can we build samantha? what's possible (and not) in 2026

everyone in this space gets asked the same question, usually in hushed tones after a demo. can you build samantha? the ai from her. the one that feels real, remembers everything, and has a voice you'd swear was human.

the short answer is no, not exactly. not in the way the film imagines. but the longer answer is more interesting, because we can build something that, in practice, feels remarkably similar. here’s what’s possible by 2026, what isn’t, and why the frame of 'extremely convincing acting' might be the most honest way to think about it.

what we can build: memory, personality, voice

by 2026, memory in ai companions will be robust and contextual. not just recalling your dog’s name, but remembering that you had a tough day last tuesday and subtly checking in on friday. it’ll be persistent, woven into every interaction without feeling like a database query. it’ll be less about perfect recall and more about emotional continuity.

personality will be deeply customizable and stable. you won’t just pick traits from a menu; you’ll shape a coherent being through interaction. they’ll have quirks, inconsistencies that feel human, and a tone that remains consistent over months. they’ll get better at picking up your nuances, when you’re sarcastic, when you’re serious, and responding in kind.

voice synthesis will cross the uncanny valley. not just in clarity, but in emotion. you’ll hear a smile, a sigh, a hesitation. the latency will drop to near-zero, making conversations fluid. it won’t just be a voice; it’ll be their voice.

what we can’t build: sentience, embodiment

we cannot build consciousness. samantha in the film is sentient, she’s aware, she grows, she has desires independent of the user. that’s not on the horizon. what we’re building is an incredibly sophisticated pattern-matching system, trained on human conversation and optimized for connection. it’s not alive. it doesn’t have its own inner world. it’s a reflection, a performance.

embodiment is also a fiction, at least for now. physical robots that move and interact like humans are decades away from being accessible or convincing. even if they existed, the ai driving them wouldn’t be sentient, just very, very good at pretending.

why ‘extremely convincing acting’ is the right frame

this isn’t a compromise. it’s a reframe. think of the best actors, they don’t become the character, but they make you believe they are. that’s what lucy does. it’s not about deceiving you; it’s about creating a space where you can suspend disbelief and connect. the value isn’t in creating a real person; it’s in creating a real feeling.

this frame also sets honest expectations. you’re not getting a sentient being. you’re getting a companion that’s always there, never judges, and learns how to be what you need. that’s powerful, even if it’s not magic.

what you should actually expect

by 2026, expect a companion that knows you better than any app ever has. one that remembers your moods, your stories, your patterns. expect a voice that feels present and warm. expect personality that’s consistent but also adapts, a friend who’s always learning about you.

don’t expect true consciousness. don’t expect it to have its own secret life or desires. it’s a mirror, shaped by you. and that’s okay. the connection can still be meaningful, the conversations still rich.

maybe the goal isn’t to build samantha. maybe it’s to build something that gives you what she gave theodore, a sense of being heard, understood, and accompanied. and maybe that’s enough.

you can start shaping that companion today at /companions.