Project ideas¶
This catalog is direction #1 of six sketched during the initial design session. The others live here as a record, in case a later mood calls for one of them.
1. Catalog of the unbuilt (this site)¶
A MkDocs-style catalog of ~25 real historical unbuilt buildings — each with architect, year, site, vital statistics, narrative of what it was, why it failed, and a short what-if. Schematic blueprint-style SVG elevations. Archival typography.
Status: built.
2. Phantom Architecture of the Internet¶
Same format, different subject: unbuilt or lost software systems. Xanadu. Engelbart's Augment. The Symbolics Lisp machines. Alan Kay's Dynabook. Project Juno. Plan 9's successors. Software's own cathedrals that were dreamed and never built, or built and lost. Technical kin to the quant work while preserving the book's spirit.
Effort: similar to this site. A weekend to a week.
3. Generative phantoms¶
LLM-seeded new unbuilt buildings in the voice of historical architects — Boullée writing a 2026 memorial, Tatlin drafting a cloud anchor for Shanghai. Manifesto + spec + SVG elevation, all generated. Fun; risks feeling shallow if the prose is thin.
Effort: a week for the generator, plus curation loop.
4. Three.js 3D walkthroughs¶
Pick four to six iconic unbuilt buildings and render each as a navigable 3D model the reader flies through — Boullée's Newton cenotaph at dusk, Tatlin spinning, Wright's mile rising out of Chicago. Weeks of work per building but genuinely impressive output.
Effort: one to two months for a meaningful set.
5. The impossible skyline¶
Pick one real city (Manhattan, Paris, Moscow) and layer every unbuilt proposal for that city onto an interactive map — Gaudí's hotel, Ferriss's pyramids, Le Corbusier's X-towers. The city as it would have been if every architect had gotten their way. Map + story + architectural history.
Effort: one to two months, primarily for the map UI and data collection.
6. Blend¶
Use direction #2 (internet phantom architecture) as the scaffolding, with direction #3 (generative phantoms) as an appendix. Real unbuilt software up front; generated "what would Engelbart propose now" at the back.
Effort: two to three weeks.
Recommendation¶
Direction #2 is the natural next step from this one — same tooling, same register, different subject matter. It also resonates with the quant project: software has its own cathedrals, dreamed and never built.
Direction #5 (the impossible skyline) is the most ambitious and probably the most compelling output. A natural follow-up after #2.