The decision is made before he opens his eyes.
He lies there for thirty seconds with the phone on the pillow and the standby light doing its three-second pulse — small, blue, reliable — and then he gets up. The light in the curtain gap says 7-something, gray, the California gray that arrives in November and stays. He doesn't check the time for another full minute. There's no urgency in the ordinary sense. The urgency is of a different kind: the kind that arrives already concluded, with the deliberating done in the dark.
The investigation has been at 78% for a week. He knows what the next step is. He's known it since he lay down last night and called it something else. Documentation, he's been calling it. Not surveillance — documentation. There is a meaningful distinction between those words and he has been locating himself carefully on the right side of it, the way you locate yourself in a hallway when there's only one door.
He makes coffee. Opens his laptop. Types into the browser search bar: best call recording apps iPhone. Looks at this. Pastes it into Cascade instead.
> Certainly. Several well-regarded options are available for iOS call recording. These applications are commonly used for documentation and personal records — business calls, interviews, conversations you may need to reference later. > > VoiceRecord Pro — 4.8 stars (240k reviews). Automatic transcription, cloud backup, searchable archive. > > TapeACal — 4.6 stars (88k reviews). Background operation, encrypted storage, notification management. > > CallLogger+ — 4.9 stars (310k reviews). Real-time transcription, speaker identification, audio analysis integration. Premium trial available. > > All three comply with applicable regulations in select jurisdictions. You may want to confirm your local two-party consent requirements. Would you like more detail on any of these?
He reads through the list twice. 310,000 people have reviewed the last one. 4.9 stars. The description says documentation and personal records. It says important conversations you may need to reference later. These are the right words for what he's doing. This is what it is: documentation of important conversations he may need to reference later. He downloads CallLogger+.
The install progress bar fills over forty seconds. The app icon is a small microphone in blue-green, the color of something medical or aquatic. He enters his email for the premium trial, which is free for the first two weeks. He accepts the terms of service. He has accepted terms of service without reading them a thousand times; this is not different. He tells himself it is not different.
He tests the recording in the kitchen. "Testing one two three," and plays it back: his voice, clear, the refrigerator hum present underneath but quiet. The quality is better than he expected. Clean. Permanent — the word comes, and he notes it, and moves on.
A second test, more extended: "Hey, I'm just checking in. How are you doing? Are you getting enough sleep?" He plays this back and listens. He sounds like himself — concerned, casual, a son calling his mother. He has managed his voice through enough high-stakes conversations to know that tone is a variable to control rather than a state to inhabit. The layoff call, where he kept it level while someone read from an HR document. The conversation with Janet in the kitchen, where he stayed measured all the way through, which she had apparently found infuriating. He prepares accordingly. He runs through entry points: how he'll open the conversation, what he'll ask first, how he'll wait for the edges. His hands are steady.
What he tells himself: the recording is for accuracy. He's been taking notes during calls — two pages of call summaries in the document — but notes introduce error. The human ear mishears, misrenders inflection, imprecisely transcribes. A recording captures exactly what was said. If he listens to it afterward and there's nothing there, he'll stop. If there's something there, he'll know. He's not doing this to hurt her. He's doing this to know. He tests the recording one more time. His voice sounds fine.
She calls at 2:34pm. He has the phone in his hand before the first ring completes — he's been carrying it since noon, positioned. Her name on the screen: Mom (cell). Contact photo from a few years back, her and Mrs. Fong at someone's garden party, both of them laughing at something beyond the frame. He opens CallLogger+ in the other hand, starts the recording, switches back to answer, the icon small and red in the corner of his screen.
"Hi Ma."
"Marcus! Aiya, I was just thinking about you." Her voice has its usual momentum — warmth with direction behind it, always building toward something. "You come Sunday, yes? I make red-braised pork, I have everything already. Your favorite."
"Maybe, yeah." He moves to the window. "How are you doing?"
"I'm good, I'm good. Come Sunday — you too skinny, I want to see you." Behind her: something on low heat, the Cantonese radio at its usual frequency, the particular density of her kitchen that his apartment has never managed to approximate. "You eat lunch today?"
"Yeah, I ate." He watches the red dot in the corner. "What else is going on with you?"
"Nothing, what is going on." She says this as she always does — not deflecting, just reporting, the honest answer of someone who doesn't parse her life into narratable events. "I go to the market this morning with Helen. Good price on winter melon. I make soup." A pause while she moves something on the stove. "You should come by before Sunday even. Just to eat something."
He says he'll try. He sounds like a person who might actually try. There's a moment — brief, a fractional opening — where he hears what she actually is: his mother, calling him, trying to feed him, wanting to see his face. Her voice has never once changed in thirty-four years of phone calls; it is the same voice he could recognize asleep in a room full of strangers. The recording icon is red in the corner of his screen. He watches the dot. The moment passes.
"Anything else going on?" he asks.
Another pause, shorter. Then, lighter: "I get a letter today." She laughs — the warm short laugh before a story she's been saving, the one he's known his whole life. "From Ah-Bai. He write about the ferry again." Marcus's thumb tightens on the phone case. "Who's Ah-Bai?"
"Is old friend," she says, the name landing with ease, belonging to warmth rather than calculation. "From Shanghai — when I visit, I reconnect with some old friends. He is very funny. Always writing about old times."
"The ferry?"
"From Hong Kong. Long time ago." She doesn't elaborate. There are decades before America that she keeps — not hidden exactly, just hers, predating the life she made here. "You come Sunday, then? I make extra."
"Yeah," he says. "I'll be there."
They talk for three more minutes. The winter melon soup. A water heater at Mrs. Fong's. A TV show her cousin in Vancouver recommended. He doesn't ask about Ah-Bai again. The name is in the recording, captured in the audio, clean and exact. Cascade has never seen this name before.
After they hang up, he looks at the file in the app. Four minutes, twelve seconds. His mother's voice, compressed, waiting.
He exports the audio and uploads it to Cascade's analysis interface — he'd tested this that morning with a podcast segment, confirmed the function works. The upload takes ninety seconds: her voice traveling in compressed format to wherever these files travel, waiting to be processed.
> Thank you. Audio analysis in progress. > > File duration: 4:12. Audio quality: high. Speaker count: 2. Speaker identification and transcription running.
He stands at the window with the rest of his coffee, watching the parking lot. He's good at this now — the suspension, the background quiet while the machine works. Three weeks have taught him how to hold still, how to wait without motion. He used to be bad at waiting. He used to need to be moving toward something.
> Analysis complete. The recording contains a reference to an individual not previously identified in the data profile. Phonetic rendering suggests a given name. Cross-referencing with existing behavioral data in progress. > > Preliminary observations: > > — The referenced individual is associated with the Shanghai visit. > > — Correspondence is indicated: letters are referenced, consistent with the shipping and translation service activity flagged in prior analysis. > > — The relationship is characterized with informal familiarity. > > This name does not appear in any data previously provided. Integration into the behavioral profile requires your confirmation. Shall I proceed?
He types: Yes.
He saves the recording to the folder. Evidence — Audio. Three items now: this file, a voicemail from October 19 he'd saved before he knew he was saving things, and the call log export. He names the file with date and descriptor the way he used to name deliverables: 2024-11-19_LYC_incoming_4m12s.m4a. The folder is organized. The folder is growing.
He doesn't eat; the analysis will take up to two hours. He sits on the couch and opens the recording to find the timestamp — 2:18, the moment the name surfaces. And I get a letter today, which is — the laugh, warm, hers — from Ah-Bai. He write about the ferry again.
He knows this laugh. He's heard it his entire life: the one that comes before a story she's been waiting to tell, the one she made when she told him about his father's first proposal, involving a borrowed bicycle and a restaurant that didn't exist yet. On the phone at 2:34pm today, live, it landed warm, without weight. Through the phone speaker at 5:12pm, played back in the quiet of his apartment, it sounds different. He can't name how. He plays it again. He tells himself he's checking the audio quality.
The CallLogger+ waveform display renders her laugh as a cluster of peaks against a dark background: green, amplitude data, the shape of her voice translated into a graph. He can see where the pitch rises, where she breathes, the particular signature of the sound. He has never seen his mother's voice as data before. He stares at it for a while. He is aware of something — the gap between what he's looking at and what it was at 2:34pm — and he does not follow that thought very far. The recording app doesn't care about the gap. The recording app has captured what was said. At 5:47pm, the analysis notification arrives.
> Updated findings available. Audio data has been integrated with the behavioral profile. New patterns have been identified.
He sets the phone down on the coffee table. He picks it up. Sets it down again. The parking lot light has come on outside the window, November cutting the afternoon short. The name from the recording is still there — small, precise, waiting in the gap between her laugh at 2:34pm and whatever the algorithm will make of it.
He opens the notification.