The challenges I ran into are truly too numerous to name. I use to be a software engineer decades ago but I haven't written anything beyond a shell script on the command line or a sql query in as long. Everything I have accomplished and built over the last month has been a function of LLMs for learning and coding assistance. It has helped me navigate everything. Just everything. Libraries to use, css styling, smart contract development and testing, refactoring of code, updating wide swaths of the application... Everything. I expected this to be an exploration of blockchain tech but I learned as much about LLMs and generative AI tools.

One challenge I faced was to make sure I was asking the right questions of the LLMs. One of my very first challenges was getting chatgpt to help me write a python script to generate country code images. I was able to navigate that over an hour and felt pretty good about what I could accomplish. Later I asked chatgpt to help me convert that script into something that could turn emoji flags into images I could use. Long story short, there were MANY iterations of attempts. I needed to install a different font because my library was having trouble translating the emoji into an image. I needed a different library that worked better with fonts. I needed a different font. More alternate libraries. Rinse. Repeat. Eventually, after dozens of tweaks and failures, it mentioned it was going to download flags from a flag repository and change the format into what I wanted. I was blown away. I realized that I'd started it down a path of converting things when what I really just wanted was flags. Providing LLMs with sufficient context and specifics is absolutely necessary for success.