In my talk, I will discuss how the concept of Fisher information provides a unifying framework for understanding the fundamental limits of optical measurements. When we infer properties of an object from scattered light, Fisher information quantifies how much can, in principle, be learned from the measurement, independently of any specific reconstruction algorithm. I will show how this viewpoint leads to practical strategies for improving optical experiments, including the use of wavefront shaping to actively maximise the information carried by scattered light [1]. A particularly striking recent result is that Fisher information can be treated as a physical quantity that flows through space: its density and flux obey a continuity equation that closely mirrors the Poynting theorem for electromagnetic energy [2, 3]. This analogy provides new physical intuition for how information is redistributed in optical systems. In the final part of the talk, I will briefly outline how these ideas extend beyond wave physics and how analogous notions of Fisher-information flow can be defined in artificial neural networks, offering a bridge between optical measurement theory and modern machine-learning methods [4, 5]. [1] D. Bouchet, S. Rotter, and A. P. Mosk, “Maximum information states for coherent scattering measurements,” Nature Physics 17, 564 (2021). [2] J. Hüpfl et al., “Continuity equation for the flow of Fisher information in wave scattering,” Nature Physics 20, 1294 (2024). [3] M. Weimar et al., “Controlling the flow of information in optical metrology,” arXiv:2508.13640 [4] I. Starshynov et al., “Model-free estimation of the Cramér–Rao bound for deep learning microscopy in complex media,” Nature Photonics 19, 593 (2025). [5] M. Weimar et al., “Fisher information flow in artificial neural networks,” Physical Review X 15, 031072 (2025).