Autoscaling is now in public preview, allowing users to dynamically adjust infrastructure based on demand by setting policies around requests per second, CPU usage, and memory consumption. This feature offers flexibility and control, enabling users to tailor autoscaling triggers to meet specific needs while managing costs through a simple slider for the maximum number of instances. It supports global deployments by allowing region-specific scaling policies, ensuring optimal performance and cost-efficiency. Autoscaling decisions are made every 15 seconds based on service metrics, with a one-minute cooldown window to stabilize changes. Users can initiate autoscaling via a command line interface (CLI) or control panel, and they can define policies based on multiple metrics. Improvements in the works include autoscaling visualization, scaling down to zero for cost savings, and policy assistance recommendations. The feature is intended to enhance the serverless experience by eliminating the need for manual infrastructure adjustments, and users can explore further details in the autoscaling documentation or join the community for updates.