Press the Enter key.

The system indicator light came on, and the waveform on the main control screen began to climb. Chen Hao didn't stare at the time numbers as before, but instead focused his attention on the initial segment of the voltage curve. He knew that the real test wasn't in the last few seconds, but in the first five seconds—where the jagged fluctuations left over from the last experiment were hidden.

“Lock the parameters,” he said. “This time, only adjust the antenna angle; freeze everything else.”

Susan nodded, her finger swiping across the virtual interface to confirm that the modulation algorithm, power supply strategy, and filtering channels were all locked. She pressed Enter to submit the current configuration.

Nana's voice rang out: "V2.1a test task has been loaded, variables: azimuth ±0.3 degree range, other parameters are sealed."

Carl stood up from beside the equipment cabinet, holding a thermometer. "The heat dissipation module has been installed, the temporary fan is working normally, and the core component temperature is 1.2 degrees lower than the previous round."

“Okay.” Chen Hao walked to the whiteboard, picked up a marker, and circled the “Antenna Angle Fine Adjustment” option. “We’re not looking for the strongest signal, we’re just looking for the most stable point.”

The first group of tests has started.

The antenna angle was increased by 0.3 degrees.

For the first twenty seconds after the signal was transmitted, everything was normal. The receiver then reported a significant increase in signal strength, with the coverage radius rapidly expanding to 7.6 kilometers. Susan stared at the comparison chart, her brow furrowed slightly.

“It’s too fast,” she said. “The ascent slope is even steeper than V2.0.”

The moment the words left his mouth, forty seconds later, the main control screen flickered.

A phase lockout alarm has been triggered.

The bit error rate jumped to 0.4% instantly. Although the system automatically switched to the backup demodulation mode and recovered within 0.6 seconds, this fluctuation exceeded the preset threshold.

“No,” Karl shook his head. “Too sensitive. It will break down at the slightest disturbance.”

“It’s not a signal strength issue.” Chen Hao looked at the log playback. “It’s that the matching rhythm is wrong, like running too fast and missing the beat.”

In the second test, the angle was adjusted back to the baseline value.

This time the operation was smooth, with the bit error rate consistently kept below 0.03%, but the signal peak did not replicate the previous level. By the end of the 83rd second, the coverage area had only reached 7.2 kilometers.

“It’s stable,” Susan said, “but it’s regressed.”

"This means the optimal solution isn't at either end," Chen Hao said, leaning against the table. "There's still space in the middle."

The third test group was set to an offset of +0.1 degrees.

Nana switched to isolation mode to ensure there was no interference from external variables. She pulled up a stability heatmap, where a dark blue area was concentrated around 0.1 degrees, representing the system's strongest anti-interference capability within that range.

“The data supports this value,” she said. “Historical records show that in three simulated disturbance tests, the 0.1-degree configuration was the fastest to return to a stable state.”

“That’s it.” Chen Hao wrote the new number on the whiteboard: **V2.1a - Stable Configuration**.

Test restart.

After the countdown ended, the system entered the long-term operation phase. The voltage curve was smooth and without jagged edges for the first ten seconds; at the thirty-fifth second, an external interference signal reappeared, with a frequency of 915 MHz, lasting for 0.8 seconds.

Nana's shielding compensation protocol responded immediately, activating dual-path filtering and reducing power by 12%. The bit error rate only rose to 0.07% and fell back within 0.3 seconds.

"It held up." Carl checked the image quality on the receiver. "The picture is clear, and the voice packet is intact."

At the 60th second, the signal strength stabilized at +73dBm, and the coverage radius exceeded 8.5 kilometers.

At the 83rd second, the system was still online.

"Did it work?" Susan asked.

"It's done." Chen Hao nodded. "This time, it wasn't luck."

He turned to Nana and said, "Generate a report and tagged and archive all the operations performed this time."

"Processing is underway." Nana's robotic arm connected to the host port and began categorizing the data stream, "including 327 indicators such as environmental parameters, component status changes, and system response latency."

"We need to refine this further," Chen Hao said. "Every change, no matter how small, must be recorded. Who made the change, when it was made, and what the basis was—all of that must be written down."

Susan opened the editing interface: "I'll write a template called 'Communication System Optimization Operation Guide,' so even beginners can follow it."

"Don't make the name too formal," Chen Hao laughed. "Just call it 'How to Change It So It Won't Explode'."

“Then we need to add a subtitle,” Carl chimed in, “'Ten Iron Rules to Learn from Failure'.”

Nana didn't laugh, but the indicator light in her eyes flashed slightly, as if she had tacitly agreed with this statement.

After the report is generated, the page automatically redirects to the version change log. V2.1a is marked as the first "reproducible stable version" with a risk warning: Do not adjust more than two variables at the same time, otherwise the test will be considered invalid.

"From now on, anyone who wants to take action must read the guidelines first," Chen Hao said, pointing to the screen. "It can only be carried out after three people have confirmed it."

"You want to implement an approval system?" Susan raised an eyebrow.

“It’s not about approval,” he said. “It’s to prevent someone from coming up and resetting all the parameters, and then saying, ‘I’ve tried it, it doesn’t work.’”

"Has anyone actually done that?" Carl asked.

“That’s exactly what happened to me.” Chen Hao shrugged. “Last week, I pushed the gain to the limit and burned out the preamplifier.”

"You remember?" Susan asked in surprise.

“I can’t forget it.” He scratched the back of his head. “That day, Nana talked about it for a full forty minutes, and she didn’t repeat a single word.”

Nana calmly said, "There is a standard fault review process in the knowledge base. I suggest playing the recording before each operation."

"Stop." Chen Hao raised his hand. "If you keep doing this, I'll unplug your power."

The laboratory fell silent for a moment.

Then Susan laughed, Carl snorted, and even Nana, who had been keeping a straight face, slightly tightened her eyes.

After the third test, they decided to run another round of verification.

This time, Susan took the lead and strictly followed the guidelines. Parameters were locked, variables were unique, and each step of the process was checked off. The countdown started, the system ran smoothly, and after ninety seconds, the signal was not interrupted.

“I can hand in the results now.” Carl tucked the printed waveform into a folder. “At least I can get a good night’s sleep.”

"Don't let up." Chen Hao looked at the screen. "This is just the first step. We know how to tune one stable system now, but we don't know how to make ten of them stable yet."

"You want to copy it in bulk?" Susan asked.

“Not only,” he said. “I want others to be able to figure it out too. We can’t be the only ones who know which screw is loose and will cause problems.”

Nana began organizing the backend logs, preparing to generate the V1.0 operation guide. The document structure was clear: the first part was a list of parameters, the second part was a solution for common exceptions, and the third part was a version upgrade path diagram.

She added a note at the end:

This guide is based on limited sample testing and applies to the current hardware configuration and environmental load. If critical components are replaced, please reassess the compatibility.

"Add one more thing," Chen Hao leaned closer, "—Don't trust your intuition, trust the data."

Nana added:

Intuition can lead to system crashes; data, at least, can tell you why a system crashed.

“That’s harsh,” Carl said.

“That’s true,” Nana said.

Over the next two hours, they ran four more rounds of testing, verifying repeatability at different time intervals. Each time, they followed the established procedures and recorded everything completely. V2.1a consistently performed well, with the error range kept within 1%.

“This is science.” Chen Hao leaned back in his chair. “Before, we would shout ‘Awesome!’ if we hit the right spot, and say we were unlucky if we hit the wrong spot. Now it’s different. Now we know where the awesomeness lies, and we also know who to investigate when we’re unlucky.”

Susan closed her notebook: "We can let someone else try tomorrow."

“Who?” Carl asked.

“The people on Li Yan’s shift,” she said. “Have them follow the instructions and see if they can run it independently once.”

“What if they mess it up?” Carl said.

“That means the guidelines aren’t clear enough,” Chen Hao said. “Or we haven’t really understood this yet.”

Nana has completed the final report output. The cover title is:

Communication System Optimization Operation Guidelines V1.0

There is a line of small print below:

Dedicated to all those who try to debug equipment by guesswork.

"Don't print this sentence," Chen Hao said.

"It's been saved as an internal comment," Nana said. "Viewable only by developers."

As night deepened, the laboratory lights remained on.

Chen Hao stood in front of the control panel, a marker in his hand, his gaze fixed on the flowchart on the whiteboard. He hadn't worked for such a long time continuously in a long time, but he didn't feel tired. When he failed, he wanted to lie down; when he succeeded, he felt strangely alert.

Susan sat in the secondary control seat, her fingers hovering above the keyboard, waiting to input the next round of test commands.

Carl crouched beside the equipment cabinet, checking the circuit temperatures. He inserted a probe into the bottom of the chassis and scanned each interface one by one with a temperature gun. He straightened up and nodded, indicating, "The heat dissipation is normal."

Nana's robotic arm is still at the host port, and the data archiving progress bar has reached 98%. Her optical lens is flashing, indicating that final verification is underway.

Chen Hao turned to look at the screen; the waveform stretched smoothly, with the green line pointing straight ahead.

"Prepare for the fourth experiment," he said. "This time, we'll try to reduce power consumption and see if we can save some electricity while maintaining stability."

"Another change?" Susan asked.

“It’s not about changing,” he said. “It’s about continuing.”

He picked up a marker and added a new line to the whiteboard:

**Goal: Stability + High Efficiency**

The pen tip glided across the whiteboard, making a short, sharp friction sound.