Whole cell recording from subcortical structures has revealed much about the membrane properties of sensory systems. Since its introduction by Margrie et al in 2002, however, patch clamp recordings deep in the intact brain have suffered from low yield and high access resistance, primarily due to pipette contamination during descent to subcortical nuclei.We describe a novel algorithm and best practice for reducing pipette tip contamination and improving access resistance when targeting deep, subcortical neurons in mice (e.g., 1-3 mm deep). In this work, we introduce an automated method for dodging obstructions such as neurons and blood vessels during descent.Traditionally, pipettes are localized in the intact brain by blind, linear rapid descent to the subcortical region of interest (e.g., hippocampus, thalamus). During this descent, pipettes are often contaminated by debris that prevents gigasealing and increases access resistance. Using this new algorithm, pipette resistance is measured at 128 Hz during descent for a high spatial resolution. If an obstruction is encountered prior to the region of interest (as indicated by an increase in resistance >1 Mohm threshold), the pipette is stopped, retracted, and moved laterally in 20 um steps. Lateral steps are repeated until resistance decreases below threshold, indicating that the obstruction has been avoided and descent is resumed.Pipettes inserted using this method arrived at the region of interest without a significant change in resistance (<300 kohms at 3000 um) 69% of the time (n=46), while pipettes inserted using traditional blind, linear methods were inserted successfully 32% of the time (n=31); Whole cell recordings following successful descent resulted in lower access resistance recordings and smaller holding currents than those inserted with the traditional algorithm. This pipette localization method therefore greatly improves the quality and accessibility of subcortical tissues in the intact brain.