This paper examines the role of numerical libraries within computational infrastructure through a case study of the Basic Linear Algebra Subprograms (BLAS). First released in 1979, BLAS remains foundational for performing fast, reliable vector and matrix operations. Drawing on infrastructure studies and close readings of BLAS documentation, I show how the library’s design produces a layered artifact. A stable, abstract interface enables consistent use, while a dynamic, evolving interior allows developers to optimize algorithms over time. This synergy helps explain BLAS’s longevity. Attending to these layers illuminates patterned (in)visibility within computational systems and invites further study of their socio‑technical dynamics.
