Storygold Training foundation models for parametric CAD
We're solving the data desert in AI for 3D reasoning by collecting and generating the world's largest dataset of parametric CAD models and learning environments.
We're building a 3D reasoning model
10+ Years of AI Research Experience
Our team consists of MIT, Princeton, and International Math Olympiad alumni who spent years training large-scale deep learning models and LLMs at Google Research, OpenAI, and Waymo.
Training on terabytes of data
We're building the world's largest parametric CAD dataset. Focusing on parametric CAD lets our model preserve engineering intent — dimensions, constraints, relationships — that downstream editing, simulation, and manufacturing depend on.
Building general intelligence
True general intelligence must understand the physical world. We're building general intelligence by teaching it 3D reasoning, starting with the backbone of all manufacturing: CAD.
Research notes
All posts →LoRA fine-tuning for vision-language models
LoRA is touted as being as effective as full fine-tuning. On spatial reasoning over CAD data it isn't — the low-rank assumption breaks when a model has to learn something genuinely new.
How good is GPT-5 at 3D?
Syntactically perfect code, dramatically short on geometry. An empirical analysis of frontier-model performance on parametric CAD generation.
Building AGI requires intelligence that understands the 3D world.
Get early access to the model and occasional research updates.