Writing a SIGGRAPH Paper from Scratch
Story 6, from Lobsters
A self-taught software engineer created a terrain generator using diffusion models, published a SIGGRAPH paper, and gained significant attention online. The process involved years of studying math and ML, followed by experimentation with code and model training.
Key points
- The author built a terrain generator called InfiniteDiffusion using diffusion models and submitted it to SIGGRAPH 2026, where it was accepted.
- The project began with a single RTX 3090 Ti GPU and evolved over several years of study and experimentation.
- The author followed a structured learning plan, including math textbooks and ML courses, before diving into research and code.
- The paper was well-received, with a Minecraft mod showcasing the work and reaching millions of views.
- The author emphasizes the importance of books and foundational knowledge over AI-assisted learning.
Why it matters
This story shows how a software engineer with no formal ML background can create cutting-edge research and publish at a top conference. It highlights the value of deep learning fundamentals, careful experimentation, and the potential of diffusion models in procedural generation.
Counterpoint
The project required significant time and computational resources, which may not be feasible for all engineers, especially those without access to high-end hardware.
Sources
- Writing a SIGGRAPH paper from scratch (Lobsters)
This summary was written by a language model and may contain errors. Read the original for the full story.