Advancements in LLM coding models have made it easier than ever to work in a domain you know nothing about, but there are still plenty of challenges you'll run into. Follow along with me as I convert Mario Kart 64 into an immersive PC VR motion simulation experience using AI-assistants to code in a language I don't know and in a domain I've never worked in. N64 game code is not well-represented in LLM training data, so I was often steering AI assistants through low-information territory, trying to recognize when a model was pattern-matching instead of reasoning, and structure my collaboration to get useful output I couldn't always fully evaluate. The lessons I took from that journey now shape how I work with AI on unfamiliar codebases in my day job, and they're more practical than one might expect from a weekend project involving virtual go-karts.
Jimmy has been in AI & Data Science for over 10 years. Currently serving as Chief AI Officer for Famelia, a local startup that offers automated recipe customization to address your personal dietary needs.
