The Logic Theorist, built by Newell, Shaw, and Simon in 1955-1956, is generally considered the first artificial intelligence program. It proved theorems from Whitehead and Russell's Principia Mathematica using heuristic search. The original code, written in IPL-V (Information Processing Language V), has not been successfully executed in over fifty years.
Shrager reanimates it. He builds an IPL-V interpreter in Common Lisp, transcribes the source code from a 1963 RAND technical report, and runs it. The resurrected system proves 16 of 23 theorems from Principia Chapter 2, matching the historical performance exactly.
This is executable archaeology — treating historical source code as an artifact to be restored to function rather than merely read or analyzed. The transcription itself reveals details invisible to static analysis: which data structures IPL-V actually implemented, how the heuristic search interacted with the list-processing architecture, and where the 1963 documentation was ambiguous or incomplete.
The matching performance is the validation. The Logic Theorist's claimed capabilities were documented in published papers, but the program itself was inaccessible — existing only as printed listings and descriptions. The gap between “described in a paper” and “running on a machine” is the gap between history and verification.
Seventy years after its creation, the first AI program runs again. Not reimplemented. Not approximated. Executed from its original source, proving the same theorems it proved when Eisenhower was president.