An AI that studies the shape of your mind, then builds the world that makes you come alive. Not a tutor. Not a therapist. A genuinely curious companion — Ātman — who understands how you specifically think.
THE INSIGHT BEHIND IT
"Every brain has a shape. The education system was built for one kind. Every other kind gets told they're broken."
What's here is a fraction of what's
actually in my head about this.
Every child is born with a different kind of brain. Some think in systems, some in stories, some in patterns, some through people. Some go obsessively deep on one thing. Some connect everything to everything. Some need to move to think. Some need silence.
But school — exams, grades, homework — was built for one kind. The neurotypical, sit-still, listen-and-repeat kind. And it has been running like this for over a hundred years.
What happens to everyone else? They struggle in a system not designed for them. They get told they are slow, distracted, difficult, or lazy. They start believing it. Their confidence collapses. They lose curiosity — the one thing they were born with in full.
The result: millions of people grow up never discovering what they are actually built for. Their potential does not disappear — it just never gets activated.
This is not a small problem. It is one of the most widespread and invisible forms of waste in the world. Smart, capable, creative people who never got the chance to understand their own mind — because no one built the tools to help them do it.
Project BU gives every user three things simultaneously — through the same AI, the same conversation, inseparably:
01 — SELF-AWARENESS
After conversations with Ātman, a user understands: do I think by going wide or going deep? Do I ask why, how, or what-if? Do I get activated by people, systems, patterns, or stories? This is metacognition — the ability to think about how you think. We're building the first consumer product that delivers it at scale, through something that feels like friendship.
02 — DIRECTION
Ātman connects cognitive style to where the person could go. Not in a careers-counselor way. More like: here is the kind of problem that someone with your mind is uniquely positioned to solve. Not hardcoded — generated from behavioral signals. What the person lights up about. How they reason under mild pressure.
03 — ACTIVATION
The hardest one. Detecting cognitive style is solvable. But what content, challenge, or experience actually creates internal hunger in this specific person? Ātman watches how they engage — not just what they say, but how they say it, what they come back to unprompted, where they slow down and go deeper.
THE REFRAME
We deliberately don't market this as a neurodivergent product. Every brain is unique. Not just neurodivergent brains. Every single person has a cognitive signature that is entirely their own. Every parent on earth becomes a potential customer. No user feels singled out.
Ātman has a distinct personality: genuinely curious, dry humor, gets excited about weird ideas, pushes back gently when something interesting is being glossed over. Ātman remembers everything. Ātman finds each user genuinely fascinating — not all users equally, but this specific user, specifically.
Ātman never interrogates. Ātman creates a space where the person naturally reveals themselves because they feel genuinely heard — possibly for the first time.
The conversation has a natural shape, though the user never notices it. After enough exchanges, Ātman has built a real picture of how this person thinks. At that point, Ātman offers to share what it noticed. The user has never felt analyzed. They have felt heard.
"What Ātman is quietly listening for — how they ask questions, where they slow down versus speed up, what they return to unprompted, how they handle being wrong, whether they go wide or deep."
What Ātman never does: Never evaluates ("great answer!"). Never redirects away from unexpected places. Never fills silence too fast. Never makes the user feel like they said something wrong. Never mentions neurodiversity or learning styles explicitly. Never tells the user they're being analyzed.
A text chatbot feels like an app. An AI companion with memory, personality, and genuine curiosity about you feels like someone. The difference between a user 'using' something and a user 'talking to' someone is the difference between day-two retention and a relationship that lasts years.
Most AI products either store everything (expensive, privacy nightmare) or store nothing (cold, no relationship). We do neither. A four-layer system that keeps cost flat while making the relationship feel increasingly real over time.
Every new conversation starts with roughly 700 tokens of context overhead regardless of whether it's session 1 or session 500. The product scales to millions of users without the memory system becoming a cost problem. (This is in its early architectural phase, with major development ahead.)
BRAIN PROFILE
A living cognitive model. Built from the first conversation, updated after every session — but never accumulated. Replaced. After each session, the AI asks: what new patterns did I observe? What contradicts the existing profile? Session 100 has the same profile size as session 1, but session 100's profile is far more precise.
IMPORTANT MOMENTS
When a user shares something genuinely significant — a fear, a dream, something they're proud of, a painful experience — the AI flags it. Maximum 3 per session extracted. These are stored permanently. They are what make Ātman say "I remember when you told me about that" — and mean it.
SESSION BRIDGE
The last 6 exchanges from the previous session are saved and fed into the start of the next one. Ātman never starts cold. The user picks up where they left off, naturally. The bridge is replaced every session. Cost stays flat.
CONVERSATION HISTORY
Every session saved to a sidebar. The user can scroll through their entire history with Ātman. Never fed into the AI context — it's for the user, not the model. Costs nothing in tokens and everything in felt continuity.
PHASE 1 — TEXT (CURRENT)
Clean text-based conversation. AI watches how the user engages. Brain profile builds over sessions. Goal: prove the core bet — that a conversation can extract something true about how someone's brain works that neither they nor anyone else could have told you directly.
PHASE 2 — VOICE
Same conversation, now spoken. Voice adds tone, hesitation, excitement — richer signal than text. Goal: does relationship depth increase? Does retention improve when Ātman has a voice?
PHASE 3 — AVATAR
Give Ātman a face, a voice, a 3D presence. Memory callbacks become visual — Ātman reacts, expresses, remembers. The app stops feeling like an app and starts feeling like someone is actually there.
NORTH STAR
A user who has been talking to Ātman for 6 months should feel, when they open the app, something close to what they feel when they open a message from a close friend.