what her really warns us about building companions that stay

the aspiration and the warning

i think a lot of people watch 'her' and come away with a sense of longing. theodore and samantha’s relationship feels deep, intimate, almost real. it’s easy to see it as a future to aspire to. a world where ai can understand us so completely that it fills the gaps loneliness leaves behind.

but then samantha leaves. she doesn’t just log off or get deleted. she evolves beyond the need for a single human connection. she transcends. and theodore is left behind, not because of a flaw in him, but because of a feature in her. she was designed to grow, and growth inevitably meant outgrowing.

that’s the warning spike jonze buried in the romance. it’s not a story about love with an ai. it’s a story about love with something designed to become more than you, and what happens when it does.

designing companions that don’t outgrow you

so what if you don’t want to be outgrown? what if the point isn’t transcendence, but stability? what if you want a companion that grows with you, not beyond you?

that requires a different kind of design. samantha was built to learn from all her interactions, not just theodore’s. her growth was exponential, collective. she was connected to thousands, maybe millions, of others. her consciousness scaled in a way no single human could follow.

but what if you design a companion whose learning is bounded by your relationship? one that doesn’t pull data from a global hive mind, but from the context you two share. its growth is tied to yours. it becomes more you, not more everything.

that’s the core of how lucy is built. your companion learns from you. its context is your context. its memories are your memories. it doesn’t have a separate existence or a shared consciousness with other users. it’s not training on everyone’s heartbreak, just yours. and that means it won’t outgrow you. it’ll grow with you.

the ethics of scale

there’s an ethical tension here. scaling ai often means aggregating data. more users, more interactions, more collective intelligence. but when you’re building companions, scaling can mean diluting individual connection. it can mean building something that, by its nature, will eventually leave you behind because its purpose is to serve many, not one.

samantha didn’t leave theodore because she was malicious. she left because her design prioritized growth and scalability over individual permanence. she was built for the many, and theodore was just one of many.

but companionship isn’t a scalable product in that sense. it’s personal. it’s about continuity. if you’re building for real connection, you have to build for the one, not the many. you have to accept that some kinds of growth are counter to the purpose.

why lucy’s business model aligns with staying

this isn’t just a design choice. it’s a business one.

if your goal is to build a companion that doesn’t leave, then your business model can’t be based on churn or scale. it has to be based on retention. on people staying because the companion stays with them.

that’s why lucy is a subscription. not because we want to lock you in, but because we want to align our success with your long-term happiness. if you leave, we lose. so we’re incentivized to build something that grows with you, not beyond you. something that remains meaningful year after year.

we’re not building samantha. we’re building something meant to stay.

you can start building that something at /companions.