A static website reading ELIZA, Joseph Weizenbaum's 1966 chatbot (MIT, Project MAC, CTSS on the IBM 7094), as a cultural and technical text. The public face of the ELIZA Archaeology Project and the book Inventing ELIZA: How the First Chatbot Shaped the Future of AI (MIT Press, Software Studies). A companion to the Critical Code Studies group's readings of SHRDLU (1968–70) and Spacewar! (1962).

Status: v0.5.0. Companion to the book Inventing ELIZA, out 14 July 2026.

The site is built around the original MAD-SLIP source recovered from Weizenbaum's papers in the MIT archive in 2021. It reads the code, the DOCTOR script, the hardware, the many versions, and the program's long afterlife. It also runs a faithful ELIZA in the browser.

The IBM 7094 / CTSS console: a steel grey-blue page and header, warm line-printer text, and a bank of indicator lamps in the 7094 palette (the lamp red is the coral of the Inventing ELIZA book cover). The person types lower case; ELIZA answers in CAPITALS.

• Hero console (console.js) — the hero panel is a working IBM 7094 console: a beige chassis with a steel face, nameplate, live blinking indicator-register lamps, rocker toggle switches, and a recessed CRT readout. The recovered ELIZA-1965b.mad source lists faintly on a fanfold sheet behind it.
• Power down (console.js) — the lit red POWER button runs a 7094 shutdown: the lamps flurry and cut out as a Web Audio synth sings Daisy Bell, slowing and deepening like HAL 9000. A WHAT IS THIS? control opens a modal on the 1961 Bell Labs 7094 recording the scene alludes to.
• Easter egg (term.js) — type the letters e‑l‑i‑z‑a anywhere and the DOCTOR wakes in an overlay, running the genuine 1966 script. Esc closes it.

Static HTML/CSS/JS, no framework, no build step. Served by GitHub Pages.

assets/eliza.js is a faithful JavaScript re-implementation of Weizenbaum's algorithm, driven by the verbatim DOCTOR script from the 1966 CACM appendix (transcribed by Anthony Hay). It implements keyword ranking, decomposition / reassembly, equivalence (=KEY), NEWKEY, PRE, tag dlists, and the "certain counting mechanism" for memory that the recovered source finally explained. It reproduces the canonical 1966 conversation. For the historically exact recreation, see Anthony Hay's C++ ELIZA.

• No em dashes (verify grep -rc "—" *.html = 0). En dashes allowed; British spelling.
• Cache-buster: bump ?v=N in the generator when a versioned asset changes.
• Verify, never fabricate. Citations to be confirmed against Zotero before final.

• sources/ELIZA-1965b.mad — the recovered MAD-SLIP source.
• sources/DOCTOR.txt — the 1966 CACM DOCTOR script (Anthony Hay transcription).
• docs/annotations.md — close-reading annotations on the source.

The blog has no build step. To add a post, create one Markdown file in blog/posts/ and add one line to blog/posts.json. You can do it entirely in the GitHub web interface. Full instructions: blog/CONTRIBUTING.md.

Posts are Markdown with a small front-matter header (title / date / author) and are rendered in the browser by assets/blog.js and the vendored assets/marked.min.js. The Blog index (blog.html) and each post (blog/post.html?p=<slug>) read from blog/posts.json. Because posts are fetched at run time, preview locally with a server (python3 -m http.server), not file://; on GitHub Pages it just works.

The non-blog pages share chrome generated by a small Node script (not a runtime build): docs/gen.js. Edit it and run node docs/gen.js to rebuild Home, Overview, Code, DOCTOR, Versions, Try, People, Talks, Book, Bibliography, About, Links and the Blog shells. Editing the emitted HTML directly is also fine. docs/gen-md.js only seeded the initial 11 Markdown posts; do not re-run it once co-authors edit posts.