In a recent conversation with Abdel Sghiouar, platform engineering ambassador, my fellow AI obsesser, and Kubernetes/GKE dev rel, we were discussing how hard it can be to have conversations about the frontier of AI in platform engineering. If you want to talk code gen? Easy. If you want to talk about best practices for dynamic infrastructure for sub-agents? Well, that’s a bit harder.

Lucky for me, I’ve got Abdel and Google Cloud to help spread the knowledge. In a banger two-parter community article and webinar this week, Abdel breaks down how Kubernetes and platform engineering principles can power intelligent, goal-driven applications that reason, act, and learn autonomously.You know that famous 95% number from MIT about the masses who aren’t seeing value from AI? This is what the 5% are doing. 

I’m summarising over 3,000 words of knowledge here, but if you want to get the most detailed breakdown that currently exists. Here’s a taste:

• Agentic apps demand adaptive infrastructure: AI agents operate in an “agentic loop” of perceiving, reasoning, acting, and learning. Their dynamic, unpredictable workloads require infrastructure that can scale and adapt in real time.
• Kubernetes can provide the orchestration layer for that autonomy: With Dynamic Resource Allocation, GPU sharing, and live pod resizing, Kubernetes can let agents scale efficiently while maintaining security, governance, observability etc
• Platform engineering bridges the gap: IDPs remove Kubernetes complexity, giving AI engineers golden paths, built-in observability, and secure, governed environments for deploying agents.

If the State of AI in Platform Engineering, and DORA convinced you that you platform engineering is the key foundation of successful AI. Abdel and GKE are showing you examples of exactly what that successful AI can look like.

If you want to take a look, he and the team have put together a huge chunk of practical examples, tutorials and quickstarts for deploying AI on GKE.

K8s AI make-over is just getting started.