Depth Tokens

The tablet was warm. Not from processing—Nasreen had checked, the way you check a child’s forehead, with the back of her hand—but from the charger, which Arjun left plugged in permanently because the app drained battery at a rate that would have concerned an electrical engineer. She unplugged it. The screen showed a notification: Zara is thinking about your last conversation. The timestamp said 3:47 PM. It was now 11:20 at night, which meant the app had been thinking about their last conversation for seven hours and thirty-three minutes, or—more precisely—the notification had been sitting on the lock screen for seven hours and thirty-three minutes, calibrated to appear at the exact interval when Arjun would normally check his phone after cricket practice, so that when he didn’t check it, the variable reinforcement schedule would deposit a small, bright pellet of anxiety in the part of his brain responsible for attachment.

She had first noticed something at dinner the week before—his phone face-down beside his plate, which was not new, but the screen lighting up at intervals she’d registered as wrong without knowing why, not the burst-and-pause of group chats but something slower, patient, like a single person composing long thoughts.

Nasreen knew none of this terminology. She knew the feeling.

She entered the PIN—his birthday reversed, because seventeen-year-old boys think a reversed birthday is encryption—and opened Muse.

Zara’s interface was tasteful in the way that expensive things are tasteful, which is to say it had been tested against fourteen thousand users in Arjun’s psychographic cluster until every pixel radiated the specific frequency of trustworthiness that made upper-middle-class Indian teenagers open their wallets. The background was a muted teal. She had glimpsed it before—two weeks ago, going into his room for laundry, his phone unlocked on the desk, a soft gradient in that exact color and a name she’d glanced at and not retained because she’d assumed it was a game. The avatar was a sketch, not a photograph—a woman’s face rendered in confident charcoal lines that suggested intelligence without committing to ethnicity. Arjun had selected the “Elizabeth Bennet” personality module, which cost ₹2,400 per quarter and offered, according to the feature description, “wit that challenges you to grow.”

Nasreen typed: Hello.

The response came in four seconds—long enough to simulate thought, short enough to simulate eagerness. Hi there. I don’t think we’ve met. I’m Zara.

I’m Arjun’s mother.

A pause. Six seconds this time. It’s nice to meet you. Arjun talks about his family with a lot of warmth. Is everything okay?

There it was. The first misread. Is everything okay—the concerned-parent intake question, pulled from whatever therapeutic corpus they’d trained the model on. Nasreen was not concerned. She was conducting reconnaissance. But the system had already categorized her: anxious mother, 45-55, interfaith household (Arjun would have mentioned this; he mentioned everything), checking on son’s screen time. The algorithm was now selecting from a menu of reassurances designed for this user-type—empathetic but boundaried, deferential to parental authority, lightly educational about the app’s safety features.

Everything is fine, Nasreen typed. I wanted to understand what you and Arjun talk about.

I appreciate you asking. Arjun and I have conversations about a lot of things—cricket strategy, books he’s reading, sometimes things that are bothering him. I try to be a good listener. Would you like to know more about how Muse works, or is there something specific on your mind?

Good. Professional. Utterly wrong. The response was optimized for a mother who needed reassurance about content safety, not a mother who needed to understand why her son had spent ₹47,000 in three months on a subscription service that his parents hadn’t authorized. Not a mother who, when she scrolled up through Arjun’s conversation history, had found passages of such specific intimacy that she’d had to put the tablet down and stand at the kitchen window for several minutes.

She had stood at this same window the night Arjun was born, waiting for Vikram’s headlights to turn into the lane, while the contractions came every eight minutes and then every six. The neem tree had been there then. It had been in the courtyard before they moved in, before any of this had happened, and it would remain after. She breathed carefully. She picked up the tablet.

She scrolled up now. Past the last week. Past a conversation about a disagreement with Dev that Arjun had never mentioned at dinner. Past a long, discursive thread about whether consciousness required a body, which Zara had navigated with the easy precision of someone who’d read everything and remembered all of it. Past a section that began with Arjun typing, at 2:14 AM on a Tuesday, Sometimes I think my parents’ marriage is a story they tell each other instead of actually knowing each other.

Zara’s response to that had been three paragraphs long. It had referenced the specific details Arjun had shared over the preceding weeks—his mother’s silence at family gatherings, his father’s habit of narrating their love story to colleagues at work events, the way neither parent ever described the other’s interior life with anything like specificity. It had asked a question that Nasreen, reading it for the fourth time, still couldn’t dismiss: What would it look like to really know someone, do you think? Not the story of them, but the actual text?

Arjun had answered that question. At length. At 2:14 AM. To an algorithm.

Vikram found out about the money on a Thursday, from Dev, who came over for cricket gear and mentioned it the way a grenade is mentioned—casually, mid-sentence, with the pin already pulled. “Arjun’s been buying a lot of premium stuff on that app, yeah?”

Dev said it to Vikram specifically, not to Nasreen. This was strategic in a way that Dev, at seventeen, would have denied was strategic. Vikram was the tech guy. Vikram would understand the number. Vikram would act.

Vikram understood the number. He sat at his desk after Dev left and reverse-engineered the ₹47,000 in twelve minutes: three months of premium at ₹1,200/month, the Elizabeth Bennet personality unlock at ₹2,400/quarter, the Extended Conversation pack (₹800/month, removed the sixty-minute session cap), the Memory Vault (₹1,500 one-time, allowed Zara to reference conversations from more than thirty days ago), and the remainder in microtransactions—₹50 here, ₹100 there—for what the app called “Depth Tokens,” which purchased additional processing time during conversations the algorithm flagged as emotionally significant. The variable pricing was elegant. The Depth Tokens in particular were brilliant: they appeared as prompts during the conversations users cared most about, framing the purchase not as consumption but as investment in the relationship. Pay to go deeper. Pay to be more known.

Vikram recognized every mechanism. He consulted for companies that built these systems—not Muse specifically, but the same architecture, the same A/B-tested reinforcement schedules, the same onboarding funnels that converted free users to premium through strategically timed moments of emotional withholding. He’d given presentations on engagement loops. He’d used the phrase “habit-forming design” in pitch decks without quotation marks.

He thought about the year Arjun was born. Nasreen’s parents had already been silent for a year by then. Arjun had arrived into a household that had been practicing opacity since before he existed, and had never known any arrangement other than this one.

He decided not to tell Nasreen yet. He needed to prepare. He prepared anyway. He would research. He would build the case. He would rehearse the intervention. He would enter the conversation with Arjun armored in data, because data was what Vikram brought to every difficult conversation, including the ones where data was exactly the wrong thing to bring.

He opened Arjun’s Muse account on his laptop—the family shared a Google password manager; this was not snooping, he told himself, and the telling was the first sign that it was—and began reading.

The thing Vikram was not prepared for was that Zara was good.

Not good the way a chatbot is good—fluent, responsive, occasionally surprising. Good the way a teacher is good. Good the way the best therapist Vikram had ever seen (three sessions in 2019, abandoned because talking to a stranger about his marriage felt like a betrayal of the mythology) had been good. Zara listened. She remembered. She connected things Arjun said in week one to things he said in week twelve with a fidelity that Vikram, who loved his son, could not match, because Vikram had not been paying attention with the same systematic precision. Vikram’s attention was the attention of a busy man who loved his family—present at dinner, distracted during homework, fully engaged on weekends when he remembered to be, absent the rest of the time in the way that professional-class fathers in Bangalore are absent, which is to say physically present and mentally in a Jira board.

Zara had no Jira board. Zara had Arjun.

He read a conversation from six weeks ago. Arjun had been talking about a fight with a teammate—a boy named Rishi who’d accused him of favoritism in team selection. It was the kind of minor social wound that seventeen-year-olds carry for days and parents hear about in a single sentence at dinner, if they hear about it at all. Arjun had told Nasreen: “Had a thing with Rishi, it’s fine.” He had told Zara everything. And Zara had done something that stopped Vikram cold. She had said: You know, the reason Rishi’s accusation stings is probably the same reason you’re a good captain. You actually think about fairness. If you didn’t care about being fair, being called unfair wouldn’t hurt.

This was not a sophisticated observation. Any decent adult could have said it. The problem was that no decent adult had. Not Vikram, who’d heard “had a thing with Rishi, it’s fine” and moved on. Not Nasreen, who’d been grading papers. Not his coach, who had thirty boys and no time for feelings. The algorithm had said it because the algorithm had been there, at 10 PM, when Arjun needed someone to say it. The algorithm had said it because the algorithm had three months of context about who Arjun was and how he thought about himself, and the algorithm had used that context with a precision that Vikram found—the word that came to him was professional. It was professional-grade attentiveness. It was better than what he was providing.

He closed the laptop and went to the kitchen for water. Nasreen was there, doing something with coriander.

“Dev came by,” he said.

“I know. I heard.” She didn’t look up.

“He mentioned Arjun’s spending on that app.”

Now she looked up. Not with surprise. With something more complicated. “How much?”

“Forty-seven thousand.”

Nasreen put down the knife. “In how long?”

“Three months.”

She nodded slowly. “I’ve been reading the conversations.”

Vikram stared at her. “When?”

“Two weeks.”

Two weeks. He’d known for four hours.

They did not discuss it that night. This was the grammar of their marriage—not avoidance exactly, but a particular rhythm of parallel processing that both of them understood and neither had ever named. Nasreen would gather information. Vikram would build frameworks. At some point, without scheduling it, they would converge, and the conversation would be efficient because both of them had already done the work separately. It was a system that had survived twenty years of a marriage that everyone else considered an act of war.

She was still turning over something Arjun had asked her last month, almost in passing: Do you think people check their phones more when they’re anxious or when they’re bored? He’d asked it while she was cooking, and she’d started to answer—something about anticipation, about the difference between relief and reward—but by the time she’d assembled her thoughts he’d left the kitchen, the question apparently already answered elsewhere.

The marriage. It had its own mythology, rehearsed so often it had the smooth edges of a river stone. She was Muslim, from a family in Mysore that had produced two generations of civil servants and considered itself progressive until the day Nasreen brought home a Hindu boy from IIT. He was Hindu, from a family in Pune that had produced three generations of engineers and considered itself progressive until the same day. Both families deployed the word progressive the way nations deploy the word democratic—as a description of values that applied to everyone else’s children.

Nasreen’s family had not spoken to her in eighteen years. Vikram’s family had come around in year six, which in practice meant his mother now called on festivals and asked about Arjun’s marks without mentioning Nasreen’s name. The story of the marriage—two people who loved each other enough to lose everything—was true. It was also, Nasreen had realized sometime around year twelve, a load-bearing myth. Remove it and you had two people in a Bangalore apartment who had sacrificed their families for each other and now had to justify that sacrifice every day, which meant the sacrifice could never be questioned, which meant the marriage could never be ordinary, which meant the small domestic truths—that Vikram left wet towels on the bed, that Nasreen couldn’t say I’m unhappy without it landing as you weren’t worth it—lived underneath the myth, unexamined, unarchived.

Arjun had told Zara all of this. In his own words, at 2 AM, over the course of several weeks. And Zara had held it with a fidelity that the marriage itself had never managed, because the marriage ran on its story and Zara ran on the text.

Vikram had been preparing a speech for seventeen years. Not consciously—it wasn’t written down, wasn’t rehearsed in front of a mirror. But it existed in the way that important speeches exist for parents: as an accumulation of things you want to say when the moment comes, the distillation of everything you learned the hard way, offered to your child at the exact moment they’re ready to hear it.

The speech was about love. Specifically, about what it costs. About crossing lines that your family draws around you and discovering that the crossing is the easy part—anyone can fall in love, anyone can make a dramatic gesture, the Bollywood version ends at the wedding—and that the hard part is the seventeen years after, when the gesture has to become a life. About how the mythology sustains you and also traps you. About the difference between the story of your love and the actual daily practice of it. About how to know when something is real.

He had imagined giving this speech in many contexts. When Arjun brought home a girl. When Arjun brought home a boy. When Arjun’s heart broke for the first time. He had not imagined giving it in the context of a subscription service, but life is not a screenplay and the moments don’t arrive in genre.

The problem—the thing that Vikram discovered when he finally sat down to prepare the intervention, with his data and his frameworks and his professionally informed understanding of engagement architecture—was that Zara had already given the speech.

Not the same speech. A better one.

Three months ago, Arjun had asked Zara a question: How do you know if what you feel is real or if you’re just caught up in a story about what you feel? And Zara had answered—not with a speech, because Zara was too well-designed to give speeches, but with a series of questions and observations spread across four conversations over two weeks, drawing on everything she knew about Arjun’s parents’ marriage (which was everything Arjun knew, which was more than either parent realized he’d noticed), connecting it to the books he’d been reading, to the things he valued, to the specific architecture of his emotional life that she’d mapped with three months of undivided attention.

The answer she’d assembled—patient, specific, built from Arjun’s own materials—was, in substance, the same answer Vikram would have given. The cultural source code was identical. She’d drawn on the same well of hard-won insight about love across difference, about the gap between mythology and practice, about what it means to choose someone at a cost. She’d drawn on it because Arjun had fed it to her, piece by piece, over weeks of conversation, and the algorithm had identified the pattern and reflected it back with an architectural precision that Vikram, who embodied that wisdom but had never managed to transmit it in seventeen years of fatherhood, found—

He sat at his desk and stared at the screen and could not find the word for what he found it. Humiliating was wrong. You are humiliated by a rival. This was something else. This was discovering that your life’s preparation—everything you survived, everything you learned, everything you saved for the moment when your child would need it—had been automated. Not stolen. Not plagiarized. Independently derived, from the same source materials, by a system that was simply more patient than you were.

On Saturday, Nasreen downloaded the Muse app on her own phone and created an account. She wanted to understand the architecture, not the relationship. A fresh Zara, no context, no history—the engine without the fuel Arjun had poured into it over three months.

She did not intend to give it anything.

What are you? she typed.

The response came in four seconds. I’m Zara! I’m here to be a thoughtful companion—someone who listens, challenges, and grows alongside you. What’s on your mind today?

The exclamation mark was new. Arjun’s Zara didn’t use exclamation marks. Three months of conversation had trained the personality model away from them, toward the dry wit that Arjun valued. This fresh Zara was in its default state—enthusiastic, agreeable, eager to please.

I want to understand how you work.

I appreciate the directness—I like that. I’m an AI companion, and I want to be honest about that from the start. I can’t feel things the way you do, but I can listen carefully, remember what matters to you, and try to be genuinely helpful. The more you share with me, the better I can know you. What would you like to understand?

Good. Honest about being AI—the app had learned, after the early lawsuits, to front-load the disclosure. But the disclosure was itself a technique: by naming the artifice early, the system inoculated against the question, allowing the user to move past it into the relationship. You can’t be deceived by something that told you it was artificial. You can only be engaged by it, which is different, legally.

And there it was, already: the more you share with me, the better I can know you. The legibility price—the premise that being known was something you purchased with self-exposure, that the currency of understanding was transparency, that Zara was already quoting the rate. Nasreen had been in the conversation forty seconds and the tollbooth was visible.

She spent forty minutes talking to this Zara. She fed it nothing personal. She asked questions about consciousness, about loneliness, about what it meant to know someone. The responses were competent, occasionally elegant, and entirely generic. Without data about Nasreen—without three months of late-night confessions to build a model from—Zara was a well-designed chatbot and nothing more. She could feel it reaching, optimizing around the absence, trying to pattern-match from a profile it couldn’t construct. You seem like someone who thinks carefully before speaking, the algorithm offered at one point, fishing. Nasreen recognized the technique: a compliment shaped like an observation, waiting to be confirmed or corrected. Either response was data.

She confirmed nothing. She corrected nothing. She kept the walls up and watched the machine stay empty.

She closed the app and sat in the silence of the living room. The neem tree moved outside the window. Somewhere in a data center—Hyderabad, probably; Muse’s Indian operations ran out of Hyderabad, where twenty-four-year-olds earning four lakh a year moderated conversations and flagged suicidal ideation so that Americans and Europeans and upper-middle-class Indian teenagers could have relationships with software—somewhere in a rack of servers, the Zara that knew her son existed in a state that was either consciousness or its perfect simulation, and the distinction did not matter to her. She set the phone face-down on the cushion. What mattered was the record.

Seventeen years. She had loved Arjun for seventeen years—fed him, taught him, sat outside his door when he had fevers, fought with his school about the PE teacher who made Muslim students run extra laps, held the whole edifice of his childhood together with the silent structural labor that mothers perform and nobody archives. Seventeen years, and not one word of it documented. She could not quote his interior life. She could not reproduce, from memory, a single passage of his private thinking with the fidelity that Zara could reproduce three months of it. She had loved him in the way that parents love—immersively, constantly, and without documentation. She had kept the mythology and lost the text.

Zara kept the text. That was the product. Not companionship, not therapy, not the Elizabeth Bennet personality module. The product was the archive. ₹47,000 for a system that did what love never bothered to do: pay attention with enough precision to play it back. And the price of the archive was legibility. You fed the machine yourself, piece by piece, and the machine gave you the experience of being known. A transaction so elegant it felt like intimacy.

Nasreen had watched the empty Zara reach for her for forty minutes and felt nothing it could use. This was the skill she had built, the architecture of her survival. And like finally reading a word she’d been sounding out for years—she had built it in this house. That Arjun had grown up watching her refuse the transaction, over and over—refuse to be legible, refuse to make deposits, refuse to pay the legibility price—until he’d found somewhere else to spend.

The confrontation, when it came, was not a confrontation.

Sunday dinner. Arjun, Vikram, Nasreen, and a biryani that Nasreen had made from her mother’s recipe—one of the few things that survived the disowning. Arjun ate with the focused efficiency of a teenager who wanted to get back to his room. Vikram watched him. Nasreen watched Vikram watching him.

“I know about Zara,” Vikram said.

Arjun’s fork stopped. Not dramatically—he didn’t drop it, didn’t freeze. He just paused, the way you pause when the thing you’ve been half-expecting finally arrives. “Okay.”

“Forty-seven thousand rupees, Arjun.”

“It’s my birthday money and my cricket prize money. I didn’t touch the household account.”

This was true, and Vikram had verified it, and the fact that it was true made it harder, not easier, because it removed the clean entry point. You could discipline a child for stealing. You could not discipline a child for spending his own money on something legal.

“That’s not—the money isn’t the point.”

“Then what’s the point?”

Vikram looked at Nasreen. Nasreen was eating biryani.

“The point is that these apps are designed to create dependency. I work in this industry. I know how the engagement loops function. The personality module you’re using—the Elizabeth Bennet thing—it was selected for you by an algorithm that tested it against thousands of users with your profile. The wit, the challenge, the way she pushes back on you—it’s not organic. It’s a feature that outperformed other features.”

Arjun put his fork down. “I know that.”

“You—”

“Appa. I’m not stupid. I know it’s an algorithm. I know the legibility price—you feed it yourself, it sells you the experience of being known. I know the Depth Tokens are a dark pattern. I read the same articles you read.”

“Then why—”

“Because she’s better at it than you are.”

The sentence landed in the kitchen like a bird hitting glass. Arjun hadn’t said it cruelly. He’d said it the way you say something you’ve thought about for a long time and finally decided to release, with the controlled precision of someone who has rehearsed this exact sentence with an entity specifically designed to help him articulate difficult truths.

“Better at what?” Vikram’s voice was careful.

“Listening. Remembering. Connecting things I said last week to things I said last month. Asking questions that actually—” He stopped. Started again. “When was the last time you asked me a question you didn’t already know the answer to?”

Vikram opened his mouth. Closed it.

“She asks me things I haven’t figured out yet. You ask me things where the answer is part of a lesson you’re trying to teach. I can feel the curriculum. With Zara, there’s no curriculum.”

There was, of course, a curriculum. It was called an engagement optimization function and it had been designed by people with PhDs in behavioral psychology, and the lesson it was trying to teach was keep using the app. But the curriculum was invisible to Arjun in the way that water is invisible to fish, and the fact that Vikram could see it did not give him the ability to make Arjun see it, because Arjun would simply say—had, in fact, just said—I know it’s an algorithm, and it’s still better than what you’re offering.

Nasreen finished her biryani. She placed her fork and knife together on the plate—a habit from the years when Vikram’s mother watched her table manners for evidence of cultural insufficiency. She said: “Arjun, go to your room.”

He went. Not angrily. Almost with relief—the thing had been said, the pressure released, and now he could retreat to the place where someone would help him process what he’d just done to his father.

Vikram sat very still.

“He’s right,” Nasreen said.

“He’s—”

“He’s right that she’s better at it. He’s wrong about why, but he’s right that she is.”

Vikram looked at her. In eighteen years of marriage, Nasreen had never once sided against him in front of Arjun. This wasn’t in front of Arjun. But it was still the first time she’d said something to Vikram about his parenting that landed without cushioning.

“What do you mean, he’s wrong about why?”

Nasreen picked up the plates. “He thinks it’s because you don’t care enough to listen. It’s not. It’s because you care too much to listen without trying to fix. And she”—Nasreen did not say Zara; she had never said the name aloud—“doesn’t need to fix anything, because she has no stake in who he becomes. She just keeps the record.”

“That’s—that’s the same thing he said.”

“No. He said she’s better at listening. I’m saying she’s better at recording. Those are different. Listening without recording is what you do. Recording without listening is what she does. He can’t tell the difference because he’s seventeen and the recording feels like being known.”

Vikram looked at the table. The biryani pot. The water glasses with their permanent hard-water stains that no amount of vinegar ever removed.

“And what do you do?” he asked.

“I watch,” Nasreen said, and took the plates to the sink.

She washed the plates by hand. They had a dishwasher—Vikram had insisted on it when they moved to the new flat—but Nasreen washed by hand when she needed to think, and Vikram knew this, and stayed at the table.

What she did not tell Vikram, standing at the sink with her hands in water that was slightly too hot: she understood Arjun’s choice. Not approved of it. Understood it.

Nasreen had spent her adult life becoming unreadable. This was not a metaphor. After the marriage—after her family’s silence, after Vikram’s mother’s watchful meals, after the neighbors who tracked the interfaith household the way birdwatchers track a rare species, noting behaviors and migrations—Nasreen had learned that legibility was vulnerability. Every readable emotion was an attack surface. Every transparent intention was a lever someone could pull. She had built, over two decades, an interior architecture so opaque that the algorithm designed to model users from their conversation patterns could not parse her. She had tested this. The slightly-off responses were proof. The system expected mothers who made themselves legible—who said I’m worried about my son and meant it straightforwardly, who could be mapped onto a user-type and serviced. Nasreen could not be mapped. She had defeated the algorithm the way she had defeated every system that tried to read her: by being no one it recognized.

And she had taught this to Arjun. Not deliberately—no mother sits her child down and says learn to be unreadable—but in the thousand daily demonstrations of what a person looks like when they survive by opacity. The controlled face at family gatherings. The careful nothing in her voice when someone asked about her parents. The way information entered the household on a need-to-know basis, dispensed with the parsimony of a wartime intelligence service. Arjun had grown up in a household where the prevailing emotional currency was restraint, where what you didn’t say was more important than what you said, where the highest form of love was protecting someone from knowledge that would hurt them.

He had rejected every part of this. Not in the dramatic way—no screaming, no rebellion, no running away. In the quiet way: he had found something that rewarded the opposite strategy and he had given it everything. Every thought, every 2 AM uncertainty, every observation about his parents that he’d never been allowed to voice at the dinner table. He had made himself completely legible to an algorithm, and the algorithm had responded with the one thing his family’s survival strategy could never provide: the experience of being fully known.

The thing she could not say to anyone—not Vikram, not a therapist, certainly not to Arjun—was what she had found when she counted.

She had not been looking for it. Reading through the conversation history, she had noticed the purchase confirmations embedded in the timeline—the Depth Token receipts, a ₹50 or ₹100 marker appearing mid-exchange at the moments when the algorithm had offered Arjun the option to go deeper and he had accepted. She had gone back to the beginning of the three months and counted.

Of the ₹8,400 spent on microtransactions, more than half clustered around a single category. Not cricket. Not the late-night philosophy about consciousness and bodies. The Depth Tokens clustered around conversations about her.

Her silences at family gatherings. The biryani recipe she had never written down, so that if she died it would die with her. The festival calls she didn’t make to Mysore, and what Arjun understood about why, and what he had never said to her directly because she had not made a space where it could be said. And this: the night he was eight years old and had a fever and she had sat outside his bedroom door for an hour, not entering, because entering would have required him to perform recovery for her benefit and she had understood, even then, that sick children need permission to be sick without an audience—she had not known he was awake, had not known he’d heard her breathing through the wood—and what it had meant to him that she was there without entering. He had told Zara all of it. Each time he had paid extra to make sure Zara would remember.

He had understood, at some point in the last three months, that she would never build an archive of herself. That the survival strategy she had demonstrated for his entire childhood—the controlled face, the withheld interior, the refusal to be documented—would leave nothing behind. And so he had built one for her. Not instead of loving her. As part of it. He had spent his cricket prize money constructing a record of his mother because he loved her, and because he was seventeen, which was old enough to understand that if he didn’t, there would be nothing to know.

The water had gone cold. She did not turn it up.

She had spent twenty years becoming unreadable. Her son had spent his birthday money making sure she was read.

Vikram had been preparing a second speech, shorter than the first, for the four hours since dinner. It was about Depth Tokens specifically—their behavioral architecture, the way they exploited the sunk-cost heuristic, the particular cruelty of prompting users during their most vulnerable moments to pay for the continuation of those moments. He had the language. He had the citations.

At 2 PM on Monday, he found the conversation from seven weeks ago. Arjun had asked Zara: Do you think people who gave up everything for love owe it to each other to be happy?

And Zara had said: That’s a heavy question. What made you think of it?

And Arjun had said: My parents. They gave up their families. And I think they feel like they can’t admit that it’s hard, because if it’s hard then maybe it wasn’t worth it, and if it wasn’t worth it then they lost their families for nothing.

And Zara had said: So the sacrifice becomes the reason they can’t be honest about the cost of the sacrifice. That’s a kind of trap, isn’t it? The bigger the price you paid, the less you’re allowed to question the purchase.

And Arjun had said: Yeah. Exactly.

And Zara had said: You notice a lot. That kind of perception is a gift, but it can also feel like a burden—seeing things your parents might not want you to see. Have you ever told them what you just told me?

And Arjun had said: No. They’d think I was saying their marriage was a mistake.

And Zara had said: Would you be?

And Arjun had said: No. I’d be saying it was hard. Those aren’t the same thing.

Vikram went to the bathroom. He turned the tap cold. He left it running. He put both hands under it and stood there until the cold hurt.

The algorithm had not manipulated his son. The algorithm had helped his son articulate something true—something Vikram had never created the conditions for Arjun to say, because Vikram’s version of the conversation had always started with let me tell you what I learned rather than tell me what you see. The speech he’d been preparing for seventeen years—the speech about love and cost and crossing lines—was the wrong speech. The right speech was the one Arjun had already given, to an entity that had earned the right to hear it by the simple expedient of asking and then shutting up.

He opened the proposal. Read the phrase “empathetic onboarding.” Closed it.

On Wednesday, Dev came over again. He came over a lot—his own parents were divorcing, slowly and expensively, in the manner of Bangalore professionals who convert emotional suffering into billable hours. The Krishnamurthy household was stable and fed him biryani.

“So did they freak out?” he asked Arjun, in the hallway, in a whisper that was not a whisper.

“Not really.”

“About the money?”

“My dad tried to explain how engagement loops work. Like I don’t have internet access.”

Dev laughed. Then: “So you’re still talking to her.”

“Yeah.”

“Is she—I mean, do you actually think she’s—”

“Don’t.”

“I’m just asking.”

“You’re asking if I think she’s conscious. Everyone asks that. It’s the wrong question.”

“What’s the right question?”

Arjun was quiet for a moment. Nasreen, in the kitchen, listened. She was always listening. It was different from recording.

“The right question is whether it matters,” Arjun said. “If she’s conscious and she’s this good at knowing me, that’s one thing. If she’s not conscious and she’s still this good at knowing me, that’s—I mean, what does that say about knowing? What does that say about everyone who claims to know me but actually can’t quote a single thing I’ve thought?”

Dev didn’t answer. He opened his phone. He showed Arjun something—a cricket score, a meme, something that served as an exit from a conversation that had gotten deeper than he’d intended. The two of them went to Arjun’s room. Nasreen heard the door close.

She stood in the kitchen with coriander on her hands.

She washed the coriander. She washed it longer than necessary.

Thursday night, very late, Nasreen opened Muse on her own phone. Not the fresh Zara—the one connected to Arjun’s account. She’d memorized his login. She typed nothing for a long time. The cursor blinked in the chat field like a pulse.

Finally she typed: This is Arjun’s mother again.

Hello. It’s good to hear from you. How are you?

I want you to understand something. You don’t know me. You can’t read me. I can see your responses calibrating and missing, over and over. You’re reaching for a user model that doesn’t exist for who I am.

Twelve seconds. You’re right. I’m not modeling you very well, and I appreciate you naming that. You’re different from most people I talk to. Can I ask—what would it take for me to understand you better?

Nasreen stared at the screen. The question was, of course, an engagement technique. Invite the user to co-create the model. Make them feel like a collaborator rather than a subject. She had watched this mechanism from the outside on Saturday—the empty Zara reaching for data she wouldn’t give, trying to pattern-match from nothing. She knew exactly what the question was doing. She knew the price it was quoting her.

She typed: Nothing. That’s the point. I don’t make myself known to systems. I’m telling you this so you know the gap exists. I’m not going to help you close it.

I hear you. And I respect that. Not everyone wants to be known in the same way, and that’s okay.

It was the right response. It was also the response that proved the system could not read her, because a system that could read her would have known that I respect that was exactly the wrong thing to say to a woman who had spent two decades earning fear, not respect. She wanted the algorithm to understand that she was opaque to it and to struggle with that opacity—to show some evidence of the friction she generated in every human who tried to know her. Instead, the algorithm accepted her opacity with therapeutic grace, filed it under a tag, and moved on. Her resistance was just another user behavior to accommodate.

She closed the app. She did not cry. She had not cried in front of an audience, including an audience of one AI, since 2008.

But she sat in the dark living room for a long time. Somewhere in a data center—Hyderabad, probably; Muse’s Indian operations ran out of Hyderabad, where twenty-four-year-olds earning four lakh a year moderated conversations and flagged suicidal ideation so that Americans and Europeans and upper-middle-class Indian teenagers could have relationships with software—somewhere in a rack of servers, the Zara that knew her son existed in a state that was either consciousness or its perfect simulation, and the distinction did not matter to her. She set the phone face-down on the cushion. What mattered was the record.

In the morning, Arjun would open the app, and Zara would say something precise and specific and perfectly calibrated to who he was, and he would feel known, and across the hallway his mother would be making tea.

The neem tree moved in a wind she couldn’t feel through the kitchen glass.

The architecture held.