March 30, 2026
Retro code, fresh chaos
Douglas Lenat's Automated Mathematician Source Code
1977 “robot mathematician” code resurfaces and the internet asks “uh… what is this?” while finding a D&D sheet
TLDR: Douglas Lenat’s 1977 Automated Mathematician code hit GitHub, revealing a time-capsule of early AI and even a surprise D&D character sheet. Commenters split between awe and bewilderment, debating whether it’s a foundational breakthrough or just a fascinating museum piece—and asking how anyone runs it today.
The source code for Douglas Lenat’s AM—short for Automated Mathematician, an early attempt at a computer that discovers math ideas—just landed on GitHub, and the comments immediately turned into a mix of museum tour, meme night, and group therapy. The repo pulls files from Stanford’s old SAIL archive and even name-drops ancient hardware (a PDP-10!) and Interlisp, which is basically software Jurassic Park. The crowd reaction? Half nostalgia, half confusion, all drama.
The thread’s loudest energy is “Explain Like I’m Five.” One user literally asks, “Can someone in the know give a little summary,” while others debate whether AM actually “discovered” math or just followed clever breadcrumbs. Cue the hot takes: some call AM the grandparent of today’s AI; others say it’s more historical artifact than usable tool. And then chaos: someone finds a literal Dungeons & Dragons character sheet in the archive, instantly crowning Lenat a founding member of the Nerd Multiverse. The jokes write themselves.
Fans celebrate the drop as a rare time capsule—and it’s reportedly public domain thanks to ARPA funding—while skeptics wonder if anyone can even run this code today without a time machine. Links fly to emulators, PDFs, and the repo. Bottom line: it’s a relic with myth-level vibes, and the community can’t decide if it’s a treasure map or a museum piece.
Key Points
- •Repository hosts source code for Douglas Lenat’s AM, extracted from the SAILDART archive.
- •Original environment: SUMEX, PDP-10 KI-10 uniprocessor, 256k core memory, running Interlisp (Jan 1975 release).
- •Interlisp used 140k of 256k memory and provided an additional 256k “shadow space” for compiled code.
- •Loader and config notes: use LT as initial loader; TOP6 was later split into TA and TB; CON6 may require older versions from the archive.
- •The software is considered public domain due to ARPA funding; the thesis encourages copying, modification, and use.