Rebirth: Starting from Lighting Up the Tech Tree

Chapter 111: Sky Dome Phase 4

Zuo Cheng put down his phone and glanced out the window.

It was already evening. The setting sun bathed the office in an orange-red hue, and Shen Yiming was still at his workstation, staring at the screen. The test data for federated learning had just finished running.

"Brother Cheng, what did they say from Blue Bay?" Shen Yiming asked without looking up.

"President Zhou wants to see us," Zuo Cheng said. "Tomorrow."

Shen Yiming's fingers stopped on the keyboard. He turned his head, his brows furrowing slightly.

"So urgent?"

Federated learning has just completed its practical testing, and normally there should be a week for summarizing and optimizing. But Zhou Henian's tone didn't sound like a routine report.

"Sky Dome Phase IV," Zuo Cheng said. "He said he needs us to do a technology demonstration."

Shen Yiming's brows furrowed even more. He didn't speak, but Zuo Cheng could tell what he was thinking.

Federated learning is a distributed training method where a model learns collaboratively across multiple devices, and the results are aggregated into a complete model. This technology has been validated in the Hangzhou smart city project with good results.

But Sky Dome Phase IV is different. Those are satellites.

"I'll go with you tomorrow," Shen Yiming said.

Zuo Cheng nodded.

---

The next morning, the sky over Hangzhou was overcast, as if veiled by a layer of gauze.

Zuo Cheng drove Shen Yiming and Fang Ze to the headquarters of Lanwan Communications. This 20-story glass building is a landmark in Hangzhou's high-tech zone, but from the outside, its gray exterior makes it look lifeless.

"Mr. Zhou is waiting for you on the twelfth floor." The receptionist led them to the elevator.

In the conference room, Zhou Henian was already seated in the main seat. He was wearing a gray jacket and looked a bit more haggard than when Zuo Cheng had last seen him. The wrinkles at the corners of his eyes seemed deeper.

"Sit down." Zhou Henian raised his hand.

Zuo Cheng sat down opposite him, while Shen Yiming and Fang Ze sat on either side.

"I've seen the test results for federated learning," Zhou Henian said bluntly. "Accuracy 93%, communication overhead reduced by 58%, not bad."

Zuo Cheng remained silent, waiting for him to continue.

Zhou Henian picked up a document from the table and pushed it in front of Zuo Cheng.

"But that's not why I called you here."

Zuo Cheng picked up the document. The first page was a project overview, with the title prominently displayed: Project Sky Dome, Phase Four.

"The tasks for Phase Four are different from the first three," Zhou Henian said, leaning back in his chair with his arms crossed over his chest. "In the first three phases, we solved the communication problem: how to connect satellites to the ground, and how to transmit data between satellites. You've done a very good job with that."

"What about Phase Four?" Zuo Cheng asked.

"The fourth phase aims to solve the problem of coordination," Zhou Henian's gaze sharpened. "It's not just simple communication, but enabling each satellite to 'think.'"

Shen Yiming's breathing visibly paused for a beat.

"To be specific," Zuo Cheng said calmly.

"The fourth phase of the plan will have 480 satellites in orbit, nearly four times that of the third phase," Zhou Henian said, standing up and walking to the window. "If we continue using the traditional centralized control method, the ground stations will become a bottleneck. Once a ground station malfunctions, the entire network will be paralyzed."

"So you want to..." Shen Yiming continued.

"Let the satellites decide how to communicate," Zhou Henian said, turning around. "Each satellite is equipped with an AI module that can sense the status of surrounding satellites and dynamically adjust its own transmission strategy. Just like a school of fish, it doesn't need a leader; each fish knows where to swim."

The meeting room was silent for a few seconds.

Fang Ze broke the silence: "This means each satellite needs to run an AI model. What's the power consumption budget?"

"5 watts," Zhou Henian said.

Fang Ze's expression froze.

What does 5 watts mean? A typical mobile phone processor has a peak power consumption of around 5 to 10 watts. But that's for mobile phones, running streamlined applications. AI models, especially deep learning models, require hundreds or even thousands of times more computing power than ordinary applications.

"Cambricon's MLU270," Shen Yiming said softly, "peak power consumption 30 watts."

"Six times the standard," Fang Ze added.

Zhou Henian looked at them without saying a word.

Zuo Cheng stared at the document in his hand. The core requirements for the Sky Dome Phase IV project were clearly stated: autonomous collaborative networking among satellites, each satellite possessing edge AI capabilities, and a power consumption budget of 5 watts.

This is a technological no-go zone. No commercial project in the world has ever achieved this.

"President Zhou," Zuo Cheng looked up, "was this request proposed by you, or...?"

"It was our suggestion," Zhou Henian interrupted him, "but it's not something Blue Bay could do on its own."

He walked back to his seat and sat down opposite Zuo Cheng.

"To tell you the truth," Zhou Henian's tone turned somber, "CEFC is also conducting similar research. Their approach uses traditional satellite technology, with a large amount of redundant hardware to ensure reliability. It's cumbersome, but reliable."

"And you?" Zuo Cheng asked.

"We want to take a different path," Zhou Henian said, looking directly into Zuo Cheng's eyes, "but this path will be difficult. If you feel you can't do it, I understand."

His meaning was clear: the technical plan for Sky Dome Phase IV has not yet been finalized. 402 could be the main developer, or it could not.

Zuo Cheng glanced down at the document. The technical specifications for Sky Dome Phase IV were right before him: 480 satellites, 5 watts of power consumption, and autonomous coordination between satellites.

This is not an easy challenge.

"Two weeks," Zuo Cheng said.

Zhou Henian didn't speak, but tapped his fingers twice on the table.

"We need a complete technical proposal within two weeks," Zuo Cheng said, looking up. "If the proposal is approved, we'll be the lead developer. If it's not, then forget it."

The meeting room was silent for five seconds.

"Two weeks is a bit short," Zhou Henian finally spoke. "I'll give you three weeks."

"No need," Zuo Cheng said calmly. "Two weeks will be enough."

Zhou Henian stared at him for three seconds, then suddenly leaned back and smiled.

"Okay." He stood up. "I'll be waiting for your technical proposal in two weeks."

As they shook hands, Zhou Henian lowered his voice: "Zuo Cheng, I hope you won't disappoint me."

Zuo Cheng's palms were slightly damp. He didn't answer.

As they left the Blue Bay Building, Shen Yiming approached him: "Brother Cheng, two weeks? The federal training just finished, and we haven't even decided on a direction yet."

"It will be decided," Zuo Cheng said.

He looked up at the sky. Gray clouds hung overhead, obscuring any blue hue.

402's technological capabilities are vastly different now. Federated learning has been successful, edge AI is available, and model compression is also in place. While the Tech Radar is still cooling down, all of these things combined should be sufficient.

should.

Fang Ze slammed his laptop onto the passenger seat and cursed, "Bullshit, running AI at 5 watts."

"Yes, it can be pulled apart," Zuo Cheng said.

"How do I pull it?"

"Think about it again."

Fang Ze didn't say anything more. Zuo Cheng was the kind of person who spoke the least at crucial moments.

Outside the car window, the Hangzhou skyline slowly receded.

Two weeks. Digging a hole for yourself. But if you don't dig it, you won't know how high you can jump.

He opened the system panel and glanced at it. 277 points, four branches and thirty-nine leaves on the tech tree, and the tech radar, currently cooling down, would have to wait another day and night.

enough?

Zuo Cheng didn't know. But he knew one thing: in two weeks, he had to come up with a proposal that would satisfy Zhou Henian.

This is the biggest challenge 402 has faced since its inception.