Kubernetes 1.35 introduces ten new Alpha features that enhance its capabilities, particularly for AI workloads. Among these, two notable advancements in Dynamic Resource Allocation (DRA) are Device Binding Conditions and Partitionable Devices. Device Binding Conditions allows Pods to wait for external resources, like GPUs, to be ready before binding, improving reliability. Partitionable Devices restores the ability to dynamically partition devices, aiding resource utilization while keeping the process transparent for users. Another significant feature is the Mixed Version Proxy, which addresses version skew during upgrades by ensuring client requests are directed to the appropriate API server, preventing 404 errors. This proxy also aggregates discovery documents from multiple servers, helping controllers function correctly. The introduction of a mechanism for CSI drivers to opt-in for service account tokens in a secure way reduces security vulnerabilities associated with logging sensitive data. Kubernetes 1.35 also includes a framework for Node Declared Features, allowing nodes to declare available features, which helps prevent scheduling errors due to version mismatches. Further enhancements include the integration of CSI Volume attach limits with the Cluster Autoscaler, improving how clusters scale based on storage limits. Mutable PersistentVolume Node Affinity allows for changes in volume accessibility, reducing scheduling failures for stateful pods. The extension of Toleration Operators supports numeric comparisons, which enables better scheduling based on service level agreements. Lastly, the support for in-place pod vertical scaling enhances resource management, and gang scheduling is set to improve how workloads are handled in the scheduler. These features collectively represent a significant evolution in Kubernetes' functionality, targeting efficiency and reliability in resource management and workload orchestration.