Chapter 38: The Calibration Log Grows
Sheldon's magnetometer calibration was off again.
He stood in his office Monday morning, staring at the readout with the particular expression of a man who had seen this anomaly before and was now certain it was not random. The calibration log sat open on his desk, seven entries spanning three weeks, each one carefully documented with timestamp, magnitude, and — most importantly — adjacent personnel.
He added Entry 7.
Date: Monday, Week 11 Time: 08:47 Magnitude: +0.003 Tesla deviation from baseline Duration: 4 minutes, 23 seconds Adjacent Personnel (within 2-hour window): A. Carter, L. Hofstadter, B. Kripke
He underlined Adam's name. The third time it had appeared in the adjacent column.
Three of seven. Above random expectation for a researcher who is in the building daily.
Sheldon pulled up the building access log on his computer. Caltech's security system recorded keycard entries by hour — not precise enough for exact timestamps, but sufficient for correlation analysis.
He crossmatched the seven calibration anomaly timestamps against Adam Carter's access patterns.
The result: Adam's building access appeared within a 2-hour window of four of the seven anomalies. Not three — four. One of the adjacent entries had been masked by Leonard's presence in the same timeframe.
Four of seven.
Sheldon wrote this number in the margin of the calibration log. Beside it, he wrote: 34%
This was the probability he assigned to a non-random correlation. Not high enough to act on. Not low enough to dismiss.
He circled the number. He did not act on it. He was waiting for more data.
---
[SHELDON AND LEONARD'S APARTMENT — MONDAY EVENING]
"The building's instruments are insufficiently shielded."
Leonard looked up from his laptop. Sheldon was standing in the kitchen doorway with the calibration log, wearing the expression of someone who had prepared a presentation.
"The building's instruments are the same as they have always been," Leonard said.
"Anomalous readings require anomalous sources. I have documented seven instances in three weeks where the magnetometer in my office registered deviations outside the expected range. Seven instances, Leonard. This is not equipment drift."
"What's your hypothesis?"
Sheldon paused. The spy hypothesis was documented but unproven. The calibration anomalies supported it — an intelligence asset might carry equipment that generated detectable signatures — but the correlation was not definitive.
"I am still collecting data," he said. "However, I recall that you discussed the anomalies with Adam, and he explained them as a thermal gradient issue."
Leonard's expression shifted slightly. "He did. The building runs hotter than the instruments are rated for. It made sense."
"It made sense to you."
The specific emphasis on "you" carried weight. Sheldon did not consider Leonard's intuition a reliable variable.
"What are you implying?" Leonard asked.
"I am not implying anything. I am noting that Adam provided an explanation that satisfied you, and that his building access correlates with four of seven documented anomalies." Sheldon wrote something in the margin of the log. "This is a data point, not a conclusion."
"Sheldon, Adam is a visiting researcher. He's in the building every day. Of course his access correlates with things that happen in the building."
"Random correlation would produce approximately two of seven adjacent entries. Four of seven is above random expectation."
Leonard stared at him. "You're doing statistics on Adam's building access?"
"I am doing statistics on anomalous calibration readings. Adam's building access is a variable in that analysis." Sheldon closed the log. "I am not accusing him of anything. I am noting a pattern. Patterns require explanation."
He returned to his room.
Leonard sat with his laptop and the particular discomfort of someone who had just learned his friend was being statistically tracked by his roommate.
---
[CALTECH PHYSICS BUILDING — TUESDAY AFTERNOON]
Adam stopped by Sheldon's office for a booking question.
The calibration log was open on the desk.
He could read Entry 7 from where he stood — the date, the timestamp, the adjacent personnel column with his name underlined. The margin notation: 34%
He did not let his expression change.
"The electron microscopy suite," he said. "Is Thursday afternoon still available?"
Sheldon looked up from his computer. "Thursday, 2-4 PM. Confirmed. Please ensure the sample stage is returned to neutral position afterward."
"I will. Thank you."
Adam left.
In the hallway, alone, he ran the calculation.
Seven anomalies. Four adjacent. Sheldon had the building access log — Adam had not accounted for this. The equipment sensitivity explanation held for random distribution, where individual anomalies could be attributed to environmental factors without pattern.
It did not hold for four-of-seven adjacent.
The correlation is above random expectation. Sheldon has noticed. He has assigned a probability.
34% means he is not certain. It also means he is not wrong.
Adam pulled out his notebook and wrote, standing in the hallway:
"Miscalculated. He has the access log."
Below it:
"Adjust: 35% probability of hypothesis shift. Not 20%. 35."
He underlined the number three times.
The equipment sensitivity deflection had bought time. It had not bought safety. Sheldon was doing exactly what Adam's probability calculation had predicted — collecting data, correlating variables, waiting for enough evidence to justify a conclusion.
The conclusion was still wrong. Sheldon was looking for a spy, not an esper. But the data he was collecting would support either interpretation once he had enough of it.
If the next anomaly occurs while I am in the building, the correlation rises to five of eight. That crosses the threshold for significant pattern.
I need to be out of the building when the next anomaly occurs.
Or I need the next anomaly to not occur.
Neither option was fully within his control.
---
Leonard passed Adam in the hallway outside Sheldon's office.
"Sheldon filed an equipment sensitivity complaint with facilities," Leonard said. "Again."
Adam kept his expression neutral. "How many times has he filed?"
"Seven."
"Does facilities respond?"
"They send a form letter acknowledging receipt." Leonard paused. "It is technically a response."
This was the funniest thing anyone had said to Adam in two weeks.
He was not entirely sure why.
---
[ADAM'S APARTMENT — TUESDAY EVENING]
Adam sat at his desk and wrote the anomalous-physics deflection in full.
He had drafted it weeks ago — the version prepared for the possibility of Sheldon shifting from "spy" toward "something else." Now, with the access log correlation at 34% and the calibration log at seven entries, the deflection needed to be ready for deployment.
He wrote:
"Academy City esper field research involves proximity measurements that occasionally generate detectable electromagnetic signatures. This is documented in the literature as 'secondary field interference' — a known side effect of working with esper field equipment. The signatures are equipment-derived, not personal, and are routine for researchers in my documentation specialty."
He read it back.
Technically accurate. Specific enough to sound like real Academy City methodology. Cannot be disproved without access to my actual classification records, which are protected by the Ghost Index.
This will hold for one conversation. Maybe two. After that, the questions will get more specific, and I will need to decide how much truth to include.
He dated the page. He did not plan to need it.
He was fairly certain he would need it.
---
Entry 8 would be added to the calibration log on Thursday.
Adam did not know this yet. The Thursday anomaly would occur at 3:17 PM, during the electron microscopy booking he had just confirmed, while he was in the building operating equipment that required focused attention.
Sheldon would note the timestamp. He would check the access log. He would find Adam's name in the adjacent column for the fifth time in eight entries.
Five of eight was 62.5%. Well above random expectation.
The log did not care about any of this. It simply recorded what it recorded, accumulating data point by data point, building toward a conclusion that was still wrong but increasingly well-supported.
Adam, sitting in his apartment with the deflection written and the probability updated, had no way to prevent what was coming.
He could only prepare for it.
Get Early Access to New Chapters
Thank you for reading. For those who want to skip the wait, my Patreon is currently 21 chapters ahead of the public sites.
Schedule: 7 new chapters released every 10 days.
Benefit: Gain a significant lead of 7 to 21 chapters depending on your tier.
Support the project and start reading the next arc now: Patreon.com/IsekaiStories