I had the chance to visit the theoretical group of Karl Friston at the Wellcome Trust Centre for Neuroimaging (University College London) from November 2010 to February 2012. During this period we built a series of active‑inference models of eye‑movement control. First, we introduced a generative model of saccadic behaviour Friston et al., 2012. We then extended the framework to clinical populations, demonstrating that neuromodulation may critically shape model predictions in participants with and without schizophrenia (Adams et al., 2012). Finally, we incorporated realistic sensory‑motor transmission delays, providing a more accurate account of the timing of perception‑action cycles Perrinet et al., 2014.