The evolution from classical to quantum information technology will be greatly facilitated by the formulation of incremental approaches to making the transition. This is a far greater challenge than might seem apparent, as today we know only of completely classical and fully quantum-mechanical models as practically sensible computing paradigms and the jump to fully-quantum hardware implementations lies well beyond our current technological capabilities. There is not even a convincing road map for getting there. Photonic (optical) computing architectures, however, provide a rich setting for the exploration of practicable semi-quantum computing paradigms that could be within reach of next-generation fabrication capabilities. They offer the possibility of gradually increasing the role of coherence and entanglement in circuit dynamics, approaching the fully quantum limit in stages. The big question however is how to design architectures whose computational performance and/or energy efficiency can exploit partial quantumness in a natural way. Developing theory around this question will require highly interdisciplinary research at the intersection of physics, engineering and computer science. As a first step to establish foundations for such research our group has been developing modeling tools for photonic circuits that can be applied anywhere along the classical-quantum spectrum, and we have begun to pursue some case studies in coherent information processing architectures. In this talk I will provide an overview of this work, emphasizing the role of coherent feedback as a central design principle.