Abstract: Learning new skills requires synaptic plasticity in cortical circuits. How does the brain choose which synapses to change so learning is fast, efficient, and doesn’t degrade earlier skills or memories? Many models have been proposed, including those that underpin modern AI, but few allow direct measurement of connectivity and activity in the same circuit during learning. We developed an all‑optical method that combines large-scale connectivity mapping with a single-neuron brain computer interface, enabling us to track how both neural activity and synaptic connections evolve during learning. Our data support a 3‑factor Hebbian learning framework, in which performance feedback gates synaptic changes to reinforce useful activity patterns. I will also briefly preview ongoing efforts to identify neuromodulators such as norepinephrine, acetylcholine, and serotonin that may carry this feedback signal to the cortex.