Oracle Corporation (SDL)

Contribution to AI

Codd's relational model was nearly a decade old and still largely a paper argument when the first product built on it went on sale. IBM had the prototype — System R, and the query language that became SQL — but moved slowly on a business already earning well from hierarchical systems, and Oracle V2 reached customers in 1979, roughly two years ahead of IBM's SQL/DS and the commercial version of Ingres. What that made ordinary was the separation of a question from the procedure for answering it: a query states which rows are wanted, and the system works out how to fetch them. Version 3, rewritten in C in 1983, carried the same engine onto mainframes, minicomputers and PCs, and within a decade corporate records had migrated out of application-specific file formats into tables that anything could query. Machine learning inherited the arrangement rather than inventing it. The tabular dataset, the join, the feature computed in SQL — statistical work could assume structured, queryable data because a generation of businesses had been sold it. The traffic later ran the other way: the data-mining arm of Thinking Machines was bought in 1999 and its algorithms shipped inside the database from 2002, on the argument that moving the model to the data beats moving the data to the model.

Drafted with AI and edited by hand (claude-opus-5, reviewed 2026-08).