"You need a lawyer," Sera said. "Not the IP attorney. Someone with national security contract experience."
"I know."
"I have a name. She's expensive and she's worth it."
"Send it."
A pause on the line. "Marcus. How do you feel about this?"
He thought about the question. He was standing at the window of the office, looking down at the street where a delivery truck was double-parked and a cyclist was navigating around it with professional irritation. The afternoon was overcast, the light flat and even.
"Interested," he said. "Cautious."
"Those are good things to be." Another pause. "Hollis is going to want to know."
"Not yet. Let me understand what's actually being offered first."
"He's a board observer. He—"
"Not yet, Sera." His voice was even. Not sharp — just certain. "I'll tell him when I have enough information to tell him something useful. Right now I have a phone call with a man who was careful not to say who he works for."
A beat. "Fair."
"How's the Monitor's second story coming?"
"They're targeting publication in three weeks. The second anomaly cluster is larger than the first."
"Good."
He hung up and turned back to his monitors.
The company was real now. It had customers, investors, employees, and apparently the attention of at least one government-adjacent intelligence interest. It had his mother's surgery scheduled for next month. It had a direction.
It did not yet have the thing the Second Gate was asking for.
*Build what others cannot follow.*
He thought about this for a long time, with the specific quality of attention the System had given him — not impatient, not anxious, but thorough. He was trying to identify the gap between *very good* and *cannot follow.* Between a tool that was better than anything else in its category and a capability that simply had no comparison.
The procurement analysis engine was in the first category. It was best-in-class. Other people could reach for it, could approximate it, could eventually build something similar with enough time and the right engineers.
*Cannot follow* required something architecturally different. Not better analysis of existing data. Different data. Different questions. Different scale.
He thought about what Cho had offered. Access to non-public data and resources to process it. He thought about what you could do with non-public procurement data at the federal level — contracts that were classified or sensitive, sub-awards that were obscured by national security exceptions, vendor networks that operated in the gap between public oversight and classified operations.
The thought was large. He set it aside to be revisited when he had more information.
He thought instead about the near-term technical horizon: the enterprise compliance product, which he needed to build in the next six weeks. He pulled up the architecture document and read through it with fresh attention.
And then he saw something he hadn't seen before.
Not a bug. Not a gap. A possibility.
The current design of the enterprise product was essentially a two-layer system: data ingestion on one side, analysis and reporting on the other, with the disambiguation engine in the middle. This was the right architecture for a company with two engineers and six months of runway. It was not the right architecture for what the system could become.
What if the disambiguation engine wasn't in the middle? What if it was distributed — a set of lightweight processes that ran continuously against a live data stream, rather than a batch process against ingested snapshots? What if, instead of analyzing procurement data after the fact, the system monitored it in real time, surface anomalies *as they emerged* in the public record?
The technical challenge was not trivial. Real-time ingestion from dozens of heterogeneous government data sources, each with its own update schedule and data format, was a hard problem. The normalization layer would need to be redesigned from scratch. The entity disambiguation would need to run in streaming mode, which required a completely different algorithmic approach.
Marcus sat down and started sketching.
He wrote for three hours without stopping. The architecture that emerged from those three hours was the most sophisticated thing he had produced since the System had activated. It was not the architecture of a startup product. It was the architecture of an infrastructure.
He sat back and read what he had written.
*Others cannot follow.*
Not because the idea was obscure. The idea was, once articulated, fairly obvious in retrospect — the same way important ideas often were. But the implementation was not obvious. The implementation required the kind of full-stack architectural vision that almost no individual programmer possessed, and that took engineering teams years to converge on.
He could build it in two months.
He would need Jin, and he would need to hire two more engineers before Jin burned out trying to keep up with the pace of what Marcus was about to ask of him.
And he would need to make a decision about Daniel Cho, and what lay behind Daniel Cho, and what kind of resources and access that door opened.
He looked at the architecture document. Then he looked at the window, at the overcast afternoon.
The System updated, quiet and final:
---
**SECOND GATE: COMPLETE.**
**REWARD:**
- Architecture Authority Lv. 3 → **Lv. 5**
- Simulation Depth Lv. 3 → **Lv. 4**
- Exploit Intuition — new sub-domain unlocked: **Institutional Mapping**
**THIRD GATE:**
*Enter rooms you were not invited into. Understand what moves the world before you move it.*
---
Marcus read the Third Gate three times.
He thought about Daniel Cho, and the data behind closed doors, and the attorney Sera was about to send him.
He thought about his mother, whose surgery date was six weeks out, whose face when it hurt had a specific quality that Nadia had described and that Marcus could see clearly even though he hadn't been in the same room as his mother in two months.
He thought about the crawler, still running, still indexing, and all the things still buried in data that no one had built the tools to find yet.
He thought about what moved the world.
He saved the architecture document, encrypted it, backed it up to three locations.
Then he wrote an email to the national security attorney.
The company had entered a new phase. So had he.