Storygold Research notes
What we learn while training a foundation model for parametric CAD: benchmarks, results, and the occasional release.
04posts
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 down when the model has to learn something genuinely new.
All posts
Sep 28, 2025
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.
Research
9 min →
Aug 22, 2025
How good is GPT-5 at 3D?
GPT-5 performance on parametric CAD generation — syntactically perfect code, dramatically short on geometry.
Benchmark
12 min →
Aug 12, 2025
An open-source foundation model for parametric CAD generation
Introducing BELLA-v1.0-13B — weights, evals, and what it can and can't build yet.
News
7 min →
Apr 9, 2025
Introducing StoryGold
What we're building, why parametric CAD is the right substrate, and why now.
News
5 min →
New notes, roughly monthly.
Research posts and early model access.