April 28, 2026
Outages, apologies, and Pi-powered shade
An Update on GitHub Availability
GitHub vows big upgrades as commenters roast Azure and praise Pi power
TLDR: GitHub apologized for recent outages and says it’s scaling up massively and even considering multiple clouds to handle a surge from AI-driven coding. Commenters clapped back with Azure skepticism, Microsoft distrust, and jokes that a Raspberry Pi at home feels faster—while others warned the AI code flood is just starting.
GitHub just dropped an apology-and-action post about two recent outages, promising “availability first” and a massive scale-up after usage surged. The company says AI coding agents are flooding projects with new repos, pull requests, and automated tasks—so they’re beefing capacity from 10x to a planned 30x, isolating critical systems like Actions, moving hot paths to faster code, and even plotting to run on more than one cloud. The full update is here: GitHub’s post.
But the comments turned it into a roast. The line about a “path to multi cloud” lit a fuse: one user asked if that’s Microsoft quietly admitting Azure can’t keep up, a spicy take in a thread already heavy with side-eyes at Redmond. Another said flat-out they don’t trust Microsoft—no PR-speak will fix it. Meanwhile, a different camp thinks the real shock is the AI deluge, warning that bots from OpenAI, Google, and beyond will keep dumping code into GitHub no matter who “wins.”
Humor flew, too. One commenter riffed “when there’s a gold rush, invest in the jewelry makers,” framing GitHub as the tool-seller in an AI frenzy. And the indie crowd flexed: a user bragged their tiny Forgejo server on a Raspberry Pi feels faster—and more reliable—than the big G. Ouch.
Key Points
- •GitHub acknowledged two recent availability incidents and outlined steps to improve reliability.
- •Capacity expansion began in Oct 2025 targeting 10x, but by Feb 2026 the plan shifted to design for 30x scale.
- •Rapid growth since Dec 2025—driven by agentic development workflows—has increased repos, PRs, API usage, and automation.
- •Short-term fixes included moving webhooks off MySQL, redesigning session cache, reworking auth flows, and adding compute via Azure.
- •Long-term work includes isolating critical services, removing single points of failure, migrating code from Ruby to Go, and pursuing multi-cloud for resilience.