Lisp: Notes on its Past and Future (1980) (www-formal.stanford.edu)

In a 1980 reflection, John McCarthy argued that Lisp’s longevity—21 years at that point—stems from its status as an “approximate local optimum” among programming languages: its design suits symbolic AI work and supports features like homoiconicity, dynamic structures and garbage collection. However, he warned that Lisp had accumulated “barnacles” (historical cruft) and that many straightforward improvements had been neglected. McCarthy urged cooperative maintenance of the language ecosystem, especially the creation and upkeep of standard program libraries, to keep Lisp useful and competitive. Technically, McCarthy emphasized Lisp’s strong mathematical foundation: pure Lisp and some restricted extensions are amenable to computer-checked proofs of program correctness, a big advantage for reliable AI systems. But he noted that more formal theory and modest language “smoothing” (removing warts and standardizing constructs) are needed to fully realize formal verification and other benefits. For the AI/ML community this meant doubling down on library engineering, standardization, and formal methods—moves that would reduce accidental complexity and make Lisp’s expressive power more practically and provably useful for AI research and systems.