IBM's John Backus and his small team circulated preliminary specifications that would evolve into FORTRAN, the language that let scientists and engineers write formulas nearly as they appeared on paper instead of raw machine code. Delivered commercially in 1957, FORTRAN cut programming time dramatically and became the dominant language for scientific and engineering computation for decades, influencing nearly every high-level language that followed.
John Backus pitched the idea to IBM management with an argument that was almost heretical at the time: programmers were wasting enormous amounts of expensive computer time and human effort writing and debugging assembly code by hand, and a compiler that translated mathematical notation directly into efficient machine instructions could fix it. Skeptics doubted a compiled language could ever generate code efficient enough to satisfy scientists obsessed with performance on room-sized, wildly expensive machines like the IBM 704.
Backus's team—a small, unusually young group that included Lois Haibt, one of the only women in the group and one of the few people at the time with formal training relevant to compiler design—spent years solving problems nobody had solved before, including how to optimize loops and allocate registers automatically. The resulting compiler generated code good enough that skeptical engineers accepted it, and IBM's gamble paid off spectacularly once FORTRAN shipped commercially in 1957.
Decades later, physicists running climate models, engineers designing aircraft, and researchers modeling nuclear reactions still relied on FORTRAN code, some of it barely modified since the 1960s. Backus, characteristically modest about the achievement, later spent much of his career trying to move computing away from the very paradigm he'd helped popularize, arguing that programming should be less about detailed step-by-step instructions and more about expressing mathematical relationships directly.