Hello, i’m trying to recreate a facade that look like the image, it has the pattern of 4 different triangle. If anyone could please help me with a starting point on how can i go on, very appreciated 
It is the same pattern with different degrees of subdivision.
Using Weaverbird certainly simplifies this task. Here, I’ve created the panels and randomly assigned one of three different degrees of subdivision:
Triangular facade with different patterns VR 01.gh (25.6 KB)
however something in the way I’ve trimmed the outer triangles to a bounding rectangle is preventing Weaverbird’s Picture Frame from functioning properly. Also, from the picture it is clear that two types of subdivisioning never neighbour. This is not the case for the script above which simply assigns the subdivisioning randomly.
So for neighbors to never touch you need some sort of graph coloring algorithm (in script some vibe coded component). Or if starting from initial hexagon separate every other triangle to different branch. To get something like this checkerboard pattern.
You get two groups of faces that you can subdivide by different degree. Instead of picture frame maybe offset curves or scale, but its harder to maintain proportions.
Triangular facade with different patterns VR 02.gh (34.6 KB)
So that’s a solution, but from the picture, I see that the windows (for example) are capable of being in the the orange or blue set of triangles. Is it just a different graph colouring then?
Graph Coloring probably is overkill and its main goal is to make sure no adjacent faces touch. Starting from simple module with each face in different branch and creating array is probably easier way.
Anyway I still think its just randomized panels and maybe rotation of panels gives illusion that they are not touching neighbors. Maybe some different noise pattern - simplex, perlin. Cause here is definitely touching.
Nevertheless we can add some recursive splitting or add more different panels.
Not only there, but, upon closer inspection, in other parts of the facade aswell. It seemed to me, however, that a random assignment wasn’t producing as varied a distribution as I was expecting.
Either way, the OP has two options on his hands now.
