This paper proposes a new epistemological model of scientific discovery. We argue that scientific knowledge does not accumulate linearly but transitions from a state of epistemic superposition—multiple competing hypotheses existing simultaneously in the mind and in the representational landscape of artificial intelligence systems—to a state of determinate understanding when inquiry tests these hypotheses against reality. This “collapse” is not metaphysical and has no connection to quantum indeterminacy in nature. Rather, it is the reduction of mental uncertainty through contact with an independently existing world. We introduce the Collaborative Thought Engine (CTE) as a structured methodology that couples human intentionality with AI-assisted exploration to accelerate this epistemic-collapse process. Historical case studies illustrate the model, and we outline its implications for future discovery frameworks.