Recently I've been working on designing a board game. The efforts towards this can be broken down into:
- Thinking about gameplay and game elements
- Researching and watching videos about my game's subject
- Building a webapp to help me store and manage my game data
I feel like the area I have focused on the most is predictable. After a few days with the spreadsheet I soon had a fully featured CMS to store all my elements, scrape information from the web to fill in stats, calculate rankings and points. I built it using Svelte 5 and Pocketbase. When I say I built it, what I mean is I architected it and Cursor built it. This technology is so amazing I am still dumbfounded occasionally by how effective it is.
Last week, I felt like I had my game conceptually ready to go. All the data I needed was in my database, I'd chosen the elements. Now I wanted to create a whole lot of cards. So I requested a quick MVP: I would upload a blank template and draw boxes on it where I wanted bits of data inserted, and then with the click of a button a image would be generated that I could print out and slip into a card sleeve and use to prototype my game. I hit generate and went to clean my teeth. When I came back, it was ready.
It took me less than an hour to add a whole image editing, zone drawing template editor to map the rest of the data onto card templates, and then a template cloner, and adding icons dynamically based on data. Every feature required about 3-4 sentences for me to describe the requirements and parameters, I would hit Go and a few minutes later it was ready to test.
Would I run this code in production? No. But does it do a nearly 100% perfect job of obeying instructions, implementing requirements, and just working when it's done. Yes.
Tonight I had an hour before bed to try and add a "Print" function to the app. I wanted to press a single button and for the app to generate all the cards based on the data in the database, put them all on a grid across as many pages as needed for printing at 600 dpi and then spit out a PDF. The AI did it in less than five minutes. I barely had time to watch some NBA highlights. This kind of functionality would have probably taken me a week to get right on my own. It did it in one shot. I didn't spend all night programming, I spent most of it taking a long walk along the beach with Vanessa as the sun slowly set.
Is there an AI bubble? It's a very hot question right now. I'm sure I'll come back to this one way or another, and this anecdote is not an answer to that. But the power of this technology is so potent it's impossible not to be impressed.
I downloaded the PDF and my game dream had come to life on my screen.
Now the question is, do I have any friends to play this game with? Or do I need to add a new feature to get the AI to play it as well...