Starting from a development pilot project that aimed to introduce new water accounting procedures to Cambodia, this article examines interactions between technical experts from abroad and government officials. Drawing on STS, performativity theory, and the anthropology of development, the article shows that the dynamics at the project interface are characterized by parallel and incongruent performances. Visiting technical experts work on the assumption that they are operating at a science-policy interface. Meanwhile, officials align with the demands of technical rationality, aware of its discrepancy with the performance of politics outside the project frame. Two versions of project realities and their relation to broader Cambodian realities are thus performed simultaneously, but awareness of this is not evenly distributed, with significant consequences for the aspiration of development organizations to transfer knowledge. The case study feeds into a subsequent characterization of a more general pattern of sparks and fizzles, in which projects continuously start up, while efforts to bootstrap technology transfer peter out as they end. This pattern may be endemic to development projects that operate on the assumption of a science-policy interface that is not really there.