Kai read the folder that night.

Twenty years of zone 5 substrate monitoring data, accumulated in the standard Guild format: date, depth reading, classification result, anomaly notation. For the first three years, every reading returned standard geological results—bedrock composition, sediment layer densities, the kind of substrate data that zone management teams filed and never read again.

Then year four, entry 312: anomaly. Unknown construction signature at substrate depth 205 metres, zone 5 interior eastern section. Classification equipment returned no valid result. Notation: pre-Guild construction signature possible. Filed as unknown pending further investigation.

Further investigation had never arrived. The anomaly appeared in entry 312 and in every reading after that, the notation cycling through "unknown substrate signature" with the regularity of something the monitoring equipment had learned to flag without anyone learning to understand.

Eighty-four entries. Four years. Stage 4’s location, documented by a hunter who didn’t know what he was documenting.

He closed the folder and looked at Varn across the table.

"It’s Stage 4," he said.

Varn looked at the folder. He had been carrying it for four years. He had filed it eighty-four times. He had never received an explanation for what it contained, and he had continued filing it because unexplained data was still data and you kept data.

He closed the folder.

"What do you need from me?" he said.

He took Varn into zone 5 the next morning.

Varn’s output suppressed within two minutes of entry, as it had the day before. His Void Path null-zones went quiet. His path-expression skills went from functional to inoperative without pain or drama—simply absent, the medium they required saturated beyond what they could work through. He followed Kai through zone 5’s entity-ambient interior as an observer, his own capabilities off but his attention fully present.

He was there to see. That was enough.

Kai navigated to Stage 3’s substrate position through the sovereign seed’s road connection and initiated a full architectural read through Dragon Mode. Not the quick read he had run from outside on the first inspection. A complete read, from one end of Stage 3’s structure to the other, every element resolved.

The first junction was exactly as he had noted: coherence failure, accumulated entity-ambient in the path-energy architecture, the incompatible deposit dense but bounded. The kind of damage that the Disruption Pulse’s clearing mechanism was designed to address.

The second junction: same mechanism, slightly less dense. The entity-ambient had been pressing more evenly against the second junction than the first, distributing the incompatible deposits over a wider area but at lower concentration.

The third junction: the worst of the three, as he had assessed. A full incompatibility deposit, the entity-ambient packed into the junction at a concentration that would cause immediate structural failure under sovereign-class output load. This one needed the most work.

He continued the read. Not stopping at the three documented junctions. All the way through Stage 3’s structure, every component, every connection, the full architectural scan that Dragon Mode at King Body depth with complete sovereign seed integration could produce.

There were four junctions.

Not three.

The fourth was not damaged. It was not accumulating entity-ambient deposits. It was active—not in the way the main Stage 3 channel was inactive, waiting for activation. Active in a different sense: it had been carrying a small, continuous flow of entity-ambient in a specific direction since before Kai arrived in Vael’s Crossing, and probably for much longer.

The construction quality was identical to the other three junctions. Same materials, same engineering tolerances, the same unmistakable precision that the builders applied to everything they built. This was not a natural formation in the substrate. It was built by the same hands that built the rest of Stage 3.

But it was not in the documentation Arveth had given him. He had read the Vael’s Crossing builders’ Stage 3 records three times. Three junctions. The repair plan addressed three junctions. This was the fourth.

He held the read and mapped where the fourth junction pointed.

It was a bypass channel. Not part of Stage 3’s main flow path upward toward Stage 4 or downward toward the entity’s layer. A lateral connection—running sideways through the substrate, parallel to the zone surface rather than perpendicular to it. Routing east.

East. Away from the city. Into substrate terrain that lay beyond Vael’s Crossing’s zone system.

He did not know what was in that direction at that substrate depth. But the sovereign seed, reading the road network’s architecture, read the bypass channel as part of the same network as Stage 3. The same construction. The same purpose. Whatever it connected to was part of what the builders had built.

The builders had not written it down.

He held the read for two more minutes. Then he released Dragon Mode and came back to the surface.

Outside zone 5, Cait was at the boundary.

He gave her what she needed operationally: three damaged junction points in Stage 3’s structure, accumulated entity-ambient deposits, repair requiring Disruption Pulse application at substrate level over three days. Zone 5 permits suspended during the repair period. Zone 4 and 3 permits to remain active but above-ceiling advisories in force.

He did not mention the fourth junction. Not until he understood what it was.

Cait filed the operational notes and went to update the zone advisories.

That night he read the Vael’s Crossing builders’ folder again. Every page. The Stage 3 section twice. Nothing referenced a bypass channel, a lateral connection point, or a fourth junction. The documentation described a three-junction structure and nothing else.

He opened the Guild’s routing system and composed a message to Arveth. One question: whether the Stage 3 records in the Vael’s Crossing builders’ documentation included any reference to a secondary lateral connection point that was not part of the main flow path.

He sent it and put the folder away.

Varn was at the table with his zone 5 monitoring data spread across the surface—he had been cross-referencing the anomaly entries against the zone’s ambient readings to see if the Stage 4 signature had changed over the four years he had been recording it. He looked up when Kai sat down.

"The Stage 4 signature hasn’t shifted in twenty years of data," he said. "Not the location, not the depth, not the construction signature type. Whatever it is, it’s stable."

"Good," Kai said. Stable meant undamaged. Stage 4 would not need repair. He would be able to activate it directly.

He went to sleep.

Arveth’s response arrived before the fourth hour.

Two sentences.

The Stage 3 records in the Vael’s Crossing builders’ documentation do not reference a lateral connection point.

However: the Kael’s Seat Stage 3 records contain a notation I have never fully understood. It reads: ’fourth junction sealed at completion by builder consensus. Purpose: to be determined when carrier arrives.’

She did not add commentary.

He read it three times.

The Kael’s Seat Stage 3 had a fourth junction too. Sealed at completion. The builders had built it, sealed it, and written a single notation explaining why they were not documenting its purpose: because that purpose would be determined by the carrier, not by the builders.

Both road networks had a fourth junction in Stage 3.

Both sets of builders had left the same thing for the same person to find.

He looked at the routing message’s timestamp. Three hours until dawn. Three days of repair work ahead of him. After that, Stage 4.

After Stage 4, Stage 5.

And after Stage 5: the fourth junctions. Both of them. Connected through the substrate, routing east toward something neither set of builders had thought it was their place to name.

He went back to sleep.

He needed to be at full output for the repair.