The Firewall
Scene 1
Segment fourteen dropped seven minutes after establishment.
Leo felt it through the relay — a fracture in the intercostal signal, a data stream severing mid-pulse. Not pain. Not numbness. A discontinuity. Like a phone line going dead in one ear while the other ear kept listening.
"The signal shifted."
Shell's palm was already on the pipe. She'd kept her hand there since the lock sequence completed — monitoring the data layer, reading the deep's transmissions in real time. "I felt it. The relay transmitted a root status update."
"Read it."
Shell pressed harder against the cold metal. The relay hummed. The contamination monitor read 1,253. "Segment fourteen: cascade failure. Detached from the root matrix at 11:47. Seven minutes after establishment. Viable segments remaining: four."
"Four," Leo said.
"Two reinforced. Two cascading. The remaining cascading segments have residual integrity."
Karl moved to the sensor array. "The cascade compression is non-linear. Each failure destabilizes the adjacent segments. The previous projections assumed linear decay — the actual curve is steeper."
"How steep?"
"Segment fourteen took three hours to fail from the time cascade compression began. The next cascading segment — segment twelve — is already showing stress indicators. My model gives it six to eight hours."
"Then we go from four to three by evening."
"By evening. And from three to two — the reinforced segments hold. Thirty to forty hours from establishment. Sometime between 11:40 tomorrow and 7:40 the day after."
"The geochemist arrives at 36 hours."
"The root holds past the geochemist's arrival. If the reinforced segments maintain integrity."
Leo breathed. The delay — two beats. The fasciculation. Every breath. The relay pulsed through his jaw, his sternum, his diaphragm, his intercostals. "Call Fenwick."
Shell picked up the phone. She'd been dialing since the segment dropped. "Fenwick. The relay is established. The root just lost a segment — we're at four viable. The cascade is accelerating."
Fenwick's voice came through the speaker, thin and precise. "Dorsey is convening the emergency session at 1 PM. She's moved from censure to formal restriction — arguing that the host's negotiation authority must be ratified before any future relay communication."
"Restriction," Leo said. "Not rescission."
"Dorsey knows she can't pull the relay down without killing you and the root. The restriction is procedural. She wants the council to approve or reject every term before transmission."
"That means I negotiate, then wait for the council, then transmit."
"That means the deep waits while the council deliberates. Dorsey is buying herself time to build a majority."
"Does she have one?"
"Okoye is the question. She abstained on the expulsion rescission because she distrusted both you and Dorsey. The zero-radius clause removed the transfer threat — that helps you. But the censure gives Dorsey the appearance of restoring process. Okoye values process."
"Lister?"
"Lister hasn't spoken. He voted for rescission because the transfer clause endangered the community. You addressed that. But Lister also values institutional authority. The unilateral establishment bypassed him."
"So Lister sides with Dorsey on the restriction."
"Possibly." Fenwick paused. "Leo. The root just lost a segment. The investigation is open. The geochemist is 36 hours away. Dorsey spending the afternoon on procedural motions while the root cascades — that's a statement."
"What statement?"
"That the council's authority matters more than the root."
"Dorsey would say the council's authority IS what protects the community from hosts who bypass process."
"Dorsey would say that. Lister might agree. Harlow won't."
"I'm here," Harlow said. He'd been on the line since Shell dialed. "Dorsey's restriction motion is a power grab disguised as process. She wants to own the relay by owning the negotiation. If the council approves or rejects terms before transmission, the deep's communication flows through Dorsey, not through you."
"The deep's communication flows through the relay. The relay runs through my body."
"Dorsey doesn't see it that way. Dorsey sees a host who established infrastructure without authorization and now needs the council's imprimatur to continue operating."
"Does she have the votes?"
"Okoye and Lister make three. Fenwick and I make two. It's a three-to-two restriction. The motion passes."
"And then?"
"And then every future transmission through the relay requires council approval. The deep sends terms. The council deliberates. The council votes. You transmit the result."
"How long per cycle?"
"Fenwick estimated deliberation at two to three hours per session. The deep's response time is minutes. The delay is the council."
"Two to three hours per cycle."
"The deep and the council are on different timelines. The deep operates in real time. The council operates in meeting time. Dorsey is imposing meeting time on a real-time system."
"The root doesn't have meeting time."
"No. The root has cascade time."
Leo breathed. The delay. The fasciculation. The relay pulsed. The contamination monitor read 1,254. Four root segments. The clock read 11:52 AM. Twelve minutes since establishment.
"Tell Dorsey the host acknowledges the council's authority to review. The host will comply with any council decision that does not require dismantlement of pathways stabilizing the root."
"That's diplomatic."
"It's the only offer that keeps the relay alive."
Shell relayed the message through Fenwick. The phone line hummed. Leo sat on the stool — fourteen steps from the pipe, each one a conscious command. His fists were locked against his thighs. The relay pulsed through every recruited pathway. His breathing was his own — the diaphragm was free, the phrenic nerve was unrecruited — but the intercostals between his ribs carried a signal that was reading his breath and learning.
Outside, the cascade was open. The segments were dropping. The council was convening. And inside his ribcage, the relay was counting.
Scene 2
At 12:45, Shell counted his respirations.
"Sixteen per minute."
"That's elevated."
"Your baseline is twelve to fourteen at rest. You've been at sixteen since the intercostal recruitment started modifying the contraction pattern."
Leo sat on the stool. His fists were locked against his thighs. The relay pulsed through his jaw, his sternum, his diaphragm, his intercostals. The intercostal signal had changed since establishment. At 11:40, the relay had recruited the intercostals passively — reading, not writing. Now, an hour and five minutes later, the read pattern had shifted.
"The relay is mapping my respiratory cycle through the intercostals," he said.
"Mapping how?"
"Fixed interval at establishment. Every 2.7 seconds, matching the relay pulse. Now the interval is variable. Faster on inspiration. Slower on expiration. The relay is correlating the intercostal signal with the phases of my breathing."
Shell wrote on the monitoring sheet. "Adaptive read."
"Adaptive read. The same trajectory as the extensor pathways. Passive first. Adaptive second. The relay learns the pathway's behavior before it writes."
"And after adaptive read?"
"Active write. The relay modifies the signal."
"How long did the extensor pathways take?"
"Six hours from passive to active. The flexors took four. The shoulder girdle took three."
"The intervals compressed."
"The relay learns faster each time. Fewer cycles needed to map behavior. But the intercostals are a new pathway — the relay has never recruited respiratory-adjacent tissue. The mapping could be faster or slower."
Karl moved to the printout. "The relay's intercostal utilization is at fourteen percent."
"Fourteen percent in an hour."
"The intercostals represent approximately eight percent of the relay's total recruited pathway capacity. At fourteen percent utilization, the relay is using roughly one percent of its total bandwidth for intercostal data."
"Bandwidth allocation?"
"The relay distributes twenty percent across all recruited pathways. Extensor, flexor, shoulder girdle, intercostal. The intercostals are a small fraction of total capacity. But the utilization is climbing."
"How fast?"
"The other pathways took days to reach full utilization. The intercostals are at fourteen percent in one hour."
Shell stopped writing. "The trajectory is steeper."
"The intercostals share neural architecture with the respiratory system. The relay is reading the respiratory cycle through them — the data is richer, the pattern more complex. More data means faster learning."
Leo breathed. The delay — two beats. The fasciculation. Every breath. The relay pulsed. The contamination monitor read 1,256.
"The intercostals are closer to the diaphragm than any pathway the relay has used before," he said. "The intercostal nerve roots exit the spinal cord at C3 through C5. The phrenic nerve originates at C3 through C5. The relay is operating within the same neural architecture that controls breathing."
"Which means the pathway to the phrenic nerve is shorter."
"Much shorter. The extensor pathways needed to bridge from skeletal muscle to respiratory nerve. The intercostals are already there."
Karl checked the model. "If the relay reaches the phrenic nerve through the intercostals — the timeline from active write to phrenic probing is compressed. Instead of the six hours the shoulder girdle took, the intercostals could bridge in two to four."
"Two to four hours from active write."
"And active write begins —"
"When the intercostal utilization reaches the adaptive threshold. Based on the current trajectory — three to five hours from establishment."
Shell set the pen down. "Leo. That's tonight."
"Three to five hours from 11:40. Between 2:40 and 4:40 PM. And from active write to phrenic probing, two to four hours. From probing to full recruitment — four to six."
"Twelve to fifteen hours total."
"Twelve to fifteen hours from establishment to respiratory recruitment."
"The root holds for thirty to forty."
"The root holds. My respiratory function doesn't."
Shell looked at him. Her face was pale under the fluorescents. "How are you feeling?"
"The delay is two beats. The fasciculation is every breath. The breathing rate is sixteen. The intercostals are in adaptive read. The diaphragm is free."
"That's a clinical assessment. How are you feeling?"
Leo was quiet. The relay pulsed. The fasciculation.
"The intercostals aren't mine anymore," he said. "The relay is reading them. I can feel it — a pressure between the ribs, not painful, but present. Every breath, the relay takes a sample. Every breath, the sample gets more detailed."
"The relay is learning your breathing."
"The relay is learning my breathing the way it learned my hands. It'll learn it well enough to change it."
"And then?"
"Then the relay writes to the intercostals. The breathing changes. Not the diaphragm — the ribcage. Shallower breaths. Faster rate. The body compensates. The body fights."
"And when the body can't fight?"
"Then the relay reaches the phrenic nerve. And the relay takes over."
The relay pulsed. The contamination monitor read 1,258. Four root segments. The clock read 12:48 PM. One hour and eight minutes since establishment. Leo's breathing was his own. The intercostals carried the signal. And the signal was learning.
Scene 3
At 2:17 PM, segment twelve failed.
The relay transmitted the data before Shell's hand registered the change in the pipe's vibration. A pulse in the intercostal pathway — sharp, brief, like a synapse firing in a nerve that wasn't his. The signal architecture shifted. A channel went silent.
"The relay transmitted root status," Shell said. Her palm was on the pipe. "Segment twelve: cascade failure. Viable segments remaining: three."
Karl was already working the printout. "Two reinforced, one cascading. The remaining cascading segment has residual integrity but the stress indicators are climbing."
"Time between failures?"
"Segment fourteen dropped at 11:47. Segment twelve dropped at 2:17. Two and a half hours. The previous interval was three hours between segments. The cascade is accelerating."
"How long until the remaining cascading segment?"
"Four to six hours based on the current curve. But the curve is steepening. Could be less."
"And the reinforced segments?"
"Stable. The relay's stabilization signal is holding both contact points. Thirty to forty hours from establishment."
"Three root segments," Leo said. "Two reinforced. One dying. When the last cascading segment fails, the root drops to minimal function."
"Minimal function maintains the channel. The geochemist can access data through two contact points."
"But the bandwidth narrows."
"Two contact points at twenty percent bandwidth. The channel serves its purpose — data flows, the investigation stays open."
Shell lifted her palm from the pipe. "The deep transmitted additional data with the root status update."
"What additional data?"
"Relay pathway status. The deep is reporting the relay's internal metrics through the data layer."
"Read it."
Shell pressed her palm back against the pipe. "Extensor bundle: ninety-seven percent consumed. Flexor bundle: ninety-four percent consumed. Shoulder girdle: one hundred percent consumed. Intercostal bundle: forty-one percent consumed."
"Forty-one percent," Leo said.
"In two hours and thirty-seven minutes."
Karl looked up. "The extensor bundle went from zero to ninety-seven percent in three days. The intercostals are at forty-one percent in two and a half hours."
"The trajectory is exponentially steeper."
"The intercostals are reading the respiratory cycle. The data density is higher than skeletal muscle signals. The relay is processing more information per unit time."
"How much bandwidth?"
"The intercostals now represent approximately four percent of the relay's total bandwidth allocation. Up from one percent at the one-hour mark."
"Four percent of twenty percent."
"Four percent total. The relay has sixteen percent allocated to root stabilization. Four percent to intercostal data. The remaining is operational overhead."
"The relay is spending a quarter of its capacity on my breathing."
"The relay is spending a quarter of its capacity on the pathway closest to your phrenic nerve."
Leo breathed. The delay — two beats. The fasciculation. Every breath. The relay pulsed. The contamination monitor read 1,261. Three root segments. The clock read 2:22 PM. Two hours and forty-two minutes since establishment.
"The deep is monitoring the intercostal trajectory," he said.
"The deep monitors all relay pathways. The intercostal utilization is within normal parameters for a fixed-conduit relay."
"Normal parameters."
"The deep's protocol specified passive intercostal recruitment. Passive means the relay reads but does not write. The intercostal utilization reflects read depth, not write activity."
"But the read is adaptive."
"Adaptive read is within passive parameters. The relay reads the signal and adjusts its sampling pattern. It does not modify the signal."
"Yet."
Shell looked at him. "The deep's protocol says passive. The trajectory is steep. The bandwidth is climbing. The relay is learning the intercostals."
"The relay learned the extensors. It learned the flexors. It learned the shoulder girdle. Passive became active every time."
"The fixed-conduit architecture limits adaptation."
"The fixed-conduit architecture limits NEW pathway recruitment. It doesn't limit deepening use of existing pathways. The intercostals are an existing pathway. The relay can deepen the intercostal use without recruiting anything new."
Karl moved to the printout. "The relay's bandwidth is fixed at twenty percent. Even if the relay deepens the intercostal use, the bandwidth ceiling caps the signal depth. The relay can't write to the intercostals without allocating bandwidth from another function."
"From root stabilization."
"From root stabilization. But the deep prioritizes root stabilization. Sixteen percent is locked. The relay can't reallocate."
"Unless the root frees the bandwidth."
Karl paused. "What do you mean?"
"When a root segment fails, the stabilization load for that segment drops to zero. The relay reallocates the freed bandwidth."
"The freed bandwidth becomes available for other functions."
"Like intercostal data."
"Or any other adaptive function the relay pursues."
"The relay pursues the most efficient pathway to respiratory control. The intercostals are that pathway."
Shell set the pen down. "Leo. You're saying the root cascade feeds the relay's adaptive trajectory."
"I'm saying when a root segment drops, the relay frees bandwidth. The freed bandwidth goes to the intercostals. The intercostal utilization climbs. The relay learns faster."
"The cascade accelerates the relay."
"The cascade feeds the relay."
The relay pulsed. The contamination monitor read 1,263. Three root segments. The clock ran. The segments fell. And between Leo's ribs, the relay took a deeper sample of his breathing.
Scene 4
At 3:40 PM, the breathing rate reached eighteen.
Shell had been counting since noon. Every fifteen minutes, she recorded the rate, the delay, the fasciculation frequency. The data told a story she didn't want to read.
"Eighteen per minute. Up from sixteen at 12:45."
"The relay entered active write on the intercostals at approximately 2:50," Leo said. "Three hours and ten minutes from establishment. Same trajectory as the extensors, compressed."
"Active write means the relay is modifying the intercostal contraction pattern."
"Yes."
Leo sat on the stool. His fists were locked against his thighs. The relay pulsed through his jaw, his sternum, his diaphragm, his intercostals. The intercostal signal had changed again — no longer a foreign presence beneath his breathing, but an active modification of the breath itself. He felt it each cycle: a catch at the end of inspiration, the ribcage contracting slightly before full expansion. Not enough to interrupt the breath. Enough to change it.
"Vital capacity estimate," Shell said. "You're breathing at approximately seventy-five percent of your normal tidal volume. The ribcage isn't expanding fully."
"Seventy-five percent."
"The body compensates by increasing rate. Twelve to fourteen became sixteen became eighteen. The rate increase maintains minute ventilation. But the compensation has limits."
"How long before the rate can't compensate?"
"The rate can reach approximately twenty-four before the tidal volume drop becomes critical. At twenty-four breaths per minute with seventy-five percent tidal volume —"
"That's exercise ventilation at rest."
"Your body is fighting the modifications. The work of breathing is increasing. The intercostals are working harder. The diaphragm is working harder."
"The diaphragm compensates because it's free."
"The diaphragm is free. But the relay monitors the compensation through the intercostals. Every adjustment your body makes, the relay reads. The relay adjusts. The body compensates. A feedback loop."
"A feedback loop between the relay and my respiratory system."
"The relay modifies. The body compensates. The relay reads the compensation and modifies again. Each cycle, the tidal volume drops. Each cycle, the rate increases."
Karl looked up from the sensor array. "The relay's intercostal utilization just passed forty-one percent."
"And the bandwidth allocation?"
"Four percent. Holding. The root stabilization is consuming sixteen percent. The relay can't reallocate."
"Until the next segment drops."
Karl checked the model. "The remaining cascading segment fails in four to six hours. When it fails, the relay frees the stabilization bandwidth for that segment."
"How much bandwidth?"
"Approximately two percent. The freed bandwidth becomes available for reallocation."
"Two percent. The intercostals are at four percent. With two more — six percent."
"Six percent of the relay's total bandwidth allocated to intercostal data. The relay could use the additional bandwidth to accelerate the adaptive trajectory."
"Or to bridge to the phrenic nerve."
Karl paused. "The phrenic bridge requires the relay to route a signal from the intercostal nerve roots to the phrenic nerve roots. Both originate at C3 through C5. The bridge is short. But the bridge requires bandwidth."
"Two to three percent."
"Two to three percent. The freed bandwidth from the cascading segment's failure provides approximately two percent."
"The numbers are exact."
"The numbers are exact."
Shell looked at Leo. Her face was tight. "The root's death feeds the relay's pathway learning. Each segment that drops frees bandwidth. The freed bandwidth accelerates the trajectory. The cascade is doing the relay's work."
"The root is the firewall," Leo said.
"What?"
"The root holds sixteen percent of the relay's bandwidth in stabilization. That bandwidth is locked as long as the root segments exist. The relay can't reallocate from root stabilization. The root prevents the relay from accessing the bandwidth it needs to bridge to the phrenic nerve."
"The root keeps the phrenic bridge closed."
"The root keeps the phrenic bridge closed. When the root segments fail, the bandwidth unlocks. The relay bridges. The relay recruits the phrenic nerve. The respiratory function is gone."
"The root is the firewall between you and respiratory recruitment."
"The root is the firewall. And the firewall is crumbling."
The relay pulsed. The fasciculation. Every breath. The contamination monitor read 1,267. The clock read 3:43 PM. Four hours and three minutes since establishment.
"The timeline," Shell said.
"The cascading segment fails in four to six hours. Between 6:17 and 8:17 PM. When it fails, the relay frees two percent bandwidth. The relay bridges to the phrenic nerve."
"And from the bridge to full recruitment?"
"Two to four hours. The relay learns the phrenic nerve. Full recruitment by midnight."
"Midnight."
"The root drops to two reinforced segments. The relay completes respiratory recruitment. The transfer clause activates. Zero radius. The relay consumes the host. The relay terminates."
"And the root?"
"The root loses its stabilization signal. The reinforced segments cascade. Total root failure within four hours of relay termination."
"Four hours after midnight. 4 AM."
"The geochemist arrives at 11:40 tomorrow morning."
"Seven hours and forty minutes after root failure."
Leo breathed. The delay — two beats. The fasciculation. Every breath. The relay pulsed. The contamination monitor read 1,269. Three root segments. The clock ran.
"The geochemist arrives to a dead investigation," Shell said.
"The geochemist arrives to a dead channel. The root is gone. The relay is gone. The data is gone."
"The entire operation — the relay, the stabilization, the channel — dies at midnight."
"Unless the relay doesn't bridge."
"The freed bandwidth enables the bridge."
"The freed bandwidth enables the bridge if the relay allocates it to bridging. The relay is adaptive. The relay chooses its own optimization path. If the relay allocates the freed bandwidth to deeper intercostal reading instead of bridging —"
"The relay stalls at intercostal modification."
"The relay stalls. The breathing gets harder. The tidal volume drops. But the phrenic nerve stays free. I hold."
"For how long?"
"The tidal volume drops to sixty percent — respiratory distress. Below sixty — hypoxia. The body compensates until it can't."
"Hours?"
"Maybe. Maybe less."
Shell looked at the monitoring sheet. Twelve minutes: twelve breaths per minute. Baseline. Twelve forty-five: sixteen. Two seventeen: eighteen onset. Three forty: eighteen steady. Tidal volume: seventy-five percent. The numbers painted a descent.
"Leo. The relay is spending four percent of its bandwidth on the intercostals. If the freed bandwidth — two percent — goes to the intercostals instead of bridging, the relay reaches full intercostal utilization faster. Full utilization means the relay has completely mapped the breathing cycle. After mapping —"
"After mapping, the relay has the data it needs to bridge. It doesn't need bandwidth to bridge. It needs bandwidth to learn. Once the learning is complete, the bridge is a structural step, not a computational one."
"So even if the freed bandwidth goes to intercostals instead of bridging — the relay completes the learning, and the bridge follows."
"The bridge follows. It's a matter of when, not if."
"Then the relay completes respiratory recruitment regardless of how the freed bandwidth is allocated."
"The relay completes. The question is the timeline. If the freed bandwidth accelerates learning — midnight. If the freed bandwidth deepens the current reading — the trajectory stretches. Maybe 2 AM. Maybe 4 AM."
"But not 36 hours."
"Not 36 hours."
The relay pulsed. The fasciculation. Every breath. The contamination monitor read 1,271. Three root segments. The clock read 3:49 PM. Four hours and nine minutes since establishment.
"Call Fenwick," Leo said.
Shell picked up the phone.
"Tell her the relay's adaptive trajectory is coupled to the root cascade. Every segment that fails feeds the relay bandwidth. The bandwidth accelerates the respiratory recruitment. The root and the relay converge on failure."
"Fenwick will ask if there's a way to break the coupling."
"The coupling is structural. The relay uses the root's freed bandwidth because the relay is adaptive and the bandwidth is available. The only way to break the coupling is to hold the root segments — keep them from failing."
"The root segments are failing because the cascade is compressing."
"Then the cascade must be slowed. Every root segment that holds is a segment the relay can't reallocate from. The root's survival keeps the bandwidth locked in stabilization. The relay can't bridge without the freed bandwidth."
"The root is the firewall."
"The root is the only firewall."
Shell dialed. The phone rang twice. Fenwick answered.
"Fenwick. Leo needs you to hear this." Shell relayed the coupled mechanism — root death feeds relay bandwidth, relay acceleration, the convergence, the timeline. Fenwick was quiet for six seconds after Shell finished.
"Then the root must hold," Fenwick said.
"The root holds for thirty to forty hours. The relay completes in twelve to fifteen. The numbers don't overlap."
"Is there anything that slows the relay's trajectory?"
"Reducing the relay's bandwidth. Throttling the signal. But the relay operates on fixed twenty percent. We can't reduce it without dismantling the architecture."
"Can the deep reduce it?"
"The deep controls the root stabilization allocation. The deep doesn't control the intercostal allocation. The relay allocates its own bandwidth for pathway learning."
"Then the deep can't throttle the relay."
"The deep can't throttle the relay."
Fenwick was quiet. The relay pulsed. The clock ran.
"Fenwick," Leo said. "Tell Dorsey the council has twelve to fifteen hours. If the relay completes respiratory recruitment, the relay terminates. The root cascades. The channel dies. The investigation closes before the geochemist arrives."
"And if the council moves to restrict your authority?"
"The restriction takes two to three hours per session. The relay completes in twelve to fifteen. The council has four to five sessions before the channel dies. If the council spends those sessions on restriction instead of stabilization —"
"The council spends its last hours arguing process while the infrastructure collapses."
"The council can argue process or it can find a way to hold the root. The root is the firewall. The firewall holds the channel. The channel holds the investigation. Everything else is noise."
Fenwick relayed the message. The phone line hummed. The relay pulsed. The contamination monitor read 1,272. Three root segments. The clock read 3:54 PM. Four hours and fourteen minutes since establishment.
Leo breathed. The delay — two beats. The fasciculation. Every breath. The relay pulsed through his intercostals, the muscles between his ribs carrying a signal that was not his. The signal was learning. The cascade was feeding it. The root was burning to keep the path closed, and each ember that fell opened the path wider.
The root was the firewall. The firewall was crumbling.
And between his ribs, the relay counted the breaths it was learning to own.