Researchers at Carnegie Mellon University have developed LegoGPT, an open-source artificial intelligence model designed to create custom Lego sets from text prompts. The model is a self-regressive large language model trained on over 47,000 Lego structures, capable of predicting the next brick to ensure stable builds. LegoGPT converts 3D meshes into Lego constructions using various brick sizes including 1×1, 1×2, 1×4, 1×6, 1×8, 2×2, 2×4, and 2×6. It achieves a 98.8% physical stability validation rate, indicating high reliability in the structural integrity of its designs. Beyond Lego sets, the technology has potential applications in Lego-like building of homes and other structures, integrating computer-aided design principles with large language models.