Rumah Adat Nusantara

The problem

Most Indonesians meet rumah adat, the traditional houses, as a list to memorise for a school exam: a name, a province, a picture of a roof. Nothing in that explains why the roof has that shape. The knowledge exists in ethnographies, provincial survey volumes and architectural monographs. But it is scattered, mostly untranslated, and written for specialists.

The usual digital answer, a photorealistic 3D model, makes this worse in a quiet way. A smooth render states every dimension with full confidence, including the ones nobody has ever measured.

I built Rumah Adat Nusantara for a curious Indonesian reader on a phone, in Indonesian first. It also serves architecture and anthropology students and teachers who want the logic and the citations, and will check them.

What I built

Each building is computed from its cultural rules

Every building is generated from the rules that govern it: a household's rank, the number of families under one roof, a room count, the length of the owner's forearm. Change one of those numbers and the structure rebuilds in front of you. There is no roof-curve slider, because a roof shape is a result. The controls only offer things a household would say about itself. I wrote every post, rafter and joint as plain arithmetic, with no CAD imports or modelling libraries.

I grew the shared core only from repeated patterns

A shared core holds what is true of any building: parts, joints, build order, symmetry and sources. Everything cultural stays with the building it belongs to. Nothing moves into the core until two buildings have needed it independently, because an abstraction built from one example takes that example's shape. I chose the Minangkabau rumah gadang as the second building because it would break things: its ridge runs the other way and it is asymmetric. Tests fail the build if the core names a tradition.

Every dimension says where it came from

Each dimension is labelled measured, canon or interpolated, with a source. A source page lists every dimension with its citation and how much the house changes if that number moves. I never average figures across two buildings, because that would let an unsurveyed building hide behind a surveyed one. Each building also has a caution that states its simplifications. Today no dimension in any building is measured, and a bar on every screen shows that split.

Every picture comes from the model

Landing silhouettes, catalogue plates, exported drawings and all 70 share cards are computed from the same parts the tests check. CI compares them pixel by pixel against the generator, so an illustration can't disagree with the model. The sun is computed too. Each building has its actual site, and the scene shows the actual sun over that site on the chosen date.

Buildings sit side by side and never morph

A comparison page places any two buildings on one ground line at one scale. I don't morph one tradition into another, because that would assert a continuity that doesn't exist. Motion is limited to five animations that each show a mechanism, such as the frame going up joint by joint or rain running off the roof. Nothing spins for decoration.

Result

Rumah Adat Nusantara is live and public in Indonesian and English. It has 35 buildings: 26 houses, two granaries, three halls, a palace, a houseboat, a stone tomb, a cremation tower, a nine-building compound and a pair of ancestor shrines. Their sites span three time zones. Each building has four pages, 280 in total, and its own structural checks run over at least five rule combinations. A failing check stops the release. The site has no backend and makes no network calls, so it works with wifi off. I built it in eight days, from 25 August to 1 September 2026.