Key Takeaways

  • AI agents are fueling an unprecedented 14x commit growth at GitHub, pushing the platform to a staggering 275 million commits per week and on pace for 14 billion this year.
  • This isn't just a traffic spike; AI workloads expose novel permissioning problems and CPU constraints in foundational systems, forcing rewrites of services that have run for 10-15 years.
  • Traditional vertical and horizontal scaling methods are insufficient. GitHub's Kyle Daigle notes they're in a "diagonal" where even CPU/GPU limits mean breaking open and rewriting decade-old core services.
  • Even a tech giant like GitHub must confront "my SQL 1" – a legacy database that still houses core permissioning layers – demonstrating that no one is immune to AI's disruptive scaling demands.

The AI Tsunami Breaking GitHub's Back

Imagine growth so explosive it cracks the foundations of a trillion-dollar company. That's GitHub today. AI agents, not just human developers, are driving an unprecedented "14x" commit growth. Swyx, the podcast host, laid out the stakes: "1 billion commits in 2025. Now it's 275 million per week on pace for 14 billion this year." This isn't just abstract, it's real-time chaos. Kyle Daigle, COO of GitHub, put it plainly: "It's speeding it's still speeding up."

This isn't merely more load on existing infrastructure. It means GitHub is battling significant uptime challenges, increased CPU demand from GitHub Actions, and performance bottlenecks from growing monorepos and job queuing. The sheer scale of AI agent activity is fundamentally different from human usage patterns, creating a new class of problems that legacy systems were never designed to handle. It's a reminder that even battle-tested tech infrastructure can crumble when a new kind of workload hits like a tidal wave.

Legacy Systems Aren't Ready for AI's Scale

For most founders, scaling means more servers, bigger databases, or sharding. But what happens when the very logic of your core services breaks? Daigle revealed the pain: “Now what we're finding isn't just the like isn't the the the simple stuff that folks are on the you know sometimes on Twitter or on the internet are like hey like why is this like this? Sure there's absolutely you know silly problems that shouldn't exist. But now we're talking about like unique novel permission problems that happen only at a scale across all different objects or whatever that now we have to go rewrite this underlying system.”

The most glaring example? GitHub's permissioning layer. Daigle admitted, “The place that we continue to have pain is in permissioning. And so right now many of our permissioning layers sit into a database that we like internally call my SQL 1.” This isn't a small bug; it's a foundational issue, born from systems that predate the current era by decades. The problem isn't just capacity; it's that the rules encoded in these old systems simply don't make sense, or outright fail, at AI-driven scale. They've been pulling things out of "my SQL 1" for years, but AI's pressure is forcing a full-scale reckoning.

When Vertical and Horizontal Scaling Isn't Enough

When standard scaling fails, what's left? Daigle calls it a "diagonal" problem. “Now like we're sort of in a like diagonal where like vertical doesn't really work anymore. horizontal isn't work either because like we're all we all have some CPU or GPU constraints in the world now and now we have to go and like crack open services that have been running for 10 or 15 years and go okay the rules of this service have like legitimately changed and now we have to rewrite them.”

This is a critical insight for any founder. The problem isn't just throwing more hardware at it. It's that the core assumptions, the very "rules" embedded in services built a decade or more ago, are now fundamentally broken by AI workloads. This means not just optimizing, but rewriting services from scratch. If GitHub – with its massive engineering resources – is facing this, it's a wake-up call for every startup relying on infrastructure that hasn't been re-evaluated for a post-AI world.

What to Do With This

Don't wait for 14x growth to expose your architectural vulnerabilities. Take a hard look at your oldest, most critical services – especially permissioning, job queues, or core data stores. Identify the systems that haven't seen a significant rewrite in 5+ years. Assume AI agents, not just human users, will hit them. Can they handle novel interactions and unprecedented scale, or will you be cracking open "my SQL 1" under pressure like GitHub?