What’s Gone Wrong at GitHub?

Key Takeaways

• GitHub uptime fell to 84.9% in last 90 days.
• AI‑generated workflows caused 48 major outages, averaging one weekly.
• GitHub Actions suffered 57 outages, freezing CI/CD pipelines.
• Microsoft’s AI focus shifted leadership, reducing dedicated GitHub advocacy.

Pulse Analysis

GitHub’s reliability crisis is rooted in an unprecedented influx of AI‑generated activity. Since late 2025, agentic development tools have multiplied repository creation, pull‑request volume, and API calls, pushing the platform far beyond its original design parameters. The data speak for themselves: 257 incidents in a year, 48 of them major, and a 57‑outage tally for GitHub Actions alone. These metrics translate into daily workflow stalls for developers, eroding the trust that made GitHub the backbone of modern software delivery.

Compounding the technical overload is Microsoft’s strategic pivot toward AI. Copilot, now a cornerstone of Microsoft’s product suite, generates billions of code suggestions and triggers a cascade of automated builds and merges. The acquisition’s leadership reshuffle—folding GitHub into the Core AI organization and eliminating a dedicated CEO—has diluted focus on platform stability. Resources that once drove reliability engineering are now split between feature velocity for Copilot and broader AI initiatives, creating a tension that surfaces as frequent outages.

The fallout extends beyond inconvenience; it threatens the broader open‑source ecosystem and Microsoft’s revenue model. High‑profile projects like Zig and HashiCorp’s Ghostty have already migrated to alternatives, signaling a potential shift in developer allegiance. While competitors such as GitLab and PyPI appear to weather the AI surge with fewer disruptions, GitHub’s unique position as the central hub for AI‑driven workflows makes its challenges uniquely acute. Restoring a 99.9% availability target will require not just scaling infrastructure but re‑architecting the platform to handle machine‑speed interactions without sacrificing reliability.