Inspiration
I'm an an avid reader and I've noticed how often great writers describe things that aren't really central to the plot - what was sitting on a character's table, the objects in their room, the music playing in the background, or the little details of the place they were in. They might only get a few lines, but those details made the scenes feel more complete. They give me a sense of the character's world beyond what is happening in the story, and some times made me able to connect with the character. It made me wonder how much work goes into getting those details right. Finding the right objects, references, places or cultural details can mean hours of research, even when they're only briefly mentioned. And getting them wrong can be just as noticeable: a detail might be culturally out of place, or simply have no meaningful connection to the character. I also realised that asking an LLM to fill in those gaps doesn't necessarily solve the problem. It can produce details that sound convincing but are generic, predictable and disconnected from the character's actual background and interests. That observation became the starting point for Poppies.
What it does
Poppies is an AI character development and research agent for anyone creating a story or developing a fictional character. Whether you're writing a novel, screenplay, game or simply building a fictional world, Poppies helps you flesh out the details that make characters feel believable and distinct. You start by describing your character and what you'd like to explore. Poppies breaks this down into areas of development, creating character threads that you can explore over time. Using Qloo's Taste Graph, the agent researches cultural connections across music, film, books, places and more, then uses those discoveries to inform character details, scenes, habits, environments and creative possibilities. You can direct the agent through conversation: ask questions, provide additional context, investigate discoveries, explore alternatives or change direction. You remain in control of what gets accepted into the character. Rather than simply generating a biography or a list of personality traits, Poppies helps build a richer picture of the character's life and the world around them.
How we built it
Poppies is built with Next.js, TypeScript and CSS, with Gemini powering the AI agent and Qloo providing cultural intelligence through its Taste Graph. Gemini interprets character briefs, guides the development process and turns research into creative suggestions. Qloo helps the agent discover real cultural connections across music, film, books, places and more. We brought both together in an interactive workspace where creators can explore different aspects of a character, direct the agent, and decide which discoveries become part of their story.
Challenges we ran into
One of our biggest challenges was maintaining the agent's autonomy while keeping the experience collaborative. We wanted Poppies to actively research and develop characters, without taking creative control away from the person using it. Working with Qloo also meant figuring out how to translate open-ended character descriptions into useful cultural queries, and how to make sure the results were genuinely relevant rather than just loosely related recommendations. We also ran into occasional API delays and timeouts, which made handling responses and keeping the experience responsive important.
Accomplishments that we're proud of
We're proud to have brought AI agents into the creative process in a way that supports rather than disrupts the craft. Poppies doesn't aim to replace the creator's imagination or make creative decisions on their behalf. Instead, it takes on some of the research and exploration that goes into developing a character, giving creators more room to focus on the story itself. We're also proud of making the experience genuinely collaborative. The agent can investigate, suggest and explore, but the creator remains in control: guiding its work, challenging its findings and deciding who the character ultimately is.
What we learned
Building Poppies taught us that bringing AI into creative work is about more than generating content. The real value is in giving creators useful research, fresh perspectives and possibilities without taking away their authorship. We also learned that agent autonomy needs to be balanced with human direction. An agent should be able to independently investigate and make suggestions, but remain responsive to the creator's intent. Finally, working with Qloo showed us that discovering cultural connections is only part of the process. The agent still needs to interpret those connections in context and determine why they might matter to a particular character.
What's next for Poppies
We see Poppies evolving into a richer creative research companion for anyone building fictional characters and worlds. We want to make the agent's research more adaptive, allowing it to follow interesting cultural connections, build on previous discoveries and develop a deeper understanding of each character over time. We also want to expand the collaborative experience, giving creators more ways to guide, refine and shape the agent's work across multiple sessions. Ultimately, we'd love to see Poppies become useful beyond traditional writing, supporting character development across film, games and other storytelling mediums.
Log in or sign up for Devpost to join the conversation.