Chapter 222: Variables and Anchors

11:40 a.m. Lin Chen wrapped crushed ice in a double layer of plastic bags and pressed it against the outside of his left ankle. The swollen skin showed a dark red flush. There was a dull pain when he pressed it with his fingertips, but the joint’s dorsiflexion was still within a controllable range. He stared at the electronic clock on the wall, watching the numbers tick forward one by one. Twenty minutes. He couldn’t ice it for too long, or the blood vessels in the extremity would constrict too much. If his walking stiffened that afternoon and his gait deformed, all the compensatory pressure from sitting for long periods would end up concentrated on his lumbar spine.

The only sound in the hospital room was the steady ticking of the monitor. Xiaoman was asleep, breathing evenly, her chest rising and falling faintly in rhythm with the IV line. Lin Chen got up as quietly as he could and went to the nurses’ station to sign for the official EEG report. In the results column it read: mild abnormal discharges, no typical epileptiform waves observed. The attending physician stopped him in the corridor and said in an even tone, “It’s good that the low-grade fever is gone. The epilepsy is under pretty good control, but lately there’s been too much staying up late and too much emotional fluctuation, and that makes it easy to trigger. Family needs to keep a closer eye on her. The medication can’t be interrupted.”

“Understood.” Lin Chen nodded, folded the report in half, and slid it neatly into the inner compartment of his briefcase.

Back in the room, he changed into a light blue shirt that had been ironed, buttoning it to the second button. The dark circles under his eyes couldn’t be hidden in the mirror, but his shoulders and back were straight. He checked his bag once: notebook of mistakes, black gel pen, laptop, power bank, two printed architecture sketches, spare power adapter. Phone battery: 84%. Everything in place.

At 2:50 in the afternoon, he took a taxi to Building C in the tech park. The ride took twenty minutes. He kept his left foot resting on the front passenger floor mat, avoiding putting weight on it. The driver glanced at him in the rearview mirror and said nothing. An old song was playing in the car at very low volume. Lin Chen closed his eyes and went over the framework of the afternoon’s conversation in his head. No small talk. No probing. Straight to the core. Su Man came from a technical background; people who understood the field didn’t need preamble. He only needed to confirm two things: the safety margin of the cash flow, and whether the business logic was truly irreplaceable.

At 2:55 p.m., in the café downstairs from Building C. A booth by the window. Su Man was already there. She wasn’t wearing her shell jacket today; instead she had on a dark gray knit sweater, her hair pinned up casually. An open MacBook sat on the table, its screen showing a data-flow topology diagram. Beside it was an untouched Americano.

“On time.” Su Man looked up and motioned for him to sit.

Lin Chen pulled out the chair and set his backpack by his feet. When his left foot touched the ground, he shifted his center of gravity slightly. “President Su.”

“Just use my name.” Su Man closed the laptop and pushed over a printed draft of a business plan. “Look at the table of contents first. We’re building a vertical annotation platform for medical imaging and industrial quality inspection. Current team is seven people: four annotators, two algorithm engineers, one product manager. We’ve got the MVP running, but the bottom-layer data circulation is the choke point. Cleaning efficiency is low, version management is messy, and client delivery cycles have stretched to three weeks. I’ve looked at your open-source engine. The lineage tracking and outlier interception logic can be reused directly.”

Lin Chen opened the draft. The table of contents was clear, but the financial projections section had only a simple summary sheet. He looked up. “How many months can your cash flow last?”

“There’s 800,000 left on the books. At the current burn rate, six months. If we get an A-round letter of intent next month, we can last to year’s end.” Su Man spoke evenly. “But I’m not planning to stay alive on financing. What I want is a product that can close the loop. If you come in, technical partner, fifteen percent equity; option pool calculated separately. No high salary up front. Dividends tied to project delivery milestones. If we make it, we all eat meat together. If we don’t, we part cleanly—no dragging things out.”

Lin Chen’s fingers rubbed lightly against the edge of the paper. Fifteen percent. Not low for an early-stage project, but what it bound was execution responsibility, not decision-making power. He asked his second question: “What’s the core barrier in your data preprocessing? There are so many open-source tools now—why would clients pay for your platform?”

“It’s not about the tools. It’s about the scenario.” Su Man leaned forward, eyes sharp. “Generic cleaning engines handle standardized structured data. But the DICOM headers in medical imaging and the RAW formats in industrial quality inspection all define fields differently. What clients want isn’t ‘cleaned up.’ It’s ‘cleaned to match their quality-control standards.’ We need to distill industry know-how into a rule base, then let the model adapt to it. You understand the underlying flow. I understand the business scenarios. Only together can we compress delivery cycles to under a week.”

Lin Chen was silent for a few seconds. The logic held. The pain point was real. But risk points surfaced in his mind immediately: vertical scenarios meant a high degree of customization, and labor costs would rise exponentially; 800,000 in cash flow could not support a large-scale annotation team; if the A-round failed to materialize for too long, technical debt would eat back into the product.

“No performance bets,” Lin Chen said. “No touching gray-market data sources. I lead the technical architecture, but business requirements have to go through SLA evaluation. If customized demands exceed capacity by thirty percent, they get cut or delayed. Dividends are settled based on actual payments received. No drawing castles in the air.”

Su Man smiled. It was a shallow smile, not reaching her eyes. “You’re more cautious than you were three years ago. Fine. I accept those bottom lines. But I want a V1.0 architecture overhaul plan from you by next week. I’ll open access to the current codebase. Bring one person with you and get the medical-imaging DICOM parsing module running first. If it runs, we sign. If it doesn’t, we’ll treat it as a technical exchange, and I’ll pay your consulting fee.”

“Fine.” Lin Chen nodded. He had no need to bind himself immediately. First verify technical feasibility, then talk equity. That was his pace.

At 3:40, the meeting ended. Su Man stood and handed him an access card. “Permissions go live tomorrow. Call me anytime if there’s a problem.”

Lin Chen took the card and slipped it into his inner pocket. When he walked out of the café, the early winter sunlight slanted across the glass curtain wall, harsh on the eyes. He stood on the steps and drew in a deep breath of cold air. A wave of soreness swelled through his left foot. He went down the steps slowly and hailed a cab back to the hospital.

On the way, his phone vibrated. It was a message from the engineer stationed on-site with the team: “Brother Lin, the demo environment on President Wu’s side has crashed. The logs show a cyclic dependency in the lineage graph nodes, causing a memory overflow. The client’s technical director is pushing in the group chat. Says if it’s still like this for Wednesday’s demo, they’ll dock 30% of the final payment.”

Lin Chen stared at the screen. Cyclic dependency. He had left anti-loop detection in the V2.0 architecture, but the raw data the client imported had mixed in dirty historical data and triggered a boundary condition. This wasn’t something a minor patch could solve. The dependency parser would need to be rewritten—two days at minimum.

He replied: “Send me the logs and raw data samples. I’ll have a fix plan by tonight before eight. Keep the client stable on-site for now—tell them we’re doing stress-test optimization.”

Sent. He leaned back in the rear seat and closed his eyes. His mind began breaking down the problem: node deduplication, topological sorting, timeout circuit breaking. The technical path was clear, but time was tight. He had to divide his energy between “fixing the demo environment” and “pushing Su Man’s V1.0 overhaul.” He couldn’t do both at full strength. He could only split the work into time blocks.

At 4:30, he returned to the hospital room. Xiaoman was already awake, sitting up in bed and looking through an old picture book. The cover had hand-drawn stars on it, crooked and uneven.

“Ge.” Xiaoman looked up. “You’re back.”

“Mm.” Lin Chen hung up his coat and sat down. “I read the report. It’s fine. Keep taking the medicine.”

Xiaoman nodded and handed over the picture book. “Look. I drew a new one.”

Lin Chen opened it. The paper was rough, the pencil strokes heavy. It showed a man sitting in front of a computer, the screen filled with densely packed code. Beside it was written: “Big brother is fixing the road.”

Lin Chen’s throat tightened. He closed the picture book, his voice steady. “It’s very good. When I get through this busy stretch, I’ll take you to the river to see the real night view.”

Xiaoman smiled, showing two pointed canine teeth.

Lin Chen opened his laptop and connected to the internal network. The log files had already been sent over. He pulled up a terminal and started running dependency-tree analysis. The light from the screen reflected across his face. The keyboard clicks came fast and regular. His left foot throbbed faintly, but he didn’t move. He first wrote a memory probe script and captured the GC logs, confirming that the stack overflow was caused by recursive parsing without a depth limit. Then he refactored the graph traversal algorithm, replacing recursion with an iterator and adding a timeout circuit-break threshold. When the code was done, he ran it in the local sandbox. The first round of testing passed. Memory usage dropped by 62%, and cyclic nodes were successfully isolated. He let out a long breath, saved the version, committed it to the repository, and attached detailed change notes.

At 7:20 in the evening, the fix script was deployed to the testing environment. The latency curve on the monitoring dashboard dropped back down. He leaned against the back of the chair and rubbed at the space between his brows.

His phone lit up. A message from Su Man: “Permissions are live. The codebase is on GitLab, branch main. Tomorrow at ten in the morning, we’ll go over the DICOM parsing interface definitions. Don’t be late.”

Lin Chen replied: “Received.”

He opened his notebook of mistakes, pen tip hovering above the page. The paper was already filled with the previous 221 entries. On a new page he wrote:

“Item 222: Variables are opportunities; anchors are bottom lines. Technical debt has to be repaid, but not with your life. The demo environment is short-term delivery; the vertical platform is long-term positioning. To run both in parallel, time must be divided into blocks. Bottom line unchanged: don’t cross the red line, don’t gamble on luck. Next step: sync with Su Man tomorrow at ten; tonight, finish the DICOM parsing prototype. Put Xiaoman’s picture book away carefully. Roads are repaired one step at a time.”

He closed the notebook. Outside the window, night settled in, and one by one the lights in the hospital corridor came on. The ticking of the monitor, the tapping of the keyboard, the distant flow of traffic from the road outside wove together into a dense net. He knew tomorrow would be another new day. No miracles—only execution.

His phone vibrated again. This time it was the number for the nurses’ station.

He answered.

“Mr. Lin?” The nurse’s voice sounded urgent. “Xiaoman suddenly had a convulsion just now. It lasted about ten seconds. The doctor is already here and is administering a sedative. Would it be convenient for you to come over right away?”

Lin Chen’s fingers stopped on the keyboard. The code cursor on the screen was still blinking. He stood up. When his left foot hit the floor, the pain was sharp, but he did not pause.

“I’m coming now,” he said.