Posts in AI
-
Hnefatafl, Viking Chess for Game Boy
On The Royal Game of Ur, the link cable nearly broke me. It was by far the hardest feature of that project, burning through most of $50 in API credits before it finally worked. Music was a close second, the part I was least happy with. So when I set...
-
Claude Design to revamp two projects
Design has always been my bottleneck. I usually have a fairly clear idea of what I want, but turning that into a consistent interface is where I get stuck. For the last redesign of my resume site, back in 2024, I ended up hiring a designer through Fiverr to get...
-
Revisiting a 20-Year-Old NDS Project with AI
Almost twenty years ago, during my PhD in bioinformatics, I needed to brush up on C++. Our lab was doing comparative genomics, studying how genomes evolve at the structural level, particularly the effects of whole genome duplications in plant genomes. One of the key tools we used, i-ADHoRe 2.0, had...
-
The Royal Game Boy of Ur
Years ago, a documentary on “The Royal Game of Ur”, one of the oldest known board games, nearly 5,000 years old, instantly piqued my interest. When I later heard Dr. Irving Finkel discuss it on a podcast, the fascination came flooding back. But this time, I had an idea: build...
-
Python to Rust: Porting My Genetic Art Algorithm
One of the first posts on this blog showed how to implement a genetic algorithm in Python. Revisiting that project, I realized it is an ideal candidate to port to Rust: it is computationally heavy, and touches on performance, parallelism, and data structures. In short, a good excuse to finally...
-
ComfyUI and SDXL: New Workflows
The latest version of our software, StableDiffusion, aptly named SDXL, has recently been launched. This has simultaneously ignited an interest in ComfyUI, a new tool that simplifies usability of these models. It’s important to note, however, that the node-based workflows of ComfyUI markedly differ from the Automatic1111 framework that I...
-
Get the most out of Stable Diffusion 2.1: Tips and Tricks
The latest version of Stable Diffusion, with its increased support for larger 768px images, is a significant improvement over previous models that only supported 512px images. While it may require some adjustment as it was trained on a more limited dataset and prompts in a different way, this new model...