Retinotopic coding organizes the interaction between internally and externally oriented brain networks.

The human brain seamlessly integrates internally generated thoughts with incoming sensory information, yet the large-scale networks that support these functions -- the (Default Network, DN) and external (Dorsal Attention Network, dATN) -- are traditionally viewed as functionally antagonistic. This raises a crucial question: how does the brain integrate information across these seemingly opposed systems? Here, using densely sampled 7 T fMRI, individualized resting-state parcellations, and voxel-wise population-receptive-field mapping, we show that these internal/external networks are more interlocked than previously thought. Although spontaneous DN and dATN activity during rest is uncorrelated at the network level, functional coupling across networks is shaped by the latent visual field preferences of individual voxels in each network, as measured during independent retinotopic mapping. Voxels that share visual field preferences exhibit stronger spontaneous coupling than those with divergent preferences. These retinotopically-specific interactions are bivalent: DN voxels with negative (suppressive) visual response amplitudes are anticorrelated with matched (positive) dATN voxels, while those with positive response amplitudes are positively correlated. Thus, distinct subpopulations of visually-tuned DN voxels participate in spatially-specific interactions with the dATN. Further, retinotopic coding is intrinsic to the DN, persisting even during periods of elevated top-down drive from the DN to the dATN. These findings challenge the prevailing view of global DN-dATN antagonism, revealing a latent, voxel-level architecture of retinotopically-grounded interactions. Taken together, our results suggest that retinotopic coding underpins the dynamic coordination of perception and thought in the human brain.