
Survive Amid the Stars. A 3D Space Thriller.
How a small team of college students built a personalized 3D psychological horror game.
Published Jan 15, 2025
Hi! Our team would like to introduce to you Conquer Chiron, a 3D psychological horror in space. Tasked on a journey toward the star system Alpha Centauri, you will receive the help of a new prototype AI. Maintain your ship by completing various maintenance tasks and see just how helpful this AI can be!
Download the zip file, unzip the file, and then run the executable. Your machine may initially say that the executable file is unrecognized and give a warning. Do not fear - this is normal and is always thrown whenever a new file is downloaded that hasn't been verified by Microsoft or Apple for at least a year (even big companies take a long time to establish their app to not have to go through this warning screen). Simply press any ignore or override button on the warning popup and anything akin to "Run anyways" to launch the game.
Note that you need a working Internet connection to play the game, as the game needs to connect to the servers where the AI assistant model is being hosted.
Our team comprised of four members, Victor Liu, Grace Halak, Jacob Leshnower, and myself. We are or were once a part of Amoriem Labs, Yale University's Undergraduate Indie Game Development Club. Victor was our primary programmer, while Jacob was our sole composer. Grace and I managed the art of the game. In each of our roles, we spent about thirty days over the course of our Winter break away from classes to build this game. Here's how the process went:
Like many other industries, the game development field is heavily impacted by the use of AI, and we wanted to understand how AI can be used as an impactful tool when making games. That’s when we came across this game jam. After parsing through the multiple services AWS offers, we were struck by the capabilities they could provide to streamline the game development process, which we have come to know first-hand can be extremely slow and time-consuming. From the list of AWS services, we particularly wanted to use Amazon Q Developer, Amazon Bedrock, and Amazon DynamoDB because of the code assistance, LLM generative AI, and database services they provide.
We started thinking about how AI impacts the gaming experience, and one thing we observed was that AI is rarely, if ever, interactively used in runtime. We wanted to create a game where the player could directly interact with a text-generative AI whose answers affect the gameplay. After brainstorming ideas on how to integrate this player-interactive AI into the setting of our game, we landed on the narrative of an AI assistant that slowly turns evil in an old rustic spaceship, set in a dystopian sci-fi world. One of the first feelings associated with AI is fear. We fear for our jobs. We fear for humanity’s future. We fear AI. We wanted to capture this feeling in our game, so we looked for inspiration in other media that do the same. The hit game Lethal Company is one such example that, aside from its comedic gameplay, draws on this sci-fi long-future feeling of dread. Thus, we wanted to put the player in outer space, in a 3D world, with a first-person point of view, while they handle an AI that pulls on those fear strings. To round out the core gameplay experience, we wanted to expand it to include minigame-like tasks, similar to the popular game Among Us, so that the game was more than just a glorified AI chatbot. Thus, the basis of our game Conquer Chiron was born.
Conquer Chiron begins with the player performing a variety of benign minigame-like maintenance tasks and interacting with the AI named Chiron, who provides AI-generated responses to the player’s questions while piloting the ship toward the nearby star system Alpha Centauri. However, the AI soon gets annoyed with the player for ignoring its weekly AI diagnostic and malware tests and soon tries to kill the player. Can the player survive the onslaught of Chiron’s assassination attempts?
Programming: To build the game, we used Unity Engine integrated with the AWS services Amazon Bedrock and Amazon DynamoDB. During the coding process, we were aided by the AI-powered coding assistant Amazon Q Developer. Amazon Bedrock provided a connection between our game to the text-generative large-language models (LLMs) available in the AWS suite of AI models. Amazon DynamoDB is a no-SQL database that stores information in AWS and very easily fetches and writes data. We used Amazon Bedrock to generate text responses to whatever the player prompted the in-game AI assistant, Chiron, with, and we used Amazon DynamoDB to create an online leaderboard where players can compare the total time they have spent completing the game, to be used in speedruns. In total, Amazon Q Developer saved roughly 10 continuous hours of developing test code and API calls, allowing us to focus on game design.
Example:
Below in our AmazonBedrockConnection.cs script, we have a function called GenerateContext(string prompt) which takes in the prompt that the player has typed into the AI terminal where Chiron is hosted and outputs the string to query the AI model Amazon Nova Lite with. As you can see, if the player enters any prompt with the strings ‘task’, ‘do’, ‘finished’, ‘else’, or ‘next’, then the string will contain another string variable called taskToBeDonePrompt. This variable taskToBeDonePrompt randomly selects an available task onboard the spaceship and maps it to a pre-set string (not shown here) describing the task.
The output of the GenerateContext(prompt string) function is then used within a broader function in the AmazonBedrockConnection.cs script, which constructs the API call to AmazonBedrock. The below lines of code describe how we have called the Amazon Bedrock API and queried for Amazon Nova Lite’s response.
We also used Amazon DynamoDB to store user stats in the construction of our online leaderboard. Below in our DynamoDB.cs script are the functions that we have used to save to and fetch data entries in our data table in DynamoDB.

Art: For the visualization of the game, given the short timeframe of the hackathon, many models were sourced freely online from TurboSquid and the Unity Asset Store. We used these models to generate the 3D environment and scenes in Unity and Blender, respectively. These were used for the game itself and the trailer, found above. The thumbnail was generated using OpenAI image generators.
Example:


Music: Jacob composed four original soundtracks using the GarageBand software, implementing notes, chords, and drum beats accordingly to satisfy the needs of each song.
Example:

Programming: Learning how to navigate the AWS console and integrate AWS services into a game was a completely new domain. Victor, who has worked with Unity extensively before but not any AWS APIs, had watched a lot of tutorials and read much AWS documentation to understand how to properly call Amazon Bedrock and Amazon DynamoDB using their APIs to ensure a seamless game experience. While integrating the AWS services into Conquer Chiron was initially difficult, exploring these services opened Victor’s eyes to the power that they offer for enhancing the gaming experience for players.
Art: Working in a 3D environment was relatively new. Learning not only how Blender and Unity work in 3D but also how they integrate together came with a lot of technical challenges to overcome. Generating a 3D environment for a game involves much more than creating a scene for the camera; it must be optimized, interactable, and stable. This was a primary limitation when designing our 3D space. For scenes used in the trailer, our artists had to overcome the typical challenges of rigging and weight-painting a humanoid model. These challenges were great learning opportunities, and our artists cannot wait to expand their horizons further!
Music: On the music side, Jacob Leshnower was tasked with composing four musical tracks: the start screen, a peaceful track; a psychological horror track; and a boss track. Each track had to have the appropriate mood for the accompanying game segment. This was not easy at first, but after experimenting at the piano, Jacob soon figured out what was best. For the “start screen”, Jacob knew that “catchiness” was key, and composed a quick, upbeat song with some catchy synthesizer licks. For the “peaceful” track, Jacob utilized acoustic guitar and a slower tempo for a more laid-back feel. For the “psychological horror” track, Jacob used some space-sounding synthesizers with distortion for an Interstellar-inspired spectacle. For the boss theme, Jacob used the start screen track as inspiration, but created a melody that was more menacing and has a sense of despair and urgency. As an easter egg, you can hear Jacob playing his own piano on this track!
This is our final build of Conquer Chiron. This game could have a lot of potential directions and developments, but, as was always the case, the final goal of this project was to learn from the experience of developing it. This game was just us dipping our toes. We hope to use our experience and these tools in future game development, especially to fast-track the process. Please look forward to other games we make as a part of Amoriem Labs.
We have thoroughly enjoyed the hackathon process and would love to pursue similar opportunities in the future. Let us know your favorite aspects of the game and whether or not we should continue development in the comments!
We would like to thank AWS Services for the opportunity to use Amazon Bedrock, Amazon Q Developer, and Amazon DynamoDB. We would also like to thank the judges for their time spent reviewing ours and the thousands of other hackathon submissions. Finally, the team would like to thank you for playing our game and reading our blog!





