Although mobile phones are ideal platforms for continuous human centric sensing, the state of the art phone architectures today have not been designed to support continuous sensing applications. Currently, sampling and processing sensor data on the phone requires the main processor and associated components to be continuously on, creating a large energy overhead that can severely impact the battery lifetime of the phone. In this paper, we present the design and implementation of Little Rock, a novel sensing architecture for mobile phones, where sampling and, when possible, processing of sensor data is offloaded to a dedicated low-power processor. This approach enables the phone to perform continuous sensing at a low power overhead. We highlight and discuss in detail various design choices, tradeoffs and lessons learned. Using a pedometer application as an example, and by integrating Little Rock into an actual phone, we show that the proposed sensing architecture can be three orders of magnitude more energy efficient compared to the normal approaches.