Chapter 7: Tanner Presses
The discovery response dropped at 9:07 PM on day nineteen.
I was at my desk in the associate bullpen, reviewing billing documentation for Louis's inevitable next inquiry, when the email notification appeared. Tanner's response to our document request — the filing I'd spent two weeks preparing for, the one my synthesis had mapped in precise detail.
I opened the attachment.
The response was half the volume I'd predicted.
I stared at the document count for three seconds, running the numbers again. My synthesis had projected two hundred forty documents minimum, clustered in five production categories. Tanner had produced one hundred eighteen documents across three categories — and the three he'd chosen weren't the ones I'd flagged as high priority.
The Ledger pressed against my sternum. The pressure was different this time — not synthesis activating, but something else. Something that felt like the pages of a book being turned backward.
[CASE FILE OMNISCIENCE: Prior synthesis — prediction variance detected. Volume assumption: INCORRECT. Target documents: MISALIGNED.]
I pulled the Tanner complaint from my desk drawer. Opened it. The Ledger turned.
Synthesis. Ninety seconds.
This time, I held the flagged points at arm's length. The two corruption markers — I'd been calling them Tanner's filing inconsistencies, but they weren't Tanner's errors. They were mine. My wishful interpretation, my convenient assumptions, my brain filling gaps with what I wanted to see instead of what was there.
The volume assumption had been one of the flagged facts. The show's Tanner had been a discovery bully, burying opponents in document requests. I'd assumed this Tanner would follow the same playbook. The synthesis had confirmed what I expected because I'd expected it — and the corruption protocol had waved through my expectation as fact.
"External verification recommended."
I should have verified three weeks ago.
The correction took four hours.
I rebuilt the Tanner model from the ground up, checking every synthesis point against the original documents. One of the two flagged facts cleared — Tanner's settlement pressure timeline, which matched the exhibits. The other didn't clear — the volume assumption, which I'd invented out of meta-knowledge and wishful thinking.
Harvey's correction from last week echoed in my memory: "Volume is the cover. What's the actual target?"
He'd seen it. He'd seen it from the documents alone, without synthesis, without the Ledger. He'd asked the question I should have asked, and I'd incorporated his correction without understanding what it meant.
The new model was different. Tanner wasn't a volume bully — he was a precision hunter. He requested exactly what he needed and nothing more. The three document categories he'd produced targeted specific exhibits that established a chain of custody argument. He wasn't burying us. He was building something.
I wrote the corrected analysis. Three pages. Single-spaced. Attached a one-line note: "Tanner adapted — volume assumption was wrong, revised model attached."
I sent it to Harvey's inbox at 1:14 AM.
Day twenty. Harvey's office.
I stood in the doorway with the Tanner file under my arm while Harvey read the corrected analysis. His expression was neutral — the same neutral he'd shown when he read my first memo, the same neutral he'd shown when he put me on the team email chain.
"The chain of custody argument," he said without looking up. "Where does it land?"
"Exhibit D-17. The internal memo dated March 14th. Tanner's going to argue the memo was modified after production — there's a metadata discrepancy in the original, and if he establishes the chain was broken, half our documentary evidence becomes contestable."
Harvey looked up. His eyes moved across my face with the specific attention of someone who was re-evaluating a variable he'd previously categorized.
"You caught the metadata issue."
"The synthesis flagged it. I verified against the production log."
"How long?"
"Four hours for the rebuild. The original prediction was wrong — I made assumptions I should have checked."
Harvey set down the analysis. The silence stretched for three seconds, which was long enough to make most associates nervous. I waited.
"Wrong predictions happen," he said finally. "Wrong predictions you don't catch lose cases. Wrong predictions you catch and correct are how you get better."
He went back to reading. The dismissal was implicit.
I walked out of Harvey's office and stood at the window overlooking the city. Day twenty. First real predictive failure. First manual correction of a corruption-flagged synthesis.
The city below was doing exactly what it had been doing before — traffic flowing, lights blinking, the rhythm of Manhattan continuing without reference to my recalibration. The consistency was reassuring in a way I hadn't expected.
"Update the model," I told myself. "Adapt."
The corruption protocol wasn't a bug in the system. It was the system working as designed — flagging the facts I wanted to believe, forcing me to verify what felt convenient. I'd been treating the warnings as noise. They were signals.
The new protocol was simple: check every flagged point against the originals. Don't accept synthesis conclusions that confirm existing expectations. Verify before deploying.
"Right response," the Ledger seemed to whisper. "Wrong diagnosis."
I didn't understand what that meant yet. But I understood that the volume assumption had cost half a day's prep, and half a day's prep could cost a case, and Harvey had caught the direction of the error before I had.
The gap between us was narrower than it had been seventeen days ago. But Harvey still saw things I missed.
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