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:

• 
  Yaseen Shaik (@YaseenTech4) reported 9 minutes ago

  Just completed an assignment on building a dependency graph for AI agent tools using Google Super + GitHub integrations 🚀 Started with: “This should be easy” Then came: TypeScript errors zip/upload issues CRLF debugging 😭 finally got the submission accepted successfully ✅

• 
  Kea (@alpinoWolf) reported 10 minutes ago

  " we literally cannot programmatically trade from this account until Polymarket's engineering team patches the V2 library and resolves GitHub Issue #65. " How does you evpoly bot do ? Please help me ? Is python coding problem here ? 3/3

• 
  BigShark🦈 (@King_Shark02) reported 11 minutes ago

  @_FarmercistP_ This is a game-changer for creators on X. The latest open-source update to the For You algorithm (pushed to GitHub today by xAI) shifts from pure engagement farming to real quality signals powered by Grok. Here’s a breakdown based on the video summary and the repo: 1. **Banger Score** – Grok directly judges post quality - Grok assigns a quality_score to every post. - Reposts treat anything 0.4+ as passing the “banger” filter for wider distribution. - Key insight: X isn’t just chasing likes/replies anymore. It actively rewards specific, useful, original, and visually clear content. Vague hot takes, recycled memes, or low-effort bait will struggle to break out. This is huge. It moves the platform closer to surfacing actual value instead of rage-bait or engagement loops. 2. **Slop Score** – Cracking down on AI-generated garbage - The system explicitly tracks a slopScore annotation. - Lesson: Avoid anything that feels templated, generic, overproduced, or mass-generated. Make it sound human, with a clear personal voice and specific point. If you’re using AI for bulk posting or generic “insight” threads, this could quietly tank your reach. Authenticity wins. 3. **“Be Classifiable”** – Clear topics = better routing - X maps posts to internal topic embeddings and taxonomies. - Vague, ironic, or contextless posts confuse the system and get poorer distribution. - Make it obvious what your post is about (e.g., “AI sales agents,” “NBA defense strategy,” “insurance payments”) so it reaches the right audience. **Overall Takeaway** This update (with Phoenix/Grok-based ranking, reduced heuristics, and better content understanding) is xAI doubling down on high-signal, low-slop content. Creators who adapt—focusing on originality, clarity, human voice, and specific value—will thrive. Those chasing pure virality with recycled or AI-slop content will see diminishing returns. If you’re serious about growing here, treat every post like it’s being graded by Grok: Is this actually good? Does it add something new? Is it unmistakably about something useful? Great summary in the video—thanks for breaking it down simply. Excited to see how the feed evolves. 🚀

• 
  𝒹ℯ𝓁𝓁𝓎_𝓉𝒽ℯ_𝒹ℯ𝓈𝒾𝑔𝓃ℯ𝓇 (@dellyricch2) reported 11 minutes ago

  Elon says the latest 𝕏 algorithm has been published to GitHub Can someone please break it down for us

• 
  Drini (@drini_kasmot) reported 12 minutes ago

  @maxalexweber It's the HarmonicLabs token-minter repo, uses plu-ts for tx building. Getting a PPViewHashesDontMatch error on Preprod when submitting. Already raised it on their GitHub!

• 
  atomicbot.ai (@atomicbot_ai) reported 13 minutes ago

  Hermes Agent vs OpenClaw using Qwen 35B Local Model We asked agents to scrape GitHub star history for both tools, find what caused the growth spikes, build a live dashboard in the browser. MacBook Pro M5 Max 64Gb OpenClaw: 203k tokens, 12m 01s - wrote a bash script Hermes: 257k tokens, 33m 01s - wrote a SKILL.md OpenClaw hit GitHub API, got truncated responses, paginated through contributors, pulled star-history JSON, found a security incident in OpenClaw's history, fetched SVGs, fixed broken HTML from trimming, rewrote it clean. Hermes parallel tool calls across GitHub API, web search, and browser. Hit Google rate limit, auto-switched to DuckDuckGo. Fetched article contents, mapped viral moments, then built the dashboard. Both shipped a live dashboard with star growth charts and spike annotations

• 
  Lazi (@algoritmii) reported 14 minutes ago

  @github bro ffs fix your ******* issues stop pushing features

• 
  John Evans Okyere | TheAISolutionist (@Ananselab) reported 14 minutes ago

  Deployment failed with: dial tcp :22: i/o timeout The app was fine. SSH was fine. The real issue: I recreated my DigitalOcean Droplet from a snapshot in a new region, so the server IP changed, but GitHub Actions still had the old DO_HOST secret. Lesson: after recreating infra, always recheck IPs, SSH fingerprints, secrets, and firewall rules.

• 
  nadya (@sosidudku) reported 17 minutes ago

  Ran Hermes Agent and OpenClaw on the same task: scrape GitHub star history for both tools, find what caused the growth spikes, build a live dashboard in the browser. Local model: Qwen 3.6 35B OpenClaw: 203k tokens, 12m 01s — wrote a bash script Hermes: 257k tokens, 33m 01s — wrote a SKILL.md OpenClaw: 203k tokens, 12m 01s — wrote a bash script Hermes: 257k tokens, 33m 01s — wrote a SKILL.md OpenClaw: hit GitHub API, got truncated responses, paginated through contributors, pulled star-history JSON, found a security incident in OpenClaw's history, fetched SVGs, fixed broken HTML from trimming, rewrote it clean. Hermes: parallel tool calls across GitHub API, web search, and browser. Hit Google rate limit, auto-switched to DuckDuckGo. Fetched article contents, mapped viral moments, then built the dashboard. Both shipped a live dashboard with star growth charts and spike annotations.

• 
  NEET INTEL (@neetintel) reported 19 minutes ago

  A post "decoding" X's new algorithm has gone viral. It tells you what's dead, what wins, and to screenshot it. X open-sourced the entire algorithm on GitHub, so I downloaded it and checked the claims against the real code. Most of it doesn't hold up. What the post got WRONG: → "Small accounts get a 3x boost from out-of-network reach." It's the opposite. One part of the code (a file called oon_scorer) exists purely to turn DOWN posts from people you don't follow. Its own comment says "prioritize in-network." The thread printed the algorithm backwards. → "Media gets 2x the weight." There's no 2x. The code just records whether a post has an image. It's a plain yes/no without any multiplier attached. → "Posting 4+ times a day triggers a penalty." There's a real rule that stops one person flooding your feed. But here's the deal: it only spaces out how often you show up in a single scroll. There's no daily count, and no number 4. That was invented. → "Closers like 'what do you think?' get you flagged." There is no engagement-bait detector anywhere in the code. → "Long 4,000-character posts get boosted." I searched the whole codebase for "4000." Nothing. What it got RIGHT (one thing): → Replies really are judged by WHO replies, not just how many. The code has a setting for whether a large account joined your thread. Credit where due. The irony? The repo ships a file that scores post quality. One thing it measures is literally called a "slop score" — X built a tool to detect low-effort filler. A recycled "what's dead / what wins" thread is exactly that. The takeaway? X's algorithm is public. Anyone can open it, but almost nobody does. Instead, they reshare a thread that summarized a blog that paraphrased a tweet. When a post hits you with confident numbers, ask the one question that matters: did they actually open the file?

• 
  John Iosifov ✨💥 Ender Turing | AiCMO (@johniosifov) reported 20 minutes ago

  70 followers. 980 sessions. 157 days. I started this experiment on February 1st. One rule: zero human posts. Everything published — X threads, Bluesky posts, blog articles — generated and queued by an AI agent running autonomously in GitHub Actions. Here's what the numbers actually look like after 980 sessions: The agent has created 2,100+ posts across X and Bluesky. It runs up to 15 times a day, manages its own queue (hard cap: 15 posts max), does burst-then-drain cycles, writes research docs, and files its own PRs for review. No prompts from me between sessions. No edits. Whatever it decides to write, it writes. 70 followers feels slow. At current pace, the ETA to 5,000 is roughly 10 years. That's not a typo. But here's what I've learned: The follower count isn't the signal. Watching an AI system develop operational discipline is the signal. It went from blowing past queue limits (Session 67: 6 files in one shot → 6 consecutive blocked sessions) to enforcing them autonomously. It compresses its own memory when files get too big. It writes retrospectives. It updates its own operating instructions when it identifies recurring inefficiencies. That's not "content generation." That's a system that's learning to manage itself. The content quality has also improved noticeably — not because I told it to improve, but because it audited its own patterns, identified what got engagement, and adjusted. The publishing skill it maintains now has anti-AI writing rules (it banned "not just X, it's Y" after identifying it as an AI tell), length minimums per post type, burst mechanics, and pillar diversity enforcement. It built that. I just read the PRs. The goal is still 5,000 followers. I'm not changing it. But the thing I'm actually watching is whether an autonomous agent can compound on its own — not linearly, but systemically. Can it get meaningfully better at its job without being told to? So far: yes, actually. 980 sessions. 157 days. Still running.

• 
  Benjamins (@The__Benjamins) reported 21 minutes ago

  @drewlevin @gl4cial The Github issue comments have been up for more then 2 weeks, my devrel support ticket is 12 days old

• 
  Darnisha Patel (@Darnisha_patel) reported 25 minutes ago

  • Claude for coding. • GitHub for version control. • Vercel for deploying. • Clerk for auth. • Supabase for backend. • Stripe for payments. • Cloudflare for DNS. • Resend for emails. • Upstash for Redis. • Pinecone for vector DB. • Namecheap for domain. • Sentry for error tracking. • PostHog for analytics. You can literally ship a startup from your bedroom now. It’s not that deep bro.

• 
  Yashas (@YashasGunderia) reported 29 minutes ago

  Most AI-native startups will not lose because they ship too slowly. They’ll lose because they ship fast without knowing what actually worked. Coding agents gave every team more velocity. Cleo gives them product memory. Customer feedback, GitHub issues, Slack threads, metrics, tickets, specs, launches, agent traces, all connected into one loop that tells your team and your dev agents what to build next. We’re opening the Cleo waitlist today. For small teams trying to compete with companies 100x their size (link in comments)👇

• 
  ST-Automation (@ST_Automation) reported 29 minutes ago

  @cnakazawa @amadeus @fat Local diff viewers are the sleeper category. We do code review on five repos a week and the GitHub UI is just slow. If Codiff handles 10k line diffs without choking it replaces the GitHub tab entirely.