April Snow Chapter 08
Posted on Wed 08 April 2026 in Dead Signal
Chapter 8: Chase Your Tail
April Snow — Book 1 of Dead Signal
The testimonial was from the CISO — Chief Information Security Officer, the executive who owned the security budget and bore professional responsibility when things went wrong — of a financial services firm whose name wasn't given. Standard anonymization — security clients didn't put their names on vendor endorsements that implied they'd had vulnerabilities worth endorsing. Perfectly normal. Will read normal testimonials all the time.
"ChasePath identified behavioral patterns consistent with insider threat activity 73 days prior to our incident response being activated. We estimate it saved us four months of exposure."
He read it twice. Then he ran a search.
Significant insider threat incidents at financial services companies, publicly disclosed, last four years. Seven results. He cross-referenced them against the approximate employee size and timeline implied in the testimonial — mid-sized firm, an incident that surfaced to public consequences in the second half of 2023. Two possible matches. He pulled both.
Neither had publicly disclosed an insider threat. Both had disclosed consequences: a consent decree in one case, a regulatory fine about data handling in the other. Nothing in the public record about insider activity being the cause of either. The problems had just appeared, apparently from nowhere, and the companies had addressed them with the specific corporate language of organizations that had been told what to say by their legal teams.
CYT's testimonial referenced the insider threat anyway.
The client had signed off on the testimonial. Either they were fine with CYT describing a problem they'd never publicly acknowledged — which required explaining why — or they hadn't fully read what they'd signed off on, which required a different kind of explaining.
He had been building the folder since before coffee, which meant since 5:50 AM, which meant since before he'd been honest with himself about whether he'd slept or only lain in the dark for a while doing a reasonable impression of it.
CYT Systems. Chase Your Tail. Security intelligence that sees what's coming. Before it bites you in the ass?
The website had taken him four minutes to read and another twenty to look at properly. Premium build — the kind that cost thirty thousand dollars and was designed to communicate institutional seriousness through font choices and negative space and a hero image of a city at night threaded with signal-stream overlays in cool blue and white. The product was called ChasePath. Behavioral anomaly detection. Continuous threat monitoring. Predictions that stay accurate. The testimonials were from unnamed executives at unnamed companies in financial services and healthcare and critical infrastructure. He had read the whole site once for the story it was telling, and then gone back for the gaps.
The gaps were where the work was.
LinkedIn: forty-seven employees. He sorted by tenure. Eleven had been there more than four years. The rest had joined in the last twenty-four months, which told him the company had either grown recently or restructured or had an operational change that required new headcount. He opened all eleven long-tenure profiles and started mapping the internal org.
Three were expected: network engineers, a CISSP — Certified Information Systems Security Professional, the senior-tier practitioner credential — a former incident responder from a managed security service provider. Standard bench for an enterprise security platform. He noted them and moved on.
The other eight were not standard.
Four data scientists. Two machine learning engineers. A researcher with a Vanderbilt PhD in behavioral prediction whose dissertation title, still visible on a university research profile, was Longitudinal Behavioral Pattern Stability in High-Stress Environments — currently listed at CYT as Head of Behavioral Intelligence. A data architect whose prior role, before CYT, had been building longitudinal behavioral data pipelines for an insurance actuarial analytics company.
He paused on the data architect.
The actuarial background was specific. Actuaries built models to predict when people would die, become disabled, make insurance claims. Their models were trained on large longitudinal datasets — the same population tracked across time — precisely because behavioral patterns shifted with circumstance, and you needed to know both the pattern and its trajectory to make an accurate prediction. The data architect had spent six years building that kind of infrastructure before joining CYT.
What was a company that built enterprise security threat detection doing with a data architect who understood longitudinal behavioral tracking across years of data from the same subjects?
Unless the subjects weren't changing because they couldn't.
He pulled up a new window and started on the job postings.
Archived job listings from a paid aggregator, Wayback Machine, Google cache. He was looking for roles that didn't fit the surface story — the kind of position a company only needed if it was doing something it hadn't described publicly.
2023: Senior Behavioral Data Engineer. Responsibilities include processing and quality assurance for proprietary behavioral training datasets.
No description of the dataset. No indication of its source.
2022: Data Annotation Specialist. Experience with behavioral signal classification required. Ability to work with non-standard data formats preferred.
He wrote down non-standard data formats and stared at it for a moment.
Data annotation specialists labeled training data for machine learning models. You hired data annotation specialists when you had a dataset that required human review to characterize — when the data was ambiguous or non-standard or came in a form that automated processing couldn't handle. Standard enterprise security telemetry wasn't ambiguous. System logs, network traffic, user authentication events: all of these had established formats and established classification schemas. You didn't hire annotation specialists to handle them.
You hired annotation specialists when your data came from somewhere unusual.
2021: Behavioral Pattern Research Associate. Must be comfortable with persistent signal data and longitudinal analysis. Strong background in pattern characterization from passive sources.
Passive sources.
He wrote that down too and then read the phrase again.
Passive sources meant the data subjects weren't consenting to collection and probably weren't aware it was happening. Traffic sensors, public cameras, passive radio monitoring — all of these could be described as passive sources. None of them produced a dataset that required annotation specialists with experience in behavioral signal classification. None of them could be described as persistent signal data in a way that would require a dedicated research team to characterize.
The folder had nineteen items now. He added a working hypothesis note at the top of the document.
From the kitchen, he could hear the refrigerator cycling. Angie had come in from the front room sometime around nine and taken her position in the chair to his left — present, reading whatever was on his screen without looking directly at it, in the particular way she absorbed information that he had stopped noticing took any adjustment on his part.
"The data science headcount," she said, when he paused to write the note.
"I noticed it."
"Behavioral signal classification. Passive sources. That's not security telemetry vocabulary."
"No."
"That's neurological data vocabulary. Or —" She stopped. Started again. "That's the vocabulary for reading patterns from something that generates signal the way nervous systems generate signal."
He said: "Or did."
She was quiet.
The technical overview document was gated on CYT's website — fill out a form, receive a PDF. He had noted the phrase persistent behavioral modeling in the product overview copy two days ago and written it in the folder. He had been back to look at it three times.
The form was a lead-generation tool. The document was not actually behind any authentication; the URL was embedded in the source code of the gated landing page, which he found in twelve minutes. The PDF downloaded without credentials.
He noted the access control failure — a real finding, albeit a minor one, the kind he'd put at the bottom of a deliverable as Informational — and opened the document.
ChasePath Platform: Technical Overview, v2.3. Fourteen pages. He read it straight through for the whole picture and went back to page seven.
Fourth paragraph.
"Our predictive models leverage persistent behavioral modeling — continuous pattern analysis drawn from our proprietary dataset — to deliver threat detection that maintains accuracy across market shifts and organizational changes. Unlike point-in-time models that require periodic retraining, ChasePath's behavioral architecture ensures that predictions reflect current behavioral norms, not historical ones."
He had seen the phrase before. On their website, in the product overview: persistent behavioral modeling, behavioral signatures that don't go stale. He had written it in the folder. What the technical document gave him that the website hadn't was the rest of the sentence: continuous pattern analysis drawn from our proprietary dataset.
The dataset.
Named, in a company technical document, as the source of their competitive advantage. Not described. Not cited. Proprietary. A dataset that produced behavioral predictions with an accuracy that required it to be continuously current — predictions that didn't go stale because the source didn't go stale.
He said: "Read page seven."
Angie read it.
He watched the particular stillness that settled over her while she was absorbing something she already partially knew and was now finding stated.
"Persistent behavioral modeling," she said. The measured cadence of someone reading aloud from inside their own discipline. "In my program, that phrase would mean the models don't degrade. The training data is continuously current. The model stays accurate because the source stays accurate." She paused. "But to deliver that you need training data that doesn't go stale either." Another pause, quieter. "What are they modeling that doesn't go stale?"
"A while-true," he said. "A person being cycled."
She didn't respond immediately.
He kept going, because the logic assembled itself as he said it: "Trap a behavioral pattern. Run it in a deliberate loop. Prevent the natural degradation. The model trained on that source is always current because the source is always running — it's not going anywhere, it's not changing, it's generating continuous behavioral data for as long as you keep it cycling." He looked at the paragraph. "You'd never need to retrain. You'd never have a staleness problem. Your predictions would reflect current behavioral norms because the current state of your data source is the same as it was when you built the model."
"Because it can't advance," Angie said. Her voice had gone flat in the specific way of someone who was precisely accurate and did not find it satisfying. "The pattern cycles. Whatever's in the loop is generating data continuously. But it isn't going anywhere. It's not learning anything. It's not —" She stopped. "It's not living."
"No."
"They wrote it down," she said. "They filed it in a technical document. Under a term that sounds like every other behavioral analytics pitch I have ever read." A beat. "Anyone who knows the field and actually reads this paragraph knows exactly what they're describing. And no one reads these paragraphs. They're filler between the case studies."
He looked at the tagline at the top of page one. Chase Your Tail. Security intelligence that sees what's coming.
The name was an accurate self-description. A loop. The platform chasing its own tail. The while-true that ran because it had been told to run and hadn't been told to stop.
"They put it right in the document," she said.
"They did."
He added the technical overview to the folder with the relevant paragraph highlighted, flagged the access control finding separately, and went back to the search windows.
George appeared in the office doorway sometime around eleven.
He had been in the kitchen since morning — not visible from the desk, but present in the way large animals were present in houses they'd been in long enough to have opinions about the layout. He came through the doorway now with the unhurried certainty of something that had decided to be somewhere, crossed the floor without comment, and settled at the bookcase with his back to the wall.
Not facing Will. Not facing the screen. Facing the door.
Will looked at him.
There was something about the positioning — the bookcase, the wall, the direction of the attention — that pulled loose a memory that had been stored somewhere below the reach of ordinary retrieval. The first morning George had crossed the kitchen toward Cat on his own terms.
George had been four months old then. Already large. Already more toward the end of kitten and the beginning of something that didn't have a category yet. Cat had been on the counter — had been on the counter for years by then, since before George was a fact in the household at all, since before George was anything except a two-week-old thing yelling from a hedge that turned out to be too small to ignore. The counter was simply where Cat was.
George had been on the floor. Looking up at the counter with the expression of something that had grown large enough, finally, to have a question about the arrangement.
He had crossed toward Cat the way he approached consequential things at four months old: at three-quarter speed, head low, with the full optimism of something that didn't yet know what it was getting into. Closing the distance carefully, assessing as he went.
Cat had looked at him.
Not a warning. Not territorial. Just complete. The look of something that had been in more kitchens than this one, had already taken the full measure of what was happening here, had arrived at its conclusions before George had finished crossing the floor. The look said nothing that required translation. It was simply the look of something entire — the same way a finished argument was entire — and George had walked into it and stopped.
He had stood there for a moment.
Then he had sat down.
Not because he'd been told to. Because he had understood, through the medium of Cat's regard, that sitting down was the appropriate response to what he was looking at. That the question he'd been crossing the kitchen to ask had already been answered. He sat. He looked at Cat. Cat looked elsewhere.
The terms had been established without negotiation.
In the morning, Cat was on the counter and George was at the foot of the stairs. Both in their positions with the ease of an arrangement much older than one night.
He'd made coffee and fed both cats and gone to work.
George was at the bookcase now, watching the door. The scar across his nose was faintly visible from this angle. The same animal who had sat down on a kitchen floor twenty months ago when the look told him to. Whatever he had carried forward from that morning was in the scar, and in the way he was sitting now — back to the wall, still, watching the door.
Will turned back to the screen.
Around three in the afternoon, Angie said there was a place on Nolensville that did fried chicken.
He was about to find something in the refrigerator. A vague intention. He had not eaten since the coffee.
"Maple syrup," she added.
He stopped what he was doing. "That's not how fried chicken works."
"It is absolutely how fried chicken works and I have been eating it that way since I was twelve."
"You're describing a waffle."
"I am describing fried chicken with maple syrup, which is a dish, which has been a dish for a long time, and the place on Nolensville does not use the wrong syrup, which is the most important thing." She paused. "Real maple syrup. Not the other kind. I can’t eat it, so I’m going to need you to enjoy it for me.”
He considered the several things wrong with receiving fried chicken guidance from a dead woman who could not eat fried chicken. He considered that she had been right about the Spelling Bee pangram four days ago when he'd argued against it for forty minutes. He considered that the refrigerator contained, at this moment, three-quarters of a block of cheese, a condiment lineup, and something he'd stopped being able to identify by Tuesday.
"How do you know it's the right syrup," he said.
"Because I have eaten there approximately ninety times and I have strong opinions about their methodology."
He ordered the fried chicken.
It arrived at five-thirty. Real maple syrup, dark amber, the kind that came from Vermont and cost three times what it should. She had been right about the syrup. He held it over the food.
"You're going to pour it," Angie said. She was at the kitchen table, across from him, with the quality of attention of someone who was watching a process they already knew the outcome of.
"I'm considering my options."
"You have one option."
He poured it.
He ate a piece of chicken.
He said nothing.
She waited.
He ate another piece.
"Okay," he said.
"I know," she said, with the quiet satisfaction of someone who had been right about something before the conversation started and had simply been patient about it.
The second piece was better than the first. The crust did something to the maple syrup that had no analogy in the category of syrup on pancakes, which had been his entire relevant reference point going in and which now seemed like the wrong comparison entirely. He ate the rest and threw the containers away and went back to the desk without discussing it.
From the kitchen table, she said nothing. She didn't need to.
By eight o'clock the folder had forty-seven items.
Mail drop address: 1842 Commerce Street, Suite 400. Will had spent twenty minutes on the property records. Suite 400 was a mailbox at a virtual office service whose website offered professional business addresses starting at twenty-nine dollars a month. No commercial lease. No business improvement district filing. No parking variance. A company of forty-seven employees with no physical office on any public record.
Underground, 1.4 miles from the crash site, in a below-grade data room that had been installed before the road was last resurfaced and had no permit records anywhere in the county's construction history.
He had the full picture now. Or the full picture of what could be known from public sources, which was its own kind of picture: a security company with a mail-drop address and a data science bench heavy enough to run a behavioral research operation, job postings that described annotation specialists and passive sources and persistent signal data, a technical document that named their competitive advantage as a proprietary dataset with no data source and no citation, and a tagline that was either a coincidence or an accurate description of the architecture it was attached to.
Chase Your Tail.
A loop. A while-true. A system that ran because it had been designed to run and the design had not been told to stop.
The testimonials were real. The product worked. Forty-seven employees showed up somewhere every day and did their jobs. Whatever was underneath the last hill on the county road had been running for a decade before Angie Pierce had died above it, generating clean behavioral predictions for enterprise clients who paid for a security platform that stayed accurate because its training source couldn't change.
Persistent behavioral modeling.
He was rereading the testimonial from the financial services CISO for the fourth time when his phone rang.
Nashville number. Unfamiliar.
He looked at it for two rings. Then he answered.
"Will Hardin."
"Mr. Hardin." A woman's voice. Warm in the professional register — held at exactly the right temperature, not performed but not accidental either. The warmth of someone for whom warmth was a precision instrument. "I hope I'm not catching you at a bad time. My name is Elsie Sloane."
He looked at the laptop. Page seven. Fourth paragraph.
He said nothing.
"I'm calling about an engagement I hope you'd consider. We have a portfolio company — Meridian Systems, based here in Nashville — that we've been concerned about from a security standpoint. I've been told you're exactly the right person to call." A brief pause, professional. "We'd like a blended assessment. The full picture."
"Who referred you," he said.
"A colleague who's seen your work. I'd rather not share without asking them first." Said without apology, without defensiveness. The tone of someone explaining a protocol, not deflecting a question. "Your reputation speaks for itself, frankly."
"What kind of scope are you thinking."
"Network, physical, social engineering. We want to know where Meridian is actually vulnerable — not where their security team believes they are." Another beat. "We have concerns there are structural gaps they're not aware of."
George was in the kitchen doorway. He had been there for a while, positioned at the threshold in the way he did, facing into the office. Not at Will. Not at the phone. At some point in the east wall of the room, in the direction of the last hill on the county road.
"Send me the scope document," Will said. "We can have a conversation once I've looked at it."
"Of course." No hesitation. No adjustment. The response of someone who had expected this and was comfortable with it. "I'll send it tonight. I appreciate your time, Mr. Hardin. I look forward to working with you."
The call ended.
He set the phone on the desk. He looked at the folder. He looked at the phone. He looked at the forty-seven items in the folder and the note at the top of the first document and the phrase on page seven of the technical overview that described, in the vocabulary of a data science team that had been doing this for a decade, what was running in a below-grade facility under a road that didn't officially have a facility under it.
From somewhere behind him, from the direction of the doorway, Angie said quietly: "She called you."
"She called me."
George was looking at the east wall of the office. He had made no adjustment to his position. He was thirty-three pounds of complete stillness at the threshold, already oriented toward whatever was on the other side of the call, in the direction of the crash site.
Cat was on the kitchen counter.
He had been there all day.