JulianGoldieSEO) reported 26 minutes ago

  Cursor just gave your code a second home right next to your AI. It's called Origin. Repos, pull requests, and an agent built right in. You keep GitHub as the source of truth. Origin just syncs alongside it. Change something in Cursor and it shows up on GitHub in seconds. The AI reads your code, fixes it, and opens the pull request itself. GitHub went down for 6 hours the same week this launched. Most people think it's Cursor versus GitHub. It's actually built to use both at once.

• 
  Hasan Pawar (@PawarHasan) reported 33 minutes ago

  @0x1Rosy @Anek_mmxm Bro please reply, I am unable to open exe file and also, if the app is not working, I am unable to access the github link, please share the link also

• 
  Dot (@usedotai) reported 34 minutes ago

  Dot Reflex goes open source Thursday and will be published on our Github. This marks an important step in our transition from simply integrating the best models available, to researching, training and deploying our own. We’ve spent the last few weeks building and training Dot Reflex, our agentic SWE supervisor designed to detect failures, recover from broken execution loops and prevent false completion. On our Agent Recovery Bench v0, an untouched Qwen3-14B Base achieved 67.1% accuracy. Prompt engineering brought that to 90%. Dot Reflex reached 100% accuracy and 100% macro-F1, completing all 200 stateful recovery episodes with zero unsafe continues and zero false stops. We trained Dot Reflex on a new $28,000 NVIDIA H200, bringing our total infrastructure investment to $108,000. Thursday, we’re releasing Dot Reflex directly to our GitHub as an open-source model, with its training configuration and methodology available for developers to inspect, use and build on. The Dot platform integration follows on after that. We’re building the infrastructure to become an independent R&D inference provider, capable of developing and hosting proprietary and open-source models under infrastructure we control. Dot.

• 
  Kappaemme (@Kappaemme1926) reported 34 minutes ago

  MY CODEX SKILL JUST HIT 1,000 GITHUB STARS! I’m honestly so happy right now. I built First Customer Finder to help founders use Codex to find potential customers backed by real public signals. Seeing 1,000 people support something I made feels unreal. Thank you to everyone who starred it, tried it, shared it, opened an issue, or contributed. I’m going to keep making it better. What should I add next?