Days since last GitHub incident

Original Article ↗ Hacker News Discussion ↗

Overall reaction to GitHub instability

  • Several users noticed the outage before the official status page, citing failed releases, Actions failures, and “unicorn” error pages.
  • Some now reflexively assume CI failures are GitHub’s fault rather than their own, and argue stability should be prioritized over AI features.
  • Others feel outages are frequent enough that internal discussions have begun about moving away from Actions, Packages, or GitHub entirely, describing the platform as “decaying.”

The “days since last incident” site and humor

  • Many found the site funny and perfectly minimal, with some amused that it works even offline due to being static.
  • Others criticized it as low-effort and wished for a more elaborate meme (physical-style accident sign, octocat gags, “days without accident” templates, AI jokes).
  • A few users complained about design/usability (text too small, looks blank on phones).

Reliability, “incidents,” and SLAs

  • Some argue the counter is misleading because minor or obscure component outages reset it; others respond that what’s “trivial” varies by user.
  • Discussion touches on uptime expectations: not everyone needs “five nines,” but even short outages can be painful when they block CI, container registry pulls, or payments.
  • Users point out registries and artifact services can be single points of failure, even if read-only mirrors are conceptually simple.

Alternatives, mirroring, and decentralization

  • Suggestions include GitLab, self-hosted GitHub Enterprise Server, mirrors for dependencies (e.g., via Nixpkgs), and decentralized/p2p forges like radicle.
  • Some say moving off GitHub has high friction due to network effects and mindshare; others share negative GitLab experiences or say uptime is similar.
  • Mirroring source is seen as practical; replicating Actions, issues, registries, or Copilot is harder.

AI features and local vs cloud setups

  • One user describes heavy reliance on GitHub’s agents/Copilot for reviving old projects and is frustrated that this increases exposure to downtime.
  • Self-hosted GitHub Enterprise is mentioned but noted to lack Copilot APIs.
  • Multiple commenters explain that local LLMs still lag hosted “frontier” models; local hosting is framed as useful mainly for privacy/hobby use, not as a seamless Copilot replacement.
  • Discussion branches into hardware (Apple Silicon vs NVIDIA boxes) and building one’s own agent/tooling stack, with some arguing that investing in custom tools has high long-term ROI.

Quality issues: Actions and UX

  • Users list odd GitHub Actions behaviors (stuck jobs, inconsistent status, phantom PR badges) as evidence of brittle internals.
  • There is broader criticism of Microsoft-era quality and a sense that “AI everywhere” has crowded out core product polish.
  • Separate complaints focus on unremovable crypto spam notifications; workarounds via the GitHub CLI and a recent backend fix are shared.

IPv6 and networking

  • Some argue GitHub’s lack of IPv6 in 2025 should count as a permanent “incident”; others say residential IPv6 penetration is still too low for it to be business-critical.
  • Brief side discussion covers ISPs, router firewalls, and cloud providers that price IPv4 separately from IPv6.