The man's name was Daniel Cho, and he had been watching Marcus Vane for three weeks before he made contact.
Cho was thirty-eight, compact, with the unhurried movements of a person who had long ago learned the cost of urgency. He worked for an organization that he would describe, if pressed, as a federal contractor with cybersecurity focus, which was the specific kind of description that was technically accurate and maximally uninformative. He had been many things over the past twelve years — software engineer, penetration tester, analyst, adjunct something-or-other on two separate assignments that had required him to be officially unaffiliated with anything. He was good at his job, which was finding things that needed to be found.
What he had found, in the context of a loose monitoring brief that had been generated by the State Financial Monitor story, was Marcus Vane.
Not just Threadline. Threadline was interesting but explicable — a smart engineer with a good idea and good timing. No, what had interested Cho's team was the architecture. The way the disambiguation engine was built. The specific algorithmic decisions buried in the processing layer that were not visible from the outside but had left certain tells in the output structure of the public reports. Tells that suggested the person who had designed this system was thinking at a level of architectural abstraction that was, in Cho's assessment and in the assessment of the two other engineers he'd had quietly review it, unusual.
Not "talented startup founder" unusual.
*Unusual* unusual.
He had looked at Marcus Vane's public record. Master's degree in computer science from a state school. Several years of contract work. Nothing that suggested the level of architectural sophistication he was seeing. The record had a gap in it — not a suspicious gap, just the ordinary gap of a person whose best work had happened in private — but something about the gap bothered him in the way that good instincts bothered him when they were telling him to look more carefully.
He had had the traffic analysis run. He had seen the honeypot.
The honeypot was excellent. It was the kind of counter-surveillance layer that most commercial companies didn't build, and when they did, they built it with security consultants. The consultant engagement logs were public record — no consultant engagement. Vane had built it himself.
Cho had looked at the honeypot architecture for a long time.
Then he had told his team to stop probing Threadline and had started making arrangements for a different kind of contact.
---
He sent an email.
It came from a personal address — a real name, Daniel Cho, with a profile that had enough genuine online presence to clear a basic check but not so much as to feel staged. The message was three sentences.
*Mr. Vane — I'm a former government engineer now doing independent advisory work in the civic data space. The Monitor story caught my attention and I'd like to understand more about Threadline's technical approach. Would you be open to a brief conversation?*
Marcus read the email twice. He ran a background check on the name — he had built a lightweight version of this for himself using public data sources, because it was the kind of thing that was useful to have. Daniel Cho's online presence was real but thin. LinkedIn profile, two conference speaker credits, a GitHub with three public repositories that were competent but unremarkable.
*Former government engineer.* The phrasing was specific. It was exactly the kind of phrase that a former government engineer would use, and also exactly the kind of phrase that a current one might use if they wanted to appear former.
The repositories on GitHub were three years old. That was the kind of tell that a careful person might not notice, but that Marcus noticed. People who were genuinely active engineers had recent commits. Three-year-old repositories meant either inactivity or the use of a different account.
He thought about this for a day.
Then he replied: *Happy to talk. Available next week.*
He spent the intervening time thinking about what he was likely looking at, which was some form of government or government-adjacent intelligence interest in what Threadline was building. He was not alarmed by this. He was curious, and he was careful.
He did not tell Jin. He told Priya, partly because she had a background that made her more useful here and partly because she would find out eventually and he preferred to be the one who told her.
"Government interest," she said.
"Possibly."
"Because of the story."
"Because of the architecture," he said. "The story drew attention to us. But whoever this is, they looked at the technical layer, not just the product."
Priya was quiet for a moment. "That's a different category of attention."
"Yes."
"What do you want to do?"
"Find out what they want," Marcus said. "Then decide."
---
The call was Thursday morning. Daniel Cho appeared on video — a plain background, a desk, the kind of neutral setup that revealed nothing. He was unremarkable in appearance in a way that Marcus suspected was intentional. His eyes were attentive.
"Thank you for the time," Cho said.
"Of course."
"The Monitor story was well-executed. The visualization of the contractor network — that was Threadline's output?"
"We built the visualization engine. The journalists did the reporting."
"Right." Cho looked down briefly — at notes, Marcus assumed. "The disambiguation methodology. Is it proprietary?"
"The implementation is. The conceptual approach is derivable from published research."
"But the specific implementation."
"Is proprietary."
Cho nodded. Not disappointed — filing. "Have you had any other inbound interest from... institutions?"
"We've had journalism organizations, compliance inquiries, and one conversation with a foundation interested in anti-corruption applications." Marcus paused. "Is there a specific kind of institution you're asking about?"
A small, brief expression on Cho's face. Not quite a smile.
"I work with a group that is interested in procurement anomaly detection at a scale somewhat larger than what Threadline is currently addressing," Cho said. "Federal and multi-jurisdiction. International, eventually."
"What kind of group?"
"Advisory function. Connected to public sector intelligence needs."
*Connected to.* The grammar of careful language.
"What would engagement look like?" Marcus said.
"Initially, a technical evaluation. They'd want to understand the system's capabilities and limitations in detail. After that, potentially a contract to either license the methodology or to collaborate on an expanded version."
"Collaborate meaning?"
"Meaning access to data sources that are not publicly available." Cho said it quietly. "And resources to process them."
Marcus sat with this for a moment. He thought about what he was hearing. Private data, government adjacent, processing resources. He thought about what that meant and what it led to, in the way that the System had taught him to think — not just the immediate step but the downstream architecture of consequences.
"I'd need to understand the legal framework," Marcus said. "And the nature of the data."
"Of course."
"And who specifically I'd be working with."
"That becomes available after an NDA and a preliminary clearance evaluation."
"Clearance evaluation."
"Background check. Standard for this kind of engagement."
Marcus nodded, slowly. He was going to say yes to the next meeting. He was not going to say yes to anything beyond that until he understood more. The shape of this was interesting. The shape of this was also, he recognized, the beginning of something larger than Threadline, and possibly larger than he had fully planned for at this stage.
But the System had said: *Build what others cannot follow.*
If the government was interested in what he had built after sixty days, then what he built in the next year would produce a different category of interest entirely.
"I'm open to next steps," Marcus said. "Have your people send over a preliminary NDA and I'll have our counsel review it."
Cho nodded. "I'll be in touch by end of week."
The call ended.
Marcus looked at the blank screen for a moment.
In his peripheral vision:
---
**Second Gate: 89% complete.**
**Inference Engine — Lv. 4 → Lv. 5** *(multi-actor strategic modeling confirmed)*
---
He opened a new document and started writing notes. Everything he had observed about Daniel Cho, in the order he had observed it. Everything he had inferred, marked clearly as inference. Every question he still needed to answer.
He wrote for forty minutes.
Then he called Sera.