You don't really need to be an expert anymore to build something with AI ☠️ I don't know much physics, but I still built this endless field where the ball follows real aerodynamics. It is built with @threejs and Claude Fable 5. You can kick the ball anywhere across an endless field or score into goals that appear in different conditions, all running at 60 FPS. When I started building this, I wanted to add as much realism as possible. Things like wind, grass movement, individual grass blade structure, realistic ball flight, fog, and atmosphere. To get there, I started researching tools that could help. I tried Claude Research, but it used 11 percent of a Max session for a single research task. So I started searching & trying other research specific AI tools. One of those research tools that I liked is @Apodex_AI. I started using it for research and Claude for building. That workflow has worked really well. One model focuses on research while the other focuses on implementation. The results are noticeably better. I was genuinely impressed by the quality of the research. It handled things like the Reynolds number, drag crisis, the Magnus effect, knuckleball side force for ball flight, the Honami model for grass motion, and the Koschmieder relation for fog and atmospheric scattering. I did not need to study or fully understand every one of those concepts. The research gave me and AI enough context to build them properly and get high quality results. That is what feels different about AI. Before, you could only build what you understood. Now, if you know how to use AI well, you can build things far beyond your own expertise. Live: Code: