The Dream Atlas is an interdisciplinary scientific initiative that employs AI-driven computational methodologies to map large-scale dream patterns across spatial and temporal dimensions, while critically examining the epistemic implications of algorithmic knowledge production. The project advances the systematic study of two intelligences — the oneiric and the machine — and of what each reveals about the other. Conceptualising dreams as formations rather than narrations, it identifies, maps, and measures recurrent dream structures across a target corpus of ~ 250,000+ reports drawn from twelve philosophical and religious traditions and spanning millennia. Bridging anthropology and religious studies with machine learning and computational linguistics, it reveals structures in how humans dream that have remained invisible to traditional small-sample methods — structures that no comprehensive, large-scale analysis has been able to identify until now.