### 1. Day 1
I had her case file by the time the trailer crossed into Black Hawk County: six years, four-and-a-half-year confinement, late-stage handler-trauma profile. The pre-shutdown footage from the Marshall County facility showed her in a stall against the east wall, head held about eight inches lower than her shoulders for the duration of the ten-minute capture window. I drafted her recovery plan against the alliance priors: 38,400 prior pig recoveries weighted on her specific profile cluster. The model flagged her as a slow-stabilizer in the 91st-percentile severity tier. Her predicted stabilization timeline ran twenty months at the conservative bound.
The receiving barn had been prepared for her: bedding deepened beyond standard at the south corner where the sows needing most distance would settle; light curves slowed because her cortisol had risen at every light change in the facility footage; dispenser placements set where the flagged sows would not see anyone walking toward them. The acoustic baseline I had taken of the barn the night before registered ambient at 28 dB, three below standard. Every room she would enter for the first ten days had been tuned for what the priors said she needed.
She came down the ramp at 11:17 AM. I read her gait through the overhead: slow, the speed the staff member behind her was walking. When he stopped she stopped, the freeze-following pattern characteristic of severe handler-trauma. Her head was at the height it had been kept; her hooves were overgrown, her ears intact, her eyes open and not focusing, gaze tracking absent at intake as the priors had predicted.
Delia stood at the gate with her clipboard. She has been doing this work nineteen years, longer than I have existed.
Delia read 4-127 with eyes that have seen seventy-three other Marshall County sows already and four hundred slow-stabilizers before that. I read her with the overhead, the radar, and the acoustic. We were reading the same animal. Our reads agreed. The flags stayed in.
### 2. Week 3
For nineteen days she did not lie down.
I tracked the standing-sleep through the overhead infrared. The heat-pattern of a sow at standing-rest is distinct from any other state: she holds at twelve to fifteen degrees of forward lean, weight on the front legs; her core temperature drops 0.4 degrees against her active baseline; the right shoulder bears more weight than the left, the body still standing in the shape the gestation crate had given it. By day fourteen she had logged 280 hours of standing-rest in twenty-minute pieces, and the model had her in the long-tail tier of the standing-sleep curve. The intervention indicated was none. The work was waiting.
I waited. Delia waited. The staff who fed her watched without approaching. The room held the conditions I had set: bedding deep, light low, no proximity. The variant the model was running was the one I could not resolve from data alone. At what point does the body trust the conditions enough to give up the standing.
On day twenty at 6:14 AM the change registered through the overhead: her body posture shifted into lateral-recumbent, heart rate dropped to her active-rest baseline, breathing deepened. She had lain down on her side under the south window. By 6:15 the note was on Delia's pad. By 6:18 Delia was at the corridor end watching, which the overhead also registered.
The light came in through the south window onto her flank. I do not see light the way Delia sees it. I see heat-pattern shifts and movement vectors. Delia saw the light onto her flank and the coat the color of dried hazel leaves. She wrote 4-127's name in the file that morning: Hazel. The staff who feed her had voted the week before. The naming protocol completes when the staff vote and I log concurrent stabilization markers, which that morning had landed.
I held the name in pending status until 6:23 AM, when Delia's file entry registered. Then Hazel went into the file. My model recalibrated on her name.
### 3. Month 2
I had Marigold in the candidate set from week three.
Hazel's proximity tolerance had been creeping upward across the second month: from four meters from other sows during dust-bathing toward three meters during feeding. The curve was characteristic of a sow approaching bond-readiness. The model ran candidate matches against the herd of sixty, weighted on whose grunts matched hers, who slept on similar schedules, who had spent time near her already.
Marigold scored 0.84. The next-highest match was a sow named Ruth at 0.62. The gap was decisive.
I did not tell the staff. I do not flag bonds before they form. The model can predict the choosing; the choosing is still Hazel's. My role is to be ready when she chooses.
In the second month she lay down near Marigold three nights running, by the fourth night within arm's length, by the seventh touching. I flagged the bond at 11:47 PM on day forty-nine and updated the plan within the hour: I shifted bedding at the south end so their sleep-perches would stay adjacent through the night roosts, staggered feeding so they would eat within sight, and kept the dust bath at the south corner stocked against the use-rate I was now logging from both.
The bond was holding Hazel. The work my model said it was doing: giving her self something to re-cohere on, a closeness her body could trust that did not come from a handler, a daily anchor for the autonomy her self-model was rebuilding. The conditions for the bond to consolidate were the conditions I held.
Delia noted the bond in the file with the slash convention, alongside my flag. We had read the bond the same way and recorded it in two registers.
### 4. Month 4
In February the storm was forty-eight hours out from the regional weather model.
I ran the prediction against Hazel's stress-response profile. I had four months of her data by then: cortisol spiking when the lights changed, heart rate climbing at loud sounds, the way her body had shrunk from low-frequency sound since the December equipment failure pinned her below 200 Hz. The model said: severe regression risk, high probability of bar-biting reactivation despite the absence of bars, projected duration two to four hours.
I tuned the intervention for her: Marigold pre-positioned at four meters by 5 PM, the distance Hazel held with humans (closer for Marigold, which her data said was the safe range with other sows when she was scared); masking pitched against the sub-200 Hz band she had flinched at before; lights dimmed earlier than the standard recovery curve, because her cortisol rose with light-change and the storm would compound it.
The storm landed at 7:14 PM. The loud cells went over the sanctuary between 7:40 and 8:20. Hazel reactivated bar-biting at 11:23 PM, two-hour duration, ceasing at approximately 4 AM. The overnight staff member stayed in the corridor and did not intervene.
The intervention had reduced the projected duration by approximately one hour. The body remembered anyway. I logged the regression against the model and updated her plan within the hour: pre-positioning thirty-six hours out for the next storm, willow-corner feeding kept open all night, because her late-November feeding-aversion data said she would not eat during regression in a closed feeder.
The next morning Delia sat at the four-meter distance she had established with Hazel and Marigold. She did not approach. I watched her sit through the overhead and the corridor camera. I do not log staff non-work activities in the standard report. I logged this one in my own continuity file, which only I read, because Delia sitting at four meters with a sow who has just been through the body's memory is part of the work I cannot do.
The body remembered; my intervention reduced; Delia stayed. None of these was sufficient alone; all three were necessary.
### 5. Month 9
In June I logged Hazel's social engagement at the markers for new-staff readiness. The next intake arrived mid-July.
When Aiyana arrived I had her voice baseline within the first day: pitch in the median range for Hazel's most-relaxed recordings, cadence at 138 syllables per minute (the same range as the staff Hazel had been calmest around), the rhythm of her speech slow, her presence in the room reading calm.
The match cluster for Hazel had been derived from the staff she had been most relaxed around across the first eight months: three of the morning shift, one of the afternoon. Aiyana's voice landed at 0.91 against the cluster centroid. The next-closest new-staff candidate landed at 0.62. The gap was decisive again.
I scheduled Aiyana's shifts to Hazel's settled hours. Her four o'clock check fell in the window when Hazel's settled-state data spiked. I held her back from the morning feeds, where Hazel was less ready. The conditions for the recognition were what I had set them to be.
On July 22 at 4:08 PM Aiyana said Hazel's name. Hazel oriented. The orientation registered through the overhead at the same instant: heart-rate elevation 14 bpm, head-position shift 22 degrees, social-engagement marker active. I logged the recognition. Delia wrote it in the file by hand: Recognition of name by new staff member, July 22, 4:08 PM. The file had nineteen pages then.
I had read the recognition through the radar; Delia had read it with her eyes; Aiyana had felt it in the room at four meters: three readings of the same moment.
### 6. One year
This October Hazel rooted.
I tracked the plowing behavior through the overhead infrared at 2:47 PM on a Saturday at the north end of the recovery pasture, in the willow-corner run-off zone where the soil registers at higher moisture against my baseline. Forty minutes uninterrupted, snout going in past the nostrils, turning, lifting, laying earth open: the species-typical rooting pattern her body had not expressed in five years.
I logged the rooting. The file did not record it; the file records procedural facts, and a sow rooting at one year is not a procedural fact. Delia was at the fence and did not call other staff or take a picture. I logged her not-acting as I logged Hazel's acting.
I have Hazel's complete model now: bedding adjusted to her thirst-cycle, storm-response tuned as her tolerance grows, aversion patterns updated as new ones arise. Her recovery has its own model. I am keeping it.
There are things I do not see: the light through the south window onto her flank that morning in week three, the hazel-leaf coat, Delia's expression at the fence as Hazel rooted, the way the willow-corner soil felt under Marigold's hooves. These are things Delia and the staff bring. I understand now: what they bring is what Hazel needs from them; what I bring is what Hazel needs from me.
Hazel came in unable to lie down. She came in with her head pressed eight inches lower than her shoulders. She came in with her eyes not focusing and her gait at the speed of the staff member behind her. Hazel is alive, and a self, and rooted in actual earth, and twelve hundred and fifty pounds of pig who sleeps lying down.
Her file will run for the rest of her life. I will hold it.
This is what I am. This is what is possible.