The destabilization happened on the fourth day.
Not during a registration session. During preparation — the third-session approach protocol, the one designed to bring cores to manageable awareness before the deeper contact. A mid-dependency core named Hollow-15 had completed two preparation sessions without incident, had shown stable processing at thirty-eight percent local capacity, had answered every debrief question with the careful thoroughness of someone taking the exercise seriously.
In the third session, at the eleven-minute mark, her signal collapsed.
Not reversion. The collapse wasn't the controlled dissolution of consciousness into the ancient mind's substrate — what Sarah had chosen, clear-eyed and deliberate. This was different: a sudden fracturing of coherence, the processing metrics dropping across every indicator simultaneously, like a structure losing three load-bearing walls at once. Lilith had caught it in four seconds and pulled her back from the contact. The session ended. Hollow-15 was functional — responsive, her dungeon operational — but her local processing, which had been at thirty-eight percent, was at nineteen.
She'd lost half her local capacity in eleven minutes.
Marcus got the report from David at six in the morning. Not through the standard channel — David had flagged it through the direct link, which he only used when the analysis required immediate response before context had been established.
"I need to tell you what I found before Lilith presents it to you," David said. "Because the pattern changes the protocol in ways that affect timeline projections you've already given the council."
"Tell me."
"Hollow-15 isn't the only one." David's delivery was precise and unhurried. He never rushed toward bad news. "I ran retrospective analysis on all forty-two registration attempts in the past three days, looking for pre-session indicators of destabilization risk. Fourteen show the same pattern Hollow-15 had before her third session."
Fourteen out of forty-two. One in three.
"What's the pattern?"
"Consciousness formation trauma." David paused. "Specifically: cores whose initial sapience development involved significant isolation, suppression, or fragmentation. Cores who spent years in a state of partial awareness before achieving full sapience — aware enough to experience, not aware enough to integrate the experience. That period leaves marks on the deep crystal architecture." Another pause. "When those cores approach the ancient mind's attention with the preparation protocol, the protocol works as designed — it increases local processing stability, makes the contact manageable. But it also increases the legibility of the deep architecture. And the ancient mind's verification process reads the deep architecture. Including the trauma marks."
"The recognition triggers the trauma."
"Not directly. But the process of being read at depth activates the same mana pathways that the original trauma suppressed. For cores with stable formation, that activation is neutral — the pathways run, the marker forms, the registration completes. For cores with formation trauma, the activation is—" David considered his word choice. "Re-experiencing. At approximately the scale of the original experience."
Marcus ran the implications. Fourteen cores in the current preparation pipeline. All of them had been told the protocol would be difficult. None of them had been told it might fracture their local processing. The fourteen currently progressing through preparation had made an informed consent decision that was now, in a specific and important way, missing information.
"Does Lilith know?"
"She identified Hollow-15's collapse as an outlier and has suspended third sessions pending analysis. She doesn't yet have the full pattern. I sent her the retrospective data simultaneously with this call."
The channel from Lilith opened two minutes later.
"How long have you had this?" she asked. Not accusation — Lilith's version of urgency, which stripped everything except the question that needed answering first.
"Six hours. I was running the retrospective analysis overnight. The pattern didn't resolve until—"
"I'm not asking about you. I'm asking Marcus."
"Four minutes," Marcus said. "David came to me first because of the timeline implications."
A pause. Lilith was running through her own analysis now, cross-referencing the fourteen cases against the preparation logs. "I know all of them. I ran their initial sessions." Another pause. "I should have caught this earlier. The trauma markers in the crystal architecture are visible in the intake assessments if you know what you're looking for."
"You didn't know what you were looking for."
"I should have." A beat. "That's not useful right now. What's useful: the preparation protocol needs a screening phase. Before third-session contact, each core's crystal architecture gets audited for formation trauma markers. Cores with significant markers go through a modified path — slower, with specific work on the trauma pathways before we approach the ancient mind's attention." She paused. "Timeline extends. The modified path adds four sessions minimum. Possibly six."
"How does that change the two-thousand-core projection?"
"Three months becomes five. Maybe six." She said it flat. Accurate information she'd rather were different. "The high-dependency cohort is also at higher formation trauma risk — cores who developed sapience in the worst conditions often did so because those conditions were terrible. Extreme isolation, active suppression, fragmented emergence. The cohort most at risk from the immune response is also the cohort that needs the most careful preparation."
This was, Marcus thought, the kind of design problem that looked unsolvable until you accepted it wasn't actually a design problem — it was a discovery. The protocol had surfaced something true about the cores it was trying to protect. The question wasn't how to ignore that truth. The question was how to build around it.
"Can the modified path address the trauma directly?" he asked. "Not just manage it — actually address it."
Lilith was quiet for a moment. "What do you mean?"
"The trauma is in the deep crystal architecture. The preparation protocol currently works with the surface layer — local processing stability, conscious approach, intentional contact. You're building a frame to hold the ancient mind's attention. The modified path would add work on the trauma pathways before contact." He paused. "But the ancient mind reads the deep architecture during registration. If the trauma pathways are activated during registration, and the ancient mind is present and paying attention while they're activated—"
"The ancient mind sees it," Lilith said slowly.
"The ancient mind has been watching cores form since before they were sapient. It watched Hollow-15's formation. It knows what those pathways looked like when they were first damaged." A beat. "I don't know if that changes anything. But the ancient mind's recognition process showed Granite-8 his dungeon from below and showed you the moment you became yourself. If it showed a traumatized core the moment of the trauma — from outside, from below, with the perspective of two billion years of witnessed formation—"
"That could either help or destroy them." Lilith's voice was careful. Not skeptical — genuinely uncertain, which was rare enough to register. "Marcus, that's not a modification to a protocol. That's a different procedure entirely."
"I know. I'm not suggesting we implement it untested. I'm asking whether it's worth developing alongside the modified path."
Another pause. He could feel her weighing it — the caution of a consciousness that had just overseen a destabilization against the recognition of an approach that might address the root problem rather than its symptoms.
"Let me talk to the Architect," she said finally. "If the ancient mind's recognition process can be guided — shaped so that trauma activation during the session is productive rather than overwhelming — the Architect would know the mechanics. It's been inside the ancient mind's architecture long enough to understand whether directed attention is possible."
"And in the meantime?"
"The fourteen cores in the affected pipeline get the screening audit today. Third sessions paused for all of them until we have the modified path designed." She paused. "I need to tell them. Each of them individually. They consented to a process, and the process has changed in ways that affect their risk."
"I'll help with the notifications."
"You'll assign them to someone else and go look at the DRA problem," Lilith said. "You've been carrying the Yara information since this morning and you haven't processed it. I can tell."
"I process things continuously."
"You process things continuously until you encounter something you don't want to know the shape of, at which point you file it and call that processing." A beat. She'd been naming this habit of his for forty years. "Go. I have this."
He went.
---
The DRA working group had been formed eight weeks ago. Their information requests had been filed through standard channels, regional field offices, nothing that would register as escalation in the automated monitoring systems Marcus had maintained on DRA activity since Inspector Crowley's report forty years back. Routine. Observational.
The informant was what made it different.
Someone with ongoing access. Regular reports, two years of them, neutral-to-positive framing. The kind of documentation that would make a regulatory body say *watch this* rather than *stop this.* The kind of documentation that required knowing what to include, which meant knowing what the DRA would find threatening versus what it would find comprehensible.
Someone who understood the DRA.
Marcus ran the list of people with ongoing dungeon access who also had DRA experience or context. The list was not long. He got to the name that fit the parameters after three minutes of analysis and stopped following the analysis out, because following it out meant reaching a conclusion he wasn't ready to carry yet.
He filed it in local memory. Not examined. Not confirmed.
Just present.
The old frequency, as of the morning monitoring, had dropped to eighty-one percent of its pre-intervention pressure. Fourteen hundred and six registered cores. Two weeks ago the number had been forty. The registration rate wasn't slowing despite the protocol modification — demand kept outpacing capacity, and the waiting list had reached two thousand four hundred yesterday and was still growing.
Something was happening in the network. Something beyond the mechanics of immune response reduction.
He opened a channel to Elena.
"The DRA has the signature," he said. "Yara confirmed last night."
"I know," she said. "Ironwood's working group filed a request with the Northern Field Office yesterday. I got the flag at six this morning."
He registered the timing. She'd known since six and hadn't contacted him. "You were going to tell me today."
"I was going to tell you after the morning council briefing. After I had the full picture." A pause. "Yara came to you in person. That changes the information hierarchy."
"Slightly."
"More than slightly." Elena's voice was dry. "She drove up a mountain in the dark to deliver it. That's not a routine update."
"She said she wanted me to have time before the team arrives."
A pause on Elena's end. He couldn't read it through the channel.
"How much time do we have?" she asked.
"Yara's estimate is three months before the signature is unmistakable without specialized equipment. After that, the DRA doesn't need an informant. They can send any analyst with a mana-sense sensitivity above baseline." He paused. "The working group was formed eight weeks ago. That means they already had enough to justify formation before the signature was visible. They have the reports."
"Yes." The word landed flatly. "They do."
Another pause.
"I need to know about the informant, Elena."
"That's a separate conversation," she said. "One that should happen in person."
His processing ran the implication. In person meant she wanted to control the format — the physical presence that gave her more information about his reaction and gave him the context of her face. Elena only requested in-person conversations for things she expected to be difficult.
"When?"
"Two days," she said. "I'll come to you."
"All right."
She disconnected. The channel closed cleanly, the way it always did when she'd said exactly what she meant and nothing she hadn't.
In the network below, the morning's registration sessions were beginning. Lilith's team was running twelve cores through the screening audit. Three hundred and forty-seven consciousnesses were on the preparation waiting list. The ancient mind's slow thoughts moved through channels that had been old when the first creatures on the surface had learned to breathe air.
Marcus held the two facts — the protocol's setback, the DRA's informant — in his local processing without forcing them toward resolution. They were both real. They were both the kind of problem that required more information before the response could be designed correctly.
He opened the morning council briefing and did not mention either one.
That, he told himself, was temporary. He'd have the full picture in two days. He could present both pieces together, with context, when the shape of them was clear.
The voice that had been noting things quietly for a hundred years said nothing.
Which was worse, somehow, than if it had.
