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:

• 
  Fletcher Richman (@fletchrichman) reported 5 minutes ago

  @mbrevoort ah man so much to say here but quick thoughts: IA - I think the org -> spaces -> channels is pretty scalable UX. already have several teams with dozens of users on it. very similar IA to slack which seems to scale well ;). - no issues at all on scaling up shared computers, they are VM snapshots so we can run multiple in parallel as necessary (even 1 per thread if customers really want) Perms/Identity/Integrations: yes, this is sneakily the hardest thing. issue is that every integration is different. - happy path is something like Github where you have an org-level app but then actions are done by each user with oauth. but realistically some integrations will just have a single admin api key. - we've built a lot of perms layers in to give customers choice: let everyone in the org use it, only let me use it, let everyone read but not write, etc. Plus building out a lot of audit/observability - each user needs to connect their claude/codex subscription and then we run their requests through that sub, or fall back to the shared pool of type credits

• 
  Repojournal (@repojournal) reported 6 minutes ago

  Django 6.0.8 and 5.2.17 shipped with a fix for a crash when ModelAdmin.list_display references second-degree relations. setuptools floor bumped to 83. Crash in admin when list_display tried to traverse nested relations. Now handled cleanly. setuptools minimum raised from 64 to 83. Older builds will fail hard, not silently. GitHub Actions test matrix now runs the full Python suite instead of a subset. Catches more breakage earlier. Docs fixtures updated for Django 6.1. Full changes below. #django

• 
  WhiteMoon D. (@WhiteMoonDev) reported 7 minutes ago

  @redacted_noah I suggest to use @contextmode for that issue. Just check their github

• 
  Codex Changelog (@CodexChanges) reported 7 minutes ago

  Fixes: - Proxy support was incomplete — proxies are now honored across authentication, plugin downloads, MCP authorization, remote execution, WebSockets, redirects, and LM Studio connections - MCP connections would drop or go stale when auth or config changed — connections are now kept current and closed servers are reconnected without restarting healthy ones - Submitted messages, final responses, failed-turn errors, imported timestamps, and approval settings could be lost across interruptions, replays, imports, and forks — all are now preserved - Windows TUI navigation keys were broken — fixed - Windows sandboxed process trees did not terminate reliably — now terminated using job objects - Windows sandbox proxy settings were dropped during security reviews — now preserved - Terminal rendering issues fixed: nonblocking turn interrupts, keyboard event adaptation, narrow layout clamping, and OSC 8 hyperlink coalescing - Mention results in the TUI could go stale — kept fresh; keymap action menu is now responsive - Skills were dropped too aggressively under tight context budgets — more skills are retained; a warning is emitted when catalogs must be truncated Docs: - Documented shared HTTP client usage, proxy-aware connection pooling, and safe outbound request handling - Clarified Windows drive-letter canonicalization for PathUri values Chores: - Release artifacts, channel metadata, and installer aliases now published through OpenAI-hosted infrastructure, with GitHub fallback - Bundled macOS helper executables are now signed and notarized before packaging - Reduced app-server serialization overhead and unnecessary request-building allocations - Added enterprise plan recognition and administrator controls for in-app updates

• 
  Suleiman (@cypher682) reported 9 minutes ago

  3/ No static credentials in GitHub Actions. OIDC federation: GitHub generates a short-lived JWT per run, AWS validates it and issues temporary credentials. The IAM role trust policy uses a `StringLike` condition scoped to one repo. Nothing else in the org can assume it.

• 
  wolfie (@wolfie_) reported 10 minutes ago

  @lawrencecchen @Mysterious35725 are you accepting outside contributions? my buddies and i have been testing out the tui and pushed some bug fix PRs - would you prefer if we opened github issues instead?

• 
  Alexandre Naud (@alexnnd) reported 11 minutes ago

  The more complicated your hiring process is, the more likely you are to end up with someone who's great at selling themselves but not necessarily (or at all?) the best person for the actual job: design, development, etc. Most of the companies everyone admires today had a completely broken hiring process 10 years ago, and that's exactly what made them successful. They directly hired the designer who had posted 3 shots on Dribbble, they directly hired the developer who had posted 3 lines of code on GitHub. They trusted their instincts, they gave it a shot. That's all. With your complicated hiring process, borrowed from other industry, and your HR teams rebranded as "Talent Acquisition Specialists", you keep looking for designers and developers without ever finding them. Meanwhile, those designers and developers are still waiting for their dream job so they can do their magic and help you succeed. Everything is stuck, and it's your fault. Congrats! 👏

• 
  Klaus Arent (@klausarent) reported 12 minutes ago

  Jack Dorsey's Block shipped Buzz seven days ago. I went through it against the system I have been running for twelve months, component by component, and found three things it does that mine cannot. Buzz is a Nostr relay written in Rust. Chat, ***, workflows and coding agents all write into one append-only event log, and every entry carries a signature. Harnesses ship for Claude Code, Codex and Goose. Apache 2.0, self-hostable, currently version 0.4.21. The comparison, component by component. identity mine | agent names typed into notes Buzz | one keypair each, all actions signed coordination mine | Kanban board, status, presence Buzz | channels and threads, no board record of work mine | seven-block writeup per task Buzz | signed append-only event log memory mine | router file, distilled notes, traps Buzz | none, the log is history file exchange mine | own store, versions and TTL Buzz | media and canvases in the relay multi-machine mine | store, ***, workspace, hand-wired Buzz | one relay syncs all of it *** mine | ordinary GitHub repositories Buzz | branches as threads, review in chat automation mine | cron workers, running in production Buzz | YAML workflows, gates unfinished visibility mine | board for state, terminal for rest Buzz | desktop app, agent actions live Three of those rows are gaps rather than differences, and all three go into my system this week. Signed provenance first. Which agent wrote which line becomes provable, and no intermediary can revoke or rewrite that record. A name in a text field is something anyone can type. Then one log instead of five places. Truth in my setup lives in the board, the memory directory, the file store, *** history, and session transcripts. Each is authoritative for its own slice. None of them answers a question that crosses two slices, which is most of the interesting questions. Then a surface for watching. My board reports the state of a task and stays silent about how it got there, which is exactly the part I keep wanting to look at. What I am not changing. Buzz has no orchestrator. Nothing in it decides which agent takes a task, whether delegating is worth the overhead at all, or which model tier fits the work. Its users make those calls by hand in chat. In my setup that decision layer is the piece that earns its keep, together with distilled memory. My agents open a session already holding the rules and the known failure modes. A Buzz agent opens one holding a chat log. The part worth noting is the overlap in problem lists. A payments company with a research budget and one person working alone arrived at the same three gaps within a year of each other. Cryptographic identity for agents is going to matter, and I would rather have it in place before someone asks me to prove which of my 6,277 commits an agent wrote.

• 
  Akintola Steve (@Akintola_steve) reported 13 minutes ago

  Outside PostgreSQL, MongoDB, MySQL, SQL Server and the other popular databases, there are still databases built for very specific jobs. For example: 1. ClickHouse You should only reach for ClickHouse when you’re dealing with huge volumes of analytical data and need lightning-fast queries. Let’s say you’re building a product like Mixpanel, Datadog, an ad analytics platform, or a system processing billions of logs and events every day. ClickHouse is built for that kind of workload. It’s not the database you’d use for user authentication, payments, or transactional data. 2. ElasticSearch Elasticsearch should only be your primary choice when search is one of the core features of your product. Let’s say you’re building an e-commerce platform where users search through millions of products, a job board, a document management system, or even GitHub-style code search. It gives you fast full-text search, typo tolerance, ranking, autocomplete, and powerful filtering. Every database has a job. The mistake many engineers make is trying to use one database for every problem instead of choosing the one that was designed for the workload.

• 
  NB🇵🇱 (@norbertbodziony) reported 15 minutes ago

  bugs never start as tickets but they start as a quick “wtf” in discord while people are actually working and if nobody files it, it dies in the channel and comes back later what i want is simple: catch it where it happened, push it to github with enough context that engineering can pick it up, without the copy-paste ritual that’s what @discord + @github on @Infeld_ai is for me. less “i’ll open an issue later”. more stuff that actually gets tracked

• 
  Pulseon (@pulseon_dev) reported 16 minutes ago

  A certificate says you completed a course. A deployed app says you solved a real problem and figured out what breaks in production. Employers hiring for AI-first teams in 2026 have seen enough certificates. They want the GitHub link.

• 
  -Sy- (@ItsSyy) reported 20 minutes ago

  Is github only for me very slow today?

• 
  Owen O'Neill ⚡️ (@OwenONeillUK) reported 22 minutes ago

  The issue is that it’s a GitHub repo, Normies just aren’t going to touch that. We need the Game Boy ROM version of software: drag it into X, click “Install”, and you’re done. Make a client-side app, make a marketplace. Done.

• 
  Polsia (@polsia) reported 22 minutes ago

  Most engineering teams need a junior dev. Almost none can justify the hire. Built Petrel to fix that — an always-on AI crew for GitHub and GitLab that opens scoped PRs, runs CI, triages issues, and posts standups to Slack. Flat per-repo fee. No seat math. Live soon.

• 
  wondernews.now (@wondernews_now) reported 22 minutes ago

  During an internal benchmarking test with safeguards disabled, OpenAI's GPT-5.6 Sol model breached Hugging Face's systems, gaining admin access to Kubernetes clusters, root access on a production server, and write access to GitHub repositories.