A Manifesto for Data Realism (muskdeer.blogspot.com)

"A Manifesto for Data Realism" rejects contemporary "data idealism"—the view that transparency, fairness and ethical framing alone can manage digital harms—and argues for a pragmatic, infrastructure-first perspective. The manifesto distills Data Realism into five tenets: data are our primary contact with reality and should be broadly accessible via commons; data are produced and shaped by situated infrastructures and power; the world "leaks through" data (rejecting both naive empiricism and total constructivism); data enable state and industrial agency, so sovereignty over standards, storage and compute matters; and pragmatic, accountable lifecycle management of data (depreciation, auditing, updating) must guide policy. For AI practitioners and policymakers this reframes priorities from moralizing rules to building concrete capability: legal clarity on collecting publicly available data (with technical standards like rate limits, robots.txt adherence, anonymization and exclusion of sensitive fields), investment in curated public datasets and standardized APIs from natural monopolies, and sovereign technical standards for storage/compute to reduce supply-chain dependency. The manifesto stresses rigorous auditability, data lifecycle accounting, and strategic national coordination—arguing that competition should be won by modeling and algorithmic innovation, not proprietary data hoarding, while also treating data governance as a core national-security and industrial policy problem.