For decades the actual source code of ELIZA - Joseph Weizenbaum’s 1966 chatbot, the first program to produce the illusion of conversation - was thought lost. Most existing “ELIZA” programs were later reimplementations based on Weizenbaum’s published 1966 paper, not his original. ElizaGen, an archival project curated by Jeff Shrager, documents how that changed.
In 2021 Shrager rediscovered Weizenbaum’s original code in the Weizenbaum archives at MIT. It was written in MAD-SLIP, a now-obscure combination of the MAD programming language and Weizenbaum’s own SLIP list-processing library, running on MIT’s time-sharing system. The find was discussed in a July 2022 CoRecursive podcast interview and documented by a team including Anthony Hay, Mark Marino, Peter Millican, and David Berry.
The recovered code revealed details the published paper had glossed over, including how the program let users edit its conversational scripts. Working from it, contributors produced faithful restorations - Anthony Hay built a near-exact clone of the original MAD-SLIP ELIZA in C++ - so the program could be run again as Weizenbaum’s contemporaries had experienced it.
Why business readers should care: ELIZA’s rediscovery is a case study in software preservation - foundational systems can vanish into archives within a generation, and recovering the actual artifact, not a secondhand description, is what lets us understand how the technology truly worked.