In this paper, we use an agent-based simulation model to investigate how coordinated exploration by multiple specialists, as in new product development, is different from individual search. We find that coordinated exploration is subject to two pathologies not present in unitary search: mutual confusion and joint myopia. In joint search, feedback to one agent’s actions is confounded by the actions of the other agent. Search therefore leads to increasing mutual confusion because agents are unable to learn from feedback to correct their faulty mental models of the search space. Incorrect beliefs held by one agent lead to mistakes, and because it is unclear which agent was wrong, this confuses the other agent, either into revising (correct) beliefs or holding on to (incorrect) beliefs. Sharing knowledge aligns specialists’ mental models and counters mutual confusion by inducing coordination around particular search regions. Yet that very effort increases joint myopia, as agents prematurely reinforce each other into choosing from an increasingly narrow portion of the search space. In the extreme, high levels of shared knowledge induce agents to abandon their distinct search approach in favor of a lower common denominator. In coordinated exploration, increasing coordination efforts (such as by increasing communication) reduces mutual confusion but simultaneously increases joint myopia. Efforts to reduce joint myopia, such as by slow learning or lower levels of knowledge transfer, however, automatically increase mutual confusion. As modeled in our simulation, successful joint search needs to balance these two effects. Our results suggest that because unitary-searcher models abstract from epistemic interdependence, their predictions are potentially misleading for coordinated exploration.