Inspiration
My inspiration for Dicetrix is a lifelong love of number and puzzle games. Originally Dicetrix was envisioned quite differently, but the idea evolved over time as other puzzle games changed the playing field. I wanted to do something that was fun, playable, and relatable to other puzzle games but not make yet another clone of a successful concept. So I set out with that goal in mind, allowing for the timeline and trying to leverage as much of Kiro's automation as possible to get something functional.
What it does
Dicetrix is a spin on the 'falling-piece' puzzle game genre. In Dicetrix, groups of randomly rolled dice are combined into tetrominos and fall towards the bottom of the grid. The player's goal is to match 3 or more like numbers to make a match, destroy the pieces, and score points without running out of available space. Different levels of difficulty will increase the complexity of the game by adding dice with higher face values and other mechanics for a challenging game such as negative boosters that reroll other dice or increase the number of faces. For more casual players that don't want an excessive challenge, there is a Zen mode with mechanics aimed at a simple and relaxing gameplay.
How we built it
Dicetrix is built on the Phaser 3 template provided on Reddit's developer's forum. I leveraged Kiro's Spec agent feature to lay out a rough project outline after spending a bit of time creating a prompt specialized to the information I could find on Kiro's capabilities and features. Since the deadline was fixed, I opted to aim for a more condensed implementation that focused on having the core mechanics in place and a functional demo that can be extended with player feedback. I fell short of 100% of what I wanted implemented, but I can attribute much of that to my own lack of familiarity with how to properly harness the agent until well into the process.
Challenges we ran into
It took me a little while to learn how to rein in some of Kiro's agentic features so that they didn't get farther ahead in the implementation than I was intending. This led to quite a few unused files, variables, types and miscellaneous other things that were replaced just a prompt or two later. I ended up having to roll-back and undo changes that were made unintentionally a couple times because something got 'lost'
Accomplishments that we're proud of
I put this game demo together in around 7 days, and have it playable on Reddit. There's enough extensibility in the code that was generated to allow me to continue building more features into the game as I get feedback.
What we learned
Once I learned how to aim the AI agent in the direction I wanted more precisely, I was able to very easily get a task laid out and completed while keeping tangents to a minimum. Testing is important and making sure that you don't accidentally queue commands from two different contexts in the wrong order.
Log in or sign up for Devpost to join the conversation.