The Platform Engineer's Handbook

Ajay Chankramath — author of The Platform Engineer’s Handbook — joins Kaspar von Grünberg to unpack why he wrote a 14-chapter, code-first practitioner's guide instead of another theory-heavy platform book. The conversation's core thesis: the reason developers don't adopt platforms isn't a technology gap, it's a product discipline gap — a failure to treat developer experience as a first-class outcome with a real feedback loop. Ajay walks through the book's arc, from Kubernetes and service-mesh foundations through self-service portals to enterprise-grade concerns like policy-as-code and FinOps, and makes a pointed case for building on 100% open-source, vendor-agnostic tooling as a hedge against geopolitical and licensing whiplash.

The most relevant thread for engineers building with AI today: citing a McKinsey finding that only 6% of AI initiatives show real productivity gains, Ajay argues that agentic AI doesn't reduce the need for platform engineering — it raises the stakes. As coding agents become genuine actors in the SDLC rather than tools, IDPs need a new layer for agent context, memory, tool registries, and guardrails, and that layer "must be built, owned, and operated by you," not bought off the shelf.

His conclusion: the differentiator in the AI era isn't which frontier model you use — it's whether your platform foundations are solid enough to make agents safe and productive at all.