GitOps in 2017 as a simple model for application delivery, built around four core principles. Today, it has evolved into a central part of platform engineering, where teams operate large numbers of clusters and shared platform services.

This talk shows how GitOps has evolved over time, why early approaches break at scale, and how everything-as-code enabled automation while also creating configuration sprawl. Tools like Argo CD, Helm, and Kustomize made scaling possible, but often hide the real state of a system.

We will look at the limits of Git as a state store and why file-based, pull-driven reconciliation does not scale well. This led to OCI-based delivery, Gitless GitOps, and new approaches that aim to provide a real, operable state store.

Finally, we will explain why this matters now. Platform teams need visibility, regulation like the Cyber Resilience Act requires a clear infrastructure supply chain, and AI-driven operations depend on access to the final, real manifests.