A library organised as a building
A search box assumes you know the name of what you want. A library assumes you do not, and arranges itself so that walking through it teaches you the shape of the field.
The problem
Primary texts are abundant online and almost impossible to move through. They sit in databases built for retrieval — you arrive with a citation, you leave with a passage. Nothing in that interaction tells you what sits next to the thing you found, what tradition it belongs to, or what a reader in another civilisation was writing in the same century.
The consequence is that the record is only useful to people who already know it. A structure that requires expertise to enter cannot create any.
What I built
A building. Five wings, floors, rooms and corridors, where the spatial arrangement carries the scholarly relationship: Plato beside Confucius, the Upaniṣads beside Rūmī. You navigate by moving rather than by querying, and the position of a text is itself an argument about where it belongs.
Everything is sourced, cited and readable in the original script. The apparatus is visible — a reader can see which edition a passage came from without leaving the room.
The decision that shaped it
Search was deliberately demoted. It is present, and it is not the interface — because a search-first library silently teaches its readers that the only findable text is one they could already name. The building is the argument: the arrangement is doing scholarly work that a ranked result list cannot do.
Comparison is built in rather than bolted on. Up to four manuscripts can be read side by side, which is the actual unit of cross-civilisational reading and is nearly impossible in every other digital edition.
What it demonstrates
That information architecture is an argument, not a wrapper. At 108,021 sections the structure is the product: the same corpus behind a search box would be a database nobody learns anything from. Related: waypoints.now, where the same reasoning turns routes into reach.