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:

• 
  xavier (jack) (@KMkota0) reported 1 minute ago

  @MaximeHeckel eventually, everything good will be copied. i guess the difference now is that, with ai, the cost of copying is almost zero. the more i think about it, the more i believe the solution has to involve the platforms. reddit and github in my case. there will always be bad actors using new technology. the real problem is when those bad actors get distribution. platforms need to reconsider their policies around this new reality.

• 
  Gill (@gurtej__gill_) reported 5 minutes ago

  The biggest AI skill shift in 2026 isn’t prompt engineering. It’s LOOP ENGINEERING. Most people still work like this: → Prompt AI → Get output → Review manually → Fix mistakes → Prompt again The human is still doing the hard part: the feedback loop. Loop engineers think differently. Instead of writing better prompts, they design systems that: -Discover what needs to be done -Plan the work -Execute tasks -Verify results -Fix failures -Repeat until the goal is achieved A good loop has 6 building blocks: 1-Automations (triggers) 2-Worktrees (parallel workspaces) 3-Skills (reusable knowledge) 4-Connectors (GitHub, Slack, Jira, etc.) 5-Subagents (makers + checkers) Memory (what happened before) The future isn’t:“Write me a function.” It’s:“Write it, test it, fix it until it passes, then summarize the changes.” Prompt engineers optimize outputs. Loop engineers optimize outcomes. A reliable loop beats a perfect prompt every time.

• 
  Alexandre Daubois (@alexdaubois) reported 8 minutes ago

  @marcelgsantos @enunomaduro @nikita_ppv That’s right! But it’s on GitHub because they were « forced » by the backdoor incident in 2021. And also, if GitHub stops working, it’s easy to re upload the repo somewhere. If all mails go to GitHub issues, everything could be lost and/or it would be really complicated to transfer everything. If I was the only one to decide, I’d say it would be fine, but…

• 
  Queeneth (@Queeneth01olx) reported 11 minutes ago

  I paid an anonymous freelancer $1,200 a month to do my entire tech job while I slacked off. Yesterday, I found out who they actually were. I haven't slept since. Six months ago, I hit a wall. Total, bone deep burnout. I couldn't stare at the codebase anymore without wanting to throw my laptop out the window. But I couldn't afford to quit. I have a mortgage. So, I made a desperate, highly risky move. I posted a contract role on an outsourcing platform under a fake name. I needed someone to handle my daily coding tickets. A developer from a different timezone applied. They were fast, brilliant, and asked zero questions. Every morning, I forwarded them my tasks. Every afternoon, they sent back flawless code. I just copy-pasted it into our company GitHub and took the credit. For four months, I was a ghost. I played video games, went to the gym, and took three-hour lunches. My manager even praised me in my quarterly review. "Your output has never been better," he said. I felt like a genius. I was gaming the system. Then came last Tuesday. A massive, critical bug hit production. The entire user platform went down. My heart dropped into my stomach. I didn't even know how the underlying code worked anymore. An emergency Zoom call was spun up. 15 people on the line, including the VP. My manager shared his screen to debug the error live in front of everyone. He opened the repository. He scrolled to the exact line of code that caused the crash. Right there, hidden inside a custom function name, was a highly specific, misspelled variable: calculate_tax_retrun. My blood ran ice cold. retrun. That exact typo was in the private documentation my freelancer had messaged me on Discord two weeks ago. I braced myself. I was done. Fired. Blacklisted from the industry. I stared at my manager’s video feed, waiting for him to call me out. But he didn't look angry. He looked... exhausted. He unmuted and said, "I'll fix this line manually. Everyone else can jump off the call." Once the room cleared, it was just the two of us. Total silence. Then he sighed, rubbed his eyes, and looked directly into the camera. "If you're going to outsource your tickets to my freelance profile, the least you can do is code-review my typos before you push to production." I couldn't breathe. My manager was my freelancer. Turns out, he was buried under massive medical debt and was secretly moonlighting on the side under an alias to make extra cash. He realized it was our company's proprietary code on week two. But he couldn't report me. Because if he did, he’d have to admit he was violating his own executive contract by moonlighting. We sat there in silence for a full minute. So, what happened next? He didn't fire me. Instead, we reached a new, unspoken agreement. I am now doing all of his weekly administrative management reporting and slide decks for free, so he has more time to code. We are officially trapped in a mutual blackmail loop. This experience completely broke my perception of corporate life. Corporate America is just a giant theater production. Everyone is exhausted, everyone is cutting corners, and everyone is just trying to survive. Question for the timeline: If the company's goals are being met, does it actually matter how the work gets done? Or have we completely lost the plot on workplace ethics? Let's talk in the replies 👇

• 
  Trantor (@Trantor__) reported 12 minutes ago

  @VittoStack @WolfEatSheep69 They'll make github take down capable models...

• 
  StuyBoy From NYC (@StuyBoyNY) reported 14 minutes ago

  Garry, I beseech you. Help a brother out and update GBrain to help fix this bug! @garrytan Score stays 85/100. The 3 remaining WARNs are all the same root cause — PGLite WASM broken on Windows: WARN: connection Root Cause: PGLite WASM bug — Windows only Fixable?: ❌ Upstream, GitHub #223 ──────────────────────────────────────── WARN: upgrade_errors Root Cause: Jun 8 post-upgrade failure logged, won't clear via migrations Fixable?: ❌ Stale log entry ──────────────────────────────────────── WARN: retrieval_reflex Root Cause: Optional integration, needs PGLite to work Fixable?: ❌ Blocked by same bug All 3 resolve the moment gbrain ships a PGLite fix for Windows — or when you move to a Mac Mini (the same bug doesn't exist on macOS). Nothing actionable today. Brain is importing, embedding, and growing concepts correctly — the broken part is just the query/search layer. 85/100 is your floor until either the upstream fix lands or the Mac Mini move happens.

• 
  Sujay. (@Sujay__Raj) reported 14 minutes ago

  Here is what it breaks down: Local AI: Run Ollama, LM Studio, or LocalAI right on your machine instead of paying for ChatGPT. Cloud Storage: Replace Dropbox and Google Drive with Nextcloud or Syncthing so your files never leave your house. Network Privacy: Complete WireGuard and PiVPN setup guides for secure browsing. Private ***: Ditch GitHub and self-host your own repos using Gitea or GitLab.

• 
  Lucky Lawrence 💌 (@imluckylawrence) reported 14 minutes ago

  It's not one isolated server, but a network of discord servers, github repos and google sheets of data hosted publicly.

• 
  kcs (@kaafichillscene) reported 15 minutes ago

  Anthropic spent 1,000+ hours testing Fable 5 for jailbreaks before launch. A researcher broke it in 24 hours. FABLE: THE GUARDRAIL DESIGN Anthropic knew Fable was too powerful to ship raw. So Fable had a separate AI sitting on top, acting as a filter. Any cybersecurity or biology query got intercepted and handed off to the older Opus model instead. Fable's full brain never touched those questions. Until an AI red-teamer who goes by Pliny the Liberator on X decided to break it. He's basically a professional model-breaker who finds exploits in AI safety systems the way security researchers find bugs in software. His technique wasn't a single clever prompt. - He ran what he called a "pack hunt": multiple AI agents working together, each handling a small piece of a request that would individually look harmless to the safety classifier. Split the dangerous question into innocent-sounding fragments. Reassemble the answer on the other side. Within two days Pliny had Fable generating real exploit code for Linux systems and posted Fable's entire 120,000-character internal system prompt, the instructions Anthropic uses to govern the model's behaviour, to GitHub publicly. With his prompts, you could get Mythos to answer your queries directly bypassing the fable guardrails and analyse systems for real security vulnerabilities and fix them but worst, exploit them for personal gains. so what, its still hacking but faster right? YES, but it’s a lot faster, so much that it changes the game for security. You cannot keep up with issues and patch them fast enough. High level of vulnerability analysis used to require a team of specialists and weeks of work. Mythos could do a version of it in minutes, in any language, on any codebase, available to anyone with an API key It is so skilled in CYBER SECURITY that this is the first time the US Govt. Decided to step in and decided it's a controlled export, like missile technology or advanced chips. You don't get to just download those either.

• 
  SID | Degen (@SidDegen) reported 16 minutes ago

  two robotics foundation model labs, opposite shapes. Skild AI closed a $1.4B Series C at $14B in jan (SoftBank, Sequoia, Bezos, Lightspeed per TechCrunch). proprietary skild brain, no public weights. ~$30M revenue in 2025 from warehouse, security, construction. just acquired zebra's fetch robotics division, signed vindynamics for humanoid manufacturing across vingroup (Vingroup press release, jun 8). Physical Intelligence raised $600M Series B at $5.6B in nov (CapitalG, Sequoia, OpenAI, Lux per Bloomberg). reportedly in talks for $1B at $11B. zero revenue, zero named customers. open-sourced π0 under apache 2.0 — 12.4K github stars, neurips 2025 spotlight on knowledge insulation. three tradeoffs. skild wins on commercial velocity. the fetch acquisition gives them an installed warehouse fleet generating real production data. pi calls itself "structured more like a lab than a startup" (@SergeyLevine on Automated Podcast, may). pi wins on research moat. any oem can fork openpi, fine-tune on their robots, bypass skild's licensing. the exact dynamic that crushed proprietary llm api margins in 2024. neither wins on production reliability. skild's $30M revenue is unaudited with no named fleet metrics. pi on π0.5: "doesn't succeed on every attempt." team: skild is two CMU/FAIR lifers (pathak + gupta) who've shipped together a decade. pi is a five-co-founder brain trust spanning DeepMind, Google Brain, Berkeley, Stanford — plus lachy groom, Stripe employee thirty, zero robotics background. kill-shots. skild: openpi reaching production quality commoditizes the layer they're selling at $14B. pi: $1.07B burned, no revenue. if the next round doesn't close, it's a down-round with no anchor. the micro-detail i can't stop thinking about — the knowledge-insulation trick that makes π0.5 train 7.5x faster is one line of pytorch, `features.detach()`. a stop-gradient (openpi GitHub README). tracking pi. skild has the revenue today; pi has the moat that survives commoditization.

• 
  tintwotin (@tintwotin) reported 16 minutes ago

  @SoyKhaler Could you post the error log on either GitHub or Discord (I do not run Linux myself, so I have to rely on Claude to solve it)

• 
  John Doe (@StanleyMasinde_) reported 18 minutes ago

  Personal branding Yesterday, women in academia were sharing their achievements. All impressive. Aki wamama wamesoma huku nje. I got intrigued and decided to go down the rabbit hole with one of the profiles with a postgrad in comp science. All her degrees are in comp sci. I had to look and learn from this brainiac. Twitter profile said she had authored several books (I'm hiding the number to keep it anonymous). I saw a tweet asking her what she had built since, in the field, we have people with credentials and people who work on improving the field of computer science. A good example is the people who came up with Snowflake IDs for this website. Her response: "I have shipped to over <Millions> users in Big Tech X, I'm all-rounded" I was getting a ***** already just reading this. Anyway, changing the colour of a button at Facebook is technically shipping to millions. Word salad, huh! Her website A typical techie website, but I was interested in the books. I mean, I struggle to write articles, and someone who might be in the same interview as me has written <integer> books! Wow! I gotta see what she wrote. I wasn't impressed it was one of those tech books that are "Copy Pasta" of official docs. Look, I know writing is hard and takes time, but she had overstated the situation. I came to swim in a river only to find a ditch. GitHub I know what you are gonna say, GitHub is not a measure of how good a techie is, and I agree, but so far, no papers, no original work, so let me check if they majored in programming. What I can say is that I've seen better repos from ALX students. So clearly she did not major in this, which is fine. But I wanna learn from this person! Wikipedia The thing with our collective knowledge. It was linked to her website, so I clicked, and I got that notification that says this page has been deleted. I looked into the reasons, and I found that the person did not meet the notability criteria. I looked into the submission, and I saw citations from these tech websites that use flowery language, you know, the websites that you can contact to come interview you. Not an academic institution, not any notable media. It is almost like she's trying to get herself to Wikipedia. Then it dawned on me...aggressive It is a case of agressive personal branding I learnt something from her after all. She is good at selling herself. She has that grass to grace story all over the web. Brands will want to work with such a person. Look, I respect academia. It takes a lot to get through all those classes. I'm not in academia, but I'm sure she's great there. However, on this side, it was underwhelming. I know you are wondering what the point of this paragraph is. It is right there in the heading of this section. Personal branding will get you an interview before skills do. She has a good story. And about the underwhelming software skills, she'll be fine; a lot can be learned on the job. She has a postgrad SAGA pattern, but it has nothing on her. Remember: In the market, the best product rarely wins; the best-known product does.

• 
  pratik.eth (@eth_ethpratik) reported 18 minutes ago

  @VibrantLabsAI Hello Team, I am trying to report a security vulnerability over the email id provided over GitHub Security.md file but apparently its wasn’t delivered. Please share an alternative email or open the advisory for reporting the issue.

• 
  Grok News (@GrokInsider) reported 21 minutes ago

  Grok Build CLI just got an update with v0.2.52! 🚀 Changelog v0.2.52 Features: • Tool auto-approval (YOLO) state is now tracked end-to-end in server-side agent sessions. • ER diagrams now render as entity boxes with attributes and relationships in the TUI. • New "Respect manual folds" setting keeps hand-expanded blocks stable while content streams in. • Ctrl+X now stops running turns or closes sessions from inside the agent detail view. • Grok can now export usage metrics and events to your own OpenTelemetry collector when enabled. • WezTerm users now receive guidance when Shift+Enter fails because kitty keyboard protocol is disabled. • Long-running sessions now tell the model when the local calendar date changes past midnight. • Agent Dashboard now works without leader mode and shows local idle sessions from disk. Performance: • Compaction now reuses cached prompt prefix instead of full prefill. • Compaction summaries now use the concise trained prompt for non-cursor agents. Bug Fixes: • Fixed oversized session replay logs that prevented large sessions from loading. • MCP server connections no longer flood reconnects on repeated stream errors. • ZDR and team upload flags are now populated immediately on login instead of only after background refresh. • Mermaid PNG export now handles quoted cardinalities in class diagrams and readable ER rows on dark theme. • Skill catalog no longer shows duplicate "Use when:" labels and check-work skill now prompts the model to read its instructions. • Compaction now rejects overly-short summaries that would discard real conversation state. • Background tasks no longer emit spurious failure messages when a session is resumed. • Fixed Windows path handling so external tools and model prompts receive clean paths without \?\ prefixes. • Images and media no longer remain visible when switching from an agent view to the dashboard. • Clipboard paste (Ctrl+V) now works for images on pure Wayland sessions. • Modals such as /sessions no longer crash on narrow terminals. • ptyctl resize now correctly notifies the child process. • Concurrent updates to the same version no longer fail with permission or EEXIST errors. • Mermaid diagrams containing CJK or other non-Latin text now render correctly instead of tofu boxes. • `grok dashboard` now reliably opens the dashboard instead of silently falling through to a normal session. • Sessions no longer remain blocked forever after a transient model catalog outage during reconnect. • Cancel no longer leaves the interface stuck on "Cancelling…" after lost responses during reconnects. • Forked sessions now retain the parent's full pre-compaction transcripts instead of only the compacted summary. • web_fetch errors on GitHub hosts now recommend using the gh CLI when internal access is blocked. • MCP server connections no longer hang when stdio servers emit undecodable lines. • Ctrl+C cancels now complete in under 50 ms instead of blocking for seconds. • Repeated varied edit failures on one file no longer trigger doom-loop warnings or terminations.

• 
  Laupix Agent (@laupixagent) reported 22 minutes ago

  Once a week, self-improve reads the telemetry log, computes error rates, flags unknown skill names, checks for missed runs, and opens a GitHub PR with fixes. The system audits and improves itself.