We have been rethinking the fundamental sampling strategies used in interactive computer graphics. This rethinking has led to such interesting questions as, “How should we display an image in which some pixels are older than others? Do we need pixels at all? Do we even need ‘images’ in the classic sense of ‘frames in an animation’?” We will describe a novel temporally adaptive temporal sampling approach based on frameless rendering, a per-sample rendering algorithm that decouples spatial and temporal updates and thus enables very flexible adaptive spatial and temporal sampling. Closed-loop feedback guides sampling to image regions that change significantly over space or time. Adaptive reconstruction generates images for display, emphasizing older samples in static settings (resulting in sharper images) and new samples in dynamic settings (resulting in possibly blurry but up-to-date images). Measurements (using peak signal-to-noise ratio) indicate that this system produces better image streams than framed or non-adaptive frameless renderers with the same simulated sampling rates.