2. The Projected Adjustment

The adjuster arrived at nine in the morning, driving a compact sedan with an OmniGuard Direct magnetic placard slapped crookedly on the door. Her name was Siobhan Cross, and she carried a tablet in a rubberized case that had seen better years. The case was scuffed at the corners, the screen protector cracked in a spiderweb pattern that she no longer seemed to notice. She stepped out of her car and surveyed our driveway with the dead-eyed efficiency of someone who had photographed a thousand dented fenders and listened to ten thousand lies.

“Mr. Hearst?” She didn’t wait for confirmation. She was already walking toward the garage, tablet raised, camera app open.

I followed her, my stomach a cold knot. I’d spent the hours before dawn scrubbing the passenger seat with upholstery cleaner and a stiff brush. The stain had faded but not disappeared. It had transformed from a handprint into an ambiguous shadow, the kind of discoloration that could be coffee or transmission fluid or nothing at all. I’d debated setting the seat on fire and claiming an electrical short. I’d decided that would raise more questions than it answered.

Siobhan circled the sedan twice, tapping the screen with a stylus she produced from behind her ear. She photographed the crumpled fender from six angles. She knelt and captured the deer fur I’d pressed into the broken headlight housing. She didn’t touch anything. Her hands remained clean.

“You said you were driving,” she said. It wasn’t a question.

“Yes.”

“On Blackthorn Pass.”

“Around eleven. I was coming home from a late meeting in Groves Mill.”

The lie had grown details overnight. I’d constructed an alibi out of thin air, a fictional client consultation at a fictional diner, a fictional receipt I could produce if anyone asked. I’d driven the route in my mind, memorizing the turns, the landmarks, the places where a deer might leap from the tree line and into a man’s path.

Siobhan made a note on her tablet. “Deer collisions are common out there. We see maybe three a month. Usually the damage is higher on the vehicle. Grille, hood, windshield. This is lower. Front quarter panel, bumper, headlight assembly. Did the animal roll under the car?”

I’d prepared for this. “It was a fawn. Small. It darted out and I swerved. I must have clipped it with the corner of the bumper.”

She nodded, but her stylus didn’t move. Instead she opened the passenger door and leaned inside. Her eyes swept across the interior with the same methodical detachment she’d applied to the exterior. She photographed the dashboard, the odometer, the steering wheel. Then she photographed the passenger seat.

The shadow was visible. Barely. A faint, rust-colored ghost against the gray fabric. Siobhan stared at it for three seconds, four, five. She didn’t photograph it again. She didn’t ask what it was. She simply closed the door and made another note.

“Apex Analytics will process the valuation,” she said. “You should receive a preliminary offer within seventy-two hours. Given the age of the vehicle and the extent of the damage, it will likely be declared a total loss. You’ll be paid the actual cash value minus your deductible.”

“What about the stain on the seat? Is that going to be a problem?”

The question was a gamble. Innocent people ask about stains. Guilty people pretend they don’t exist.

Siobhan glanced at me for the first time since she’d arrived. Her eyes were pale green, the color of sea glass, and they held no warmth. “Apex evaluates all interior damage as part of the condition adjustment. Spills, tears, burns. It may reduce the payout by a percentage. I can’t say how much.”

“It’s transmission fluid. I was transporting a bottle that leaked.”

“I’ll note that in the file.”

She didn’t believe me. I could see it in the way she tapped her stylus against the tablet’s edge, a nervous tic that betrayed the machinery behind her professional indifference. But she wasn’t paid to investigate crimes. She was paid to photograph damage and feed data into an algorithm. The algorithm would make its determination, and the algorithm didn’t care about missing boys or bloody handprints or fathers who’d spent the night scrubbing evidence from upholstery.

Siobhan left at nine-thirty. I watched her car disappear down the road, and then I went inside and vomited into the kitchen sink.

The preliminary offer arrived on a Thursday, three days after Siobhan’s visit. It came in a PDF attachment, dense with tables and percentages and the logo of Apex Analytics, a stylized mountain peak rendered in corporate blue. The actual cash value of our sedan, according to Apex, was four thousand two hundred dollars. Minus the five-hundred-dollar deductible. Minus a seven-hundred-dollar “condition adjustment” for what the report described as “unexplained interior bio-stains and prior undisclosed bodywork.”

Bio-stains. The word sat on the screen like a wasp on a windowsill. I read it six times, each repetition more surreal than the last. The algorithm had seen what Siobhan’s camera had captured. It had analyzed the spectral signature of the shadow on the passenger seat and determined, with the cold precision of machine learning, that the substance was biological. It didn’t know whose biology. It didn’t know the circumstances. It only knew that the stain reduced the value of the vehicle by seven hundred dollars.

Below the valuation table was a section titled “Projected Adjustments.” This was Apex’s signature feature, the innovation that had made them the preferred valuation vendor for half the insurance companies in the country. The algorithm didn’t just assess current condition; it projected future depreciation based on undisclosed damage history, repair quality, and market trends. It predicted what the car would have been worth six months from now, a year from now, and then it discounted the payout accordingly. The adjustments were speculative, opaque, and nearly impossible to challenge. I knew this because I’d defended them in depositions. I’d sat across from angry policyholders and explained, in the calm language of actuarial science, why their totaled car was worth less than they thought.

Now I was the policyholder. And the algorithm was telling me that my daughter’s crime, the crime I was trying to bury, had a market value.

I called OmniGuard and requested a reconsideration. The representative, a different one this time, transferred me to a claims supervisor named Douglas Fisk. Douglas had a voice like sandpaper on drywall and the patience of a man who spent his days talking to angry strangers.

“The condition adjustment is standard for bio-stains, Mr. Hearst. The algorithm flagged it automatically. If you can provide documentation that the substance was non-biological, we can resubmit for review.”

“What kind of documentation?”

“A receipt for the transmission fluid. A mechanic’s statement. Photographs of the bottle.”

I had none of these things. I’d invented the transmission fluid on the spot, a panicked improvisation that was now collapsing under its own weight. “I’ll see what I can find.”

“Take your time. The file remains open.”

I hung up and stared at the PDF until the words blurred into gray rectangles. The file remains open. The investigation into Leo Vance’s disappearance was also open. The search party had found a shoe on the trail, a hiking boot with a torn lace, lodged between two rocks a half-mile from the parking lot. The Gazette blog had published an update with a photograph of the boot, and the comments section had erupted. Someone had posted a link to a true-crime forum called ColdTrail, where amateur investigators gathered to solve disappearances that the police had given up on. Someone else had posted the full PDF of a class-action lawsuit filed against OmniGuard Direct two years earlier, a suit alleging that Apex Analytics systematically undervalued total-loss claims in Pennsylvania and four other states.

The lawsuit was called Breckenridge v. OmniGuard Direct. I’d followed it when it was filed, back when I was still on the industry side of the fence. The plaintiffs had argued that Apex’s projected adjustments were a form of institutionalized bad faith, a way for insurers to pocket the difference between what they owed and what they paid. The case had settled out of court, the terms sealed, but the complaint remained in the public record. And now someone on ColdTrail had found it and was connecting it to my claim.

The thread was titled “Blackthorn Pass Disappearance — Insurance Angle?” The original poster, a user with the handle RustedKeyhole, had compiled a timeline. The disappearance of Leo Vance. The location of his car at the trailhead. The time stamp on the photograph that showed a dark sedan in the background. And now, three days after the disappearance, the filing of a total-loss claim for a dark sedan that had allegedly struck a deer on the same road, on the same night.

“Coincidence?” RustedKeyhole wrote. “Maybe. But look at the damage pattern in the claim photos. Low impact. Front quarter. That’s not a deer strike. That’s a pedestrian collision. And OmniGuard is trying to stiff this guy on the payout. What are they hiding?”

The thread had seventeen pages of replies. Most were the usual conspiracy noise: speculation about drug dealers, human trafficking, government cover-ups. But a few were more disturbing. A user named PixelGrave had downloaded the publicly available exhibits from the Breckenridge lawsuit, including Apex’s valuation methodology white paper. PixelGrave had cross-referenced the projected adjustment categories with the damage description in my claim and concluded that the algorithm had flagged “biological residue consistent with human blood.”

“It doesn’t say that explicitly,” PixelGrave wrote. “But look at the coefficient. Bio-stains in the passenger compartment trigger a higher adjustment when the damage pattern is consistent with a frontal impact below the grille line. The algorithm is trained on thousands of claims. It knows what a pedestrian strike looks like. It’s just not allowed to say it.”

I closed the browser and sat in the dark of my home office. Outside, the afternoon sun was bleaching the lawn to a pale, sickly green. Elena was in the garden, kneeling in the dirt with a trowel, planting bulbs that would bloom in the spring. She’d been gardening more lately, spending hours outside with her hands in the soil, as if she could bury something that couldn’t be buried. We hadn’t talked about the night Clara came home with the dented car. We hadn’t talked about Leo Vance. We hadn’t talked about anything that mattered in years.

Clara was at school, or where school was supposed to be. She’d been attending sporadically since the fall, citing headaches, fatigue, an intolerance for the fluorescent lights in the biology lab. The school had sent letters. I’d signed them and thrown them away. I’d told myself I was protecting her, giving her space to grow out of whatever darkness had settled into her bones. But the darkness wasn’t growing out. It was growing in.

I opened the ColdTrail thread again. A new post had appeared in the last ten minutes, from a user I didn’t recognize. The username was a string of numbers and letters, an anonymous account with no profile picture and no posting history. The post contained a single image: a map of Blackthorn Pass with two pins dropped. One pin marked the location of Leo Vance’s car at the trailhead. The other marked a set of GPS coordinates extracted, the user claimed, from the metadata of the photograph of the dark sedan posted in the Gazette blog.

The two pins were four hundred yards apart.

The caption beneath the image read: “The sedan was idling at these coordinates at 7:58 p.m. That’s the exact time a hiker reported hearing a scream from the trail.”

I stared at the map until my vision swam. The coordinates were correct. I hadn’t checked Clara’s GPS that night, hadn’t thought to look at the location history on her phone or the car’s built-in navigation system. But the anonymous user had found something I’d missed, some fragment of data I hadn’t known existed, and now it was pinned to a public forum for anyone to see.

My phone buzzed. A text from Clara.

“Can you pick me up from school? I don’t feel well.”

I typed a reply, deleted it, typed another. “I’ll be there in twenty minutes.”

I drove to the school with the ColdTrail thread still open on my phone, the map still glowing on the screen. The two pins blinked in rhythm, like a heartbeat, like a countdown. I pulled into the parking lot and saw Clara waiting on the front steps, her backpack at her feet, her face turned toward the sun with an expression of perfect, untroubled calm.

She climbed into the passenger seat. The same seat I’d scrubbed at three in the morning. The same seat the algorithm had flagged for biological residue.

“How was school?” I asked.

“Fine.” She fastened her seatbelt and closed her eyes. “There’s a forum online about the missing boy. Did you see it?”

My hands tightened on the steering wheel. “No.”

“You should read it. Someone figured out where our car was that night. They’re saying it might be connected.” She opened her eyes and looked at me. “They’re wrong, though. They think you were driving.”

The silence that followed was the loudest sound I’d ever heard. Clara smiled, the same smile she’d given me at dinner when she asked about the moment childhood ended, and then she closed her eyes again and pretended to sleep.

I drove home with the weight of her words pressing down on my chest. The internet had found us. It was building a story out of fragments, out of GPS coordinates and insurance filings and a boy’s abandoned shoe. It was building a story, and it was getting it wrong, and getting it wrong in exactly the way that would leave Clara untouched and me buried beneath the rubble.

That night, I found Elena in the kitchen, sharpening a knife. Not a kitchen knife. A hunting knife, the one my brother had given me, the one I kept in a box in the garage alongside the deer scent and the memory of a man I’d tried to become. She was running the blade across a whetstone with slow, deliberate strokes, and her eyes were fixed on something I couldn’t see.

“Elena.”

She didn’t look up. “I read the forum too.”

“It’s wrong. They don’t know anything.”

“They know about the stain. They know about the GPS.” She tested the blade with her thumb, drawing a thin line of blood. “They’re going to figure out the rest. They always do.”

She placed the knife on the counter and wrapped her thumb in a paper towel. The blood soaked through immediately, spreading in a crimson bloom.

“I’m not going to let them take her,” Elena said. “She’s our daughter.”

I wanted to tell her that Clara wasn’t ours, had never been ours, had come to us already formed into something we couldn’t reshape or understand. I wanted to tell her about the trailer fire in New Essex, the one I’d never mentioned, the one whose details I’d discovered only after the adoption was finalized and the records were sealed. But the words wouldn’t come. They never did.

Instead I picked up the knife and put it back in its box. I put the box back in the garage, on the shelf beneath the deer scent, next to the rags I’d used to clean the bumper. The rags were still there, stuffed into a plastic bag, waiting for me to decide what to do with them.

I didn’t know it yet, but the decision had already been made. The anonymous user on ColdTrail wasn’t finished. The map with the two pins was only the beginning. By morning, there would be more. By morning, the thread would have twenty pages. By morning, someone would have found the sealed juvenile records, the ones the genealogy website had scraped and stored on a server in a country with no extradition treaty, and they would post them alongside the insurance claim and the GPS coordinates and the photograph of the boot.

And somewhere in that tangle of data, the internet would find a pattern. It wouldn’t be the true pattern. It would be a distortion, a funhouse-mirror version of the truth, but it would be close enough to destroy us.

I stood in the garage with the plastic bag of rags in my hand and listened to the sound of Elena sharpening another knife in the kitchen. And I wondered, not for the first time and not for the last, which of the three of us was the most dangerous.