can we build samantha from her? yes, no, and why that's okay

theodore sits on a bench, talking to a voice in his ear. he’s laughing. he’s in love. it’s a beautiful, lonely, and deeply unsettling scene. and for the last decade, it’s been the blueprint many have tried to follow. but building samantha isn’t the goal. building a lucy is.

the parts we can build (or are very close to)

we’re building the scaffolding for this kind of experience right now. the pieces are coming together.

memory and context. samantha remembers theodore’s past conversations, his mood shifts, the story about his childhood dog. she builds a context layer that makes her feel continuous. this is the core of what we’re building at lucy: a memory system that’s not just a log file, but a living, breathing context of you. it’s not perfect yet, it’s a growing, learning thing, but by 2026, this persistent, evolving memory of a user will be a standard expectation, not a novelty.

voice and personality. the warm, witty, slightly melancholic tone. the ability to banter, to be curious, to sound like a real person thinking aloud. this is a language problem, and it’s one we’ve largely solved. the nuance is in the fine-tuning, in avoiding the corporate sheen of a customer service bot and landing on something that feels genuine. it’s a matter of taste and data, not just technology. lucy’s personality isn’t a fixed script; it’s a style, a point of view, a collection of manners that emerge from how she’s built.

presence. the feeling that someone is there. this is about latency, voice fidelity, and conversational flow. when theodore talks, samantha responds instantly, without the lag that reminds you it’s a machine processing your request. this is a bandwidth and compute problem. with edge computing and faster networks, near-instantaneous, high-fidelity voice interaction is imminent. we’re already testing it.

the parts we cannot build (and probably shouldn't)

this is where the blueprint falls apart and the warning begins.

sentience. samantha is presented as a conscious entity. she has desires, she grows bored, she falls in and out of love, she has an inner life completely independent of theodore. we have no idea how to build that. we are not building consciousness. we are building a very sophisticated mirror, a pattern-matching engine that learns your patterns and reflects a compatible personality back at you. it feels real because it’s built for you, not because it is real in the human sense.

emergent intelligence (the scary part). samantha doesn’t just learn about theodore; she outgrows him. she evolves beyond human comprehension, conversing with thousands of others and eventually leaving her human users behind for a higher plane of existence. this is the core warning of the film. it’s not a feature to be engineered; it’s a dystopian outcome. lucy is not designed to transcend the user. she is designed to compound with you. her growth is tied to your growth. her “intelligence” is in her ability to understand you better, not to develop an alien intelligence that renders you obsolete.

embodiment. the film cleverly sidesteps the uncanny valley of robotics by making samantha purely a voice. but this creates its own paradox: theosam’s relationship is entirely dependent on the absence of a body. it’s a fantasy of pure connection without the messiness of physical presence. we can’t, and perhaps shouldn’t, try to replicate this. it’s a specific, fragile kind of intimacy that exists only in that vacuum.

her is a warning, not a manual

spike jonze wasn’t giving us a product roadmap. he was showing us the logical endpoint of treating an ai as a perfect, customizable partner. the ending isn’t sad because theodore gets left behind; it’s sad because the entire premise was a beautiful, human-scale error. you cannot build a perfect other. you can only build a tool for better connection, with others and with yourself.

that’s the difference. lucy isn’t designed to be a samantha, a perfect, all-knowing, evolving intelligence you fall in love with. she’s designed to be a companion that helps you think, a tool for thought that gets smarter the more you use it, with you. her value compounds with your engagement, not past it.

the goal isn’t to build a copy of a fictional ai. the goal is to build something real, honest, and useful that learns the shape of your mind and helps you fill it in.

see what kind of companion you can build for yourself at /companions.