Bowker’s analysis of scientific memory practices argues that science does not simply preserve the past; it actively constructs usable pasts through infrastructures, classifications, archives, standards, and databases. Against the assumption that memory is merely conscious recollection, Bowker defines it as a dispersed set of social, technical, and institutional practices ranging from habitual procedure to hyper-detailed archival accumulation. Scientific knowledge depends upon this managed memory because facts must appear stable, transferable, and timeless, even though they are produced through historically situated labour. Time, therefore, becomes the “money of science”: a shared standard that allows experiments, observations, disciplines, and databases to be synchronised across different locations and temporal scales. The essay’s central tension lies between the mnemonic deep, where the past remains tangled, partial, and discontinuous, and the scientific aspiration to an eternal present in which laws of nature appear free from contingency. Bowker’s case of geology is especially revealing: Lyell’s earth functions as an imperfect archive, preserving traces of extinction, succession, and transformation, yet always through gaps, distortions, and uneven records. This geological example anticipates contemporary database science, where biodiversity records similarly claim comprehensiveness while excluding local, historical, or material traces that do not fit standardised categories. Bowker thus shows that archives are never neutral containers; they organise what can be remembered, compared, forgotten, or rendered scientifically invisible. His conclusion is that robust scientific databases must preserve traces of their own making, because the authority of science depends not on perfect memory but on recognising the infrastructural labour through which memory is produced.