Developing Blender native skills for Deepseek V4.1 Flash - Feasible?

I don't understand the apparent skill ceiling for Deepseek when it comes to modelling in Blender. Surely it's only a matter of generating orthographic views of said model and segmenting it into smaller parts, allowing the AI to analyse the measurements, coordinates and provenance of each part of the preexisting model and then using that data to build a database and then apply that database as a skill set? Am I missing something here?