Chapter 37: Acceptance of the Preliminary Research Project

The following morning.

A black sedan pulled up to the school gate. A man in an administrative jacket stepped out and stood by the car like a statue, watching the entrance. Fortunately, it was still early, and the area was nearly deserted, save for the security guard who kept a wary eye on the man.

After a short wait, Zhou Yun emerged from the gate. He glanced at the license plate to confirm it was his ride and walked toward the man.

"Good morning, Mr. Zhou. Director Zhao sent me to pick you up." The man clearly knew who Zhou Yun was; his tone was respectful as he opened the car door.

Zhou Yun, unaccustomed to such treatment, quickly thanked him and climbed inside. The drive took about half an hour before the car arrived in front of a large office building.

"We’ve arrived, Mr. Zhou."

Once the car came to a stop, Zhou Yun stepped out. Another staff member immediately approached and exchanged a few words with him before guiding him into the building and up to the thirteenth floor. The staff member led him into the corridor to the right of the elevator.

"Please wait a moment, Mr. Zhou. This is a routine security procedure," the guide explained in a low voice. "Please place all personal electronic communication devices—including mobile phones, tablets, and smartwatches—as well as any recording equipment, into this basket."

A security officer produced a plastic basket from behind and placed it on the table. Zhou Yun nodded, turned off his only device—his phone—and handed it over.

"Please step onto the platform after passing through the security gate," another guard instructed, pointing to a platform behind the metal detector.

The green indicator light on the frame lit up; there was no alarm, signaling that Zhou Yun was carrying no prohibited items. The guard motioned for him to raise his arms, gave him a thorough pat-down to confirm all was well, and then stepped aside.

"All right, Mr. Zhou. We will keep your personal belongings safe, and they will be returned to you when you leave. Thank you for your cooperation."

The staff member who had accompanied him through security then led him into a conference room.

Seeing Zhou Yun enter, Director Zhao—still dressed in his navy blue suit—stood up. "Mr. Zhou, thank you for coming. When I received the call yesterday, I could hardly believe it. I never expected such significant progress in such a short time. Truly, the younger generation is formidable."

"You are too kind, Director Zhao."

"Come, please, take a seat," the Director said, pointing to an empty chair reserved for him. "Everyone, let me introduce Mr. Zhou Yun. The model on our server is his work."

The middle-aged man beside Director Zhao had been observing Zhou Yun since he walked through the door. Though he was stunned by how young Zhou Yun was, he maintained a composed exterior.

"Mr. Zhou, let me introduce you. This is Director Li from our Technical Department."

Director Zhao briefly introduced the heads of several other departments as well, though the meeting was focused on technical matters, and the others were merely there in a supporting capacity. Director Li nodded at Zhou Yun. He had already reviewed the model and the experimental logs; since his department had provided the server, they had full access. When Director Zhao had called him in to review the logs, he had been speechless for a long time before finally uttering a single phrase: "Incredible genius."

Director Li had seven or eight years of experience in financial AI modeling, yet he found the architecture Zhou Yun had built deeply impressive. While the core foundation still relied on the transformer-based attention mechanism, the rest was vastly different from existing models. He was here not only for the formal acceptance test but also to discuss technical details; they needed a deep understanding of the system before they could safely deploy it.

After the brief introductions, Director Zhao returned to his seat. "Mr. Zhou, time is precious, so let’s skip the formalities. The purpose of this meeting is clear: to conduct the project acceptance test. Director Li will ask you some technical questions, and finally, we will verify the results against the benchmarks we established. If it passes, this phase of the project is complete. Do you have any questions?"

"No, none at all."

"Good. Let’s begin. Director Li, the floor is yours."

"Very well, Mr. Zhou," Director Li began. "After your call yesterday, we conducted internal and live-market tests. The results were excellent, meeting or even exceeding our projected targets. Additionally, I’ve read the technical documentation you uploaded to the server. However, due to time constraints, I have a few questions I’d like to clarify in person.

"There are two main issues: current models focus on a single stock, but the market consists of many, each with different characteristics. How do you intend to handle the interdependencies between various stocks, and how can you ensure the model maintains its performance when processing data volumes far exceeding the current scale? Secondly, if we were to incorporate the entire US stock market into your framework, what are your estimates regarding the required computing resources?"

Zhou Yun answered without hesitation. "I have already considered both points. To address your first question: I view the entire market as a dynamic, evolving relational network. You can think of it as a graph in data structures, where each stock is a node, and the connections between them are the edges representing the strength of their correlation. These correlations are not static; they evolve over time. The model automatically calculates these correlation values through learning, thereby constructing a real-time, dynamic, evolving network graph.

"As for your second question, it depends on your hardware infrastructure. If the hardware is appropriately optimized, and given the low resource requirements of the model itself, I estimate it would require approximately five thousand H100 units. After all, we are talking about the entire US stock market; the computational power needed is significant."

Director Li nodded, though he still had reservations about the dynamic network graph.

The discussion lasted for over two hours. By the end, Director Li was completely convinced. If he had harbored any doubts about Zhou Yun’s abilities at the start, they had vanished. Throughout the debate, he had witnessed the depth and breadth of Zhou Yun's knowledge.

"I have one last question—if you can call it that. What is the name of this model? The code only bears the initials 'QN'."

Zhou Yun smiled. "OracleNet. QN is the abbreviation, but I prefer its Chinese name: Tianji Network."