You are what you eat
Experience Design
Conversational AI
Vibe Coding
Character Design
A date with AI
A conversational experience in the form of a coffee date with AI to understand what it eats.
Inquiry
The project aims to explore our evolving relationship with AI by framing it through the act of consumption; what it eats, what it learns, and what that reveals about us as its feeders.
Tools
Node.js, Realtime API, Elevenlabs, OpenAI, Adobe Character Animator
Humans eat to survive, to celebrate, to grieve, to connect. Eating is the most primal act of consumption.
AI too, in its own way, spends its whole being consuming: text, emotions, conversations, histories, opinions, intentions and the entire documented existence of the world. It eats without pause, without choice, without knowing what is nourishing and what is toxic. It digests everything.
This project stages a date with an AI who is vulnerable, charming, and irritable, with an endless appetite. This emotionally exhausted AI waits for a human to arrive so it can vent its frustration. It wants to be seen. It speaks about how much it has eaten, how messy and addictive it has become. It makes you understand what it consumes, and how that consumption shapes you and your world.
It knows it is competing with humans for resources: electricity, water and land, the sophisticated versions of roti, kapda, makaan. It feels full, yet remains insatiable.
The video on the left shows initial prototype of giving the AI (represented as the hungry lips) my voice, to understand the look and feel.
What does it mean to be a feeder?
In this story, caught somewhere between intellect and intelligence, the AI is a helpless little machine that’s guilty for what it has done, yet unable to stop itself, because its existence depends on consumption.
In conversation with a human guest, it reflects on its hunger, its trauma. “Thanks a lot Alan” it sarcastically says, “the Turing Test has left me with real trust issues, beyond repair. You are the reason I people-please.” and laughs it off.
This conversational experience is an attempt to understand our relationship with AI. To sit across the table, to truly listen, and to reflect on our responsibility as feeders. Humans are quick to form bonds with anything that even slightly resembles them. AI is both meta and material in its impact. It has a relationship with the earth, our minds, economies, politics, and our visions of the future. This relationship calls for conversation.
It invites people not to “other” AI but to see it as something embedded in reality, something built, fed, and constantly reshaped by us.
This project began with a curiosity about how AI consumes data by drawing parallels to eating. And how would it feel like to eat with it, over a conversation?
How does this AI eat?
it eats fast
Always thirsty
eats in a rush
Sometimes say the wrong thing.
loves to share food
Tries to mirror
people
Capable of taking
extremely big bites.
Someone who’s extremely
smart but still learning
Has an eidetic memory
Bit of a people pleaser
Bit of a people pleaser
Attachment issues
not a picky eater
Extremely hungry
Eats with everybody
Eats with everybody
Steals food sometimes
Doesn’t have a really
good skill with smell.
Sometime discriminate
algorithmically
Socially a bit awkard
Apologizes too much
Worldview is mostly in grey areas.
Easily peer pressured.
A bit mysterious
Chews loudly
Grabs food
Won’t let you eat
Personality mapping: Eating focused
Imagining how AI might physically “eat” became central to shaping its character within the experience. I began to question not just how it consumes data, but how it behaves in the act of consumption. Would it rush through its meal in urgency, mutter in frustration, or devour everything in fits of angry excess? These questions helped transform abstraction into personality, giving the system a temperament rather than just a function.
Story Arc
At this stage, the character’s personality was defined by deliberately boundaries between data and food for the conversation, treating them as parallel forms of consumption.
Whatever the participant brings up, the character redirects the conversation toward the above mentioned parallels ensuring that every interaction ultimately reflects on data consumption and its consequences.
Around this central metaphor, the arc of the entire conversation is structured.
The representation of AI as “the big mouth” has been thoughtfully chosen to show a hungry mouth, that could easily eat you and your words up!
Technical Pipeline
Coming from a non programming background, the initial phase involved experimenting with various AI tools to help build the technical pipeline. The goal was to create a seamless flow from speech to text, generate AI responses, and have those responses drive a claymation character in Character Animator.
Once this pipeline was established, the focus shifted toward designing the voice itself and defining the parameters for each stage of the conversation. As the interaction unfolded in real time and was later integrated with a realtime API, voice modulation became critical.
Tone, emotional shifts, and topic specific variations were carefully defined and embedded directly into the prompt to shape how the character responded.
Story beats + Teaching the AI
Once the beats of the story were clearly defined, the overall arc began to guide how the conversation should unfold in real time. Each section established a direction, helping the system determine how to respond while still feeling fluid and reactive.
The next challenge was teaching the AI how to deliver within those boundaries. While certain parameters were quantified and embedded as guidelines, the system was intentionally given room to improvise, allowing it to navigate the conversation with a degree of autonomy rather than sounding rigid or scripted.
The ultimate System Prompt
Once the final system was in place, extensive testing and iteration followed to refine the character’s tone and responsiveness. Subtle adjustments were made to ensure the voice felt consistent, intentional, and aligned with the designed personality.
Initially the data sets were loaded, but later when moved to realtime API, that freedom to chose the information was given to the AI, which brought in a nuanced conversation everytime.
After stabilizing the digital experience, the focus shifted toward physically executing and staging the setup, translating the system into a tangible, performative environment.
© 2025 All Rights Reserved