Verne Harris was bringing it a moment ago talking about how archival scholars and practitioners must infiltrate big data disciplinary spaces and understand these infrastructures as "archontic apparatuses" and dismantle them.

Which reminds me of critical data/algorithm studies and some recent work looking at the history/provenance of ML training datasets:

https://arxiv.org/abs/1912.10389

https://arxiv.org/abs/2012.05345