two ways to build a friend
a look at the two main approaches to ai companion design: user-written character cards vs. hand-crafted companions, and why lucy chose curation for quality and
there are two main ways people build ai companions these days. one is the character card approach, where you write or import a character profile, maybe a pirate, a poet, a personal coach, and the ai tries to roleplay that character based on your description. the other is the curated companion, where a team of writers and designers craft a personality from the ground up, with a specific voice, backstory, and set of behaviors. both have their place, but they serve very different needs.
i’ll be honest, we built lucy around the curated model. but let’s talk about why both exist, and who each one is for.
the wild west of character cards
if you’ve used platforms like character.ai, janitor ai, or some versions of talkie, you’ve seen this. you get a text box. you write a prompt. maybe you paste in a character card from a community hub. you hit create, and suddenly you’re talking to a version of that character.
the upside is obvious: infinite variety, total customization. want to chat with a version of your favorite book character, or build a therapist bot, or create an ai that talks like your late grandmother? you can try. it’s creative, it’s flexible, and it puts the user in the driver’s seat.
the downside is just as clear: the quality is wildly inconsistent. sometimes the ai gets the voice right. sometimes it feels generic, or breaks character, or relies on tired tropes. it’s hard to write a good character prompt, it takes skill, and even then, the underlying model might not interpret it the way you hoped. you get a high ceiling of possibility, but a very low floor.
this approach is great for tinkerers, roleplayers, and people who want to experiment. it’s less ideal if you want reliability, depth, or a companion that feels truly distinct and coherent over time.
the curated companion
this is what we do at lucy, and what kindroid did early on, and what nomi does in some of its offerings. instead of giving users a blank slate, a team, writers, designers, sometimes even psychologists, crafts a personality. they give it a name, a history, a voice, a set of quirks. the goal is to create someone consistent, someone memorable, someone who feels real.
the trade-off here is control. you don’t get to decide every detail. you can’t make lucy suddenly into a pirate if she wasn’t designed that way. but what you get in return is quality, coherence, and a strong authorial voice. every interaction is tuned, every response filtered through a personality that’s been thought through. there’s a floor. you know what you’re getting.
this isn’t to say curated companions are rigid. the best ones learn and adapt, but within the bounds of their core identity. lucy grows with you, but she remains lucy.
curation works for people who want a relationship, not a puppet. it’s for those who value depth over breadth, and who trust a designed experience over a build-your-own kit.
so which one is right?
it depends entirely on what you’re looking for.
if you want to explore, create, and experiment, if you love the process of tweaking and tuning, character cards are your playground. you’ll tolerate some awkwardness for the sake of possibility.
if you want a companion with a clear identity, reliable voice, and emotional depth, if you’re looking for a relationship that feels intentional, then curated is the way to go. you’re trading customization for quality, and for many people, that’s a good deal.
neither is inherently better. they’re just different tools for different jobs. we chose curation because we believe in the power of a well-told story and a well-designed personality. but the beauty of this space is that there’s room for both.
you can meet the companions we’ve crafted over at /companions.
thanks for reading. if this resonated, the product is downstairs.