Something we’ve been thinking about while building @Reer_Inc for Rhino.
AI can see geometry.
But geometry isn’t meaning.
A box could be a wall, a stair core, a planter, or a column.
Without context, AI is guessing.
With context, AI can reason.
One thing we’ve learned is that AI struggles with many of the same things humans do when opening an unfamiliar Rhino file:
-
Layers called “Layer 01”
-
Objects named “Polysurface” and “Extrusion”
-
No idea which direction is north
-
No clue how the project is organized
One interesting example came up during one of our user sessions.
The user’s first request was:
Select all chairs in this interior design project.
Sounds simple. But without any project context, the agent had to guess what a “chair” was from thousands of geometries in the file. It eventually got there, but it spent far more time reasoning than actually doing the task.
That led us to a question:
What if AI had to learn the project before doing work on the project?
So we started experimenting with a process called /sync.
Before modeling, the agent learns project orientation, units, naming conventions, layer standards, and references from previous projects. It can reorganize layers, rename objects, and build an understanding of the project before touching the model.
The interesting part isn’t the cleanup.
It’s what becomes possible afterward:
Select all south-facing railings.
Find the glazing panels on the west facade.
Rename all Level 3 balcony elements.
Does this make Rhino closer to BIM? 