Shanghai, Lujiazui. This was one of the swiftest estuaries of capital flow in the East, even globally. Skyscrapers rose like a forest of cold metal, piercing the humid morning mist that carried the salty scent of the Huangpu River. Glass curtain walls reflected the rising sun, as well as the ceaseless stream of traffic below, composing a cityscape brimming with futurism and oppressive energy. High up within this forest, in a room bearing no signage and requiring multiple layers of biometric verification for entry, the air itself seemed to have solidified.
This was Mozi's "battlefield." More than a trading room, it resembled a command center from a science‑fiction film, or an ultra‑clean laboratory. No clamor of ringing telephones, no traders shouting themselves hoarse, not even many people—just him, and the immense screens that covered three walls surrounding him.
On those screens, there was no microscopic physical world as Xiuxiu confronted, nor the abstract mathematical universe that Yue'er immersed herself in. Instead, they displayed the real‑time pulse of global capital markets—countless leaping numbers, undulating curves, and flickering color blocks. The US Dollar Index, government bond yields, commodity prices, and especially that eternally alluring yet perilous gleam of yellow—international spot gold.
Today, the gold market was experiencing a classic "oscillation." No clear direction, no sustained trend; the price behaved like a beast trapped in a cage, leaping up and down within a narrow band of less than twenty dollars, erratic and unpredictable. One moment, spurred by some geopolitical news, it would surge in a pulse‑like spike; the next, it could plummet just as rapidly from a large, unidentified sell order, leaving behind long upper or lower shadows like anomalous fluctuations on an electrocardiogram. For traditional traders relying on intuition and macro judgment, such a market was torturous—mere "noise," a chaos without order.
But in Mozi's eyes, there was no irritation, no confusion—only a near‑icy focus. On the main screen before him, a clean yet complex interface was running, its title displayed in three calm Song‑type characters: [Oscillation Model].
To him, the market was never a casino, but an immensely complex dynamical system, replete with nonlinearities, feedback loops, and emergent behavior. His weapons were not insider information, nor some mystical "feel for the market"; they were quantitative models built upon rigorous mathematics and computer science. When the market showed a trend, he deployed the [Trend Model], like a vessel sailing downstream, harnessing the current's force. And in the present disorderly, fluctuating "oscillatory market," the [Oscillation Model] was his Noah's Ark, designed to capture those minute yet certain profit ripples from the chaotic waves.
One of the core engines of this model was an optimization algorithm called **gradient descent**.
Mozi lifted the now‑lukewarm espresso beside him and took a sip; the bitter liquid sharpened his concentration. Watching the gold price curve zigzag back and forth on the screen, his mind conjured a vivid metaphor—**a blind man descending a mountain**.
Imagine a blind man standing on an undulating, mist‑shrouded hillside. His goal is to find the lowest point on the slope (corresponding to the model's **objective function**—here, the error between predicted price and actual price, or, in other words, the risk of trading loss). He cannot see the entire mountain, does not know in which direction the lowest point lies, nor how far away it is.
His sole reliance is his walking stick (**computing the gradient**). He carefully probes the slope of the ground around him with the stick (**gradient direction**). He will discover that in a certain direction the ground slopes downward (**negative gradient direction**), meaning that moving that way might lower his altitude.
Thus, he tentatively takes a step in that direction (**iterative update**). The size of this step is crucial; in the algorithm, it is controlled by the parameter **"learning rate."** A step too large (learning rate too high) might cause him to overshoot the lowest point, even leap onto the opposite hillside, leading to failure to converge, or even divergence; a step too small (learning rate too low) might leave him lingering near the lowest point for a long time, inefficient, or cause him to become stuck in some small depression (**local optimum**) and mistakenly believe he has reached the bottom, missing the true global minimum.
After taking that step, he probes his new position again with the stick. Based on the new slope, he adjusts his direction and takes another step. And so the cycle repeats—step by step, through feedback from local information, continuously adjusting, continuously approaching that lowest point he cannot directly "see."
In Mozi's [Oscillation Model], the "blind man" is the algorithm itself. The "shape of the hillside" is molded by numerous variables: the market's current volatility, historical price sequences, correlations with other assets, and even "market sentiment factors" extracted from news and social media and quantified through natural language processing. The model is "trained" on vast amounts of historical data, essentially learning how to "descend the mountain" more efficiently and robustly in this dynamically changing "hillside"—that is, to find the set of trading parameters (such as entry thresholds, position size, stop‑loss and take‑profit points) that minimize prediction error (or, equivalently, capture those high‑probability, minuscule price‑reversion opportunities while controlling risk).
On the screen, lines of code scrolled; parameters fine‑tuned themselves automatically. The model was calculating the optimal "stepping" strategy based on the current market's "slope." It might judge that, within this volatility band, when the price touches the upper rail, the "downhill" path of going short was more favorable; when the price falls back to the lower rail, the "downhill" path of going long was better. It would then issue calm, emotion‑free instructions.
Mozi did not need to stare at the tick‑by‑tick chart, nor did he feel a rush of adrenaline from a sudden spike, nor did his heart race from a rapid plunge. His work was that of a model builder, monitor, and guardian. He needed to ensure the "blind man's" "sensory system" (data input) was accurate and reliable, that his "walking stick" (gradient calculation) was sensitive, and that his "step size" (learning rate) was suited to the current mountainous terrain. He had to guard against abrupt, drastic changes in the "hillside's" shape (market‑regime shifts, such as a sudden transition from oscillatory to trending), and in such circumstances, decisively switch models or intervene forcefully, preventing the "blind man" from losing his way in unfamiliar terrain.
This kind of calm decision‑making, rooted in algorithms and data, stood in stark contrast to the greed and fear that saturated the market. Mozi knew well that the market was an amplifier of the collective emotions of countless participants, a testing ground for human frailty. He had personally witnessed too many cases where money earned through luck was eventually lost through lack of skill. He had also seen so‑called "stock gods" completely wiped out by the market's cyclical fluctuations.
He held the deepest reverence for the market. This reverence was not blind worship, but an acknowledgment of its complexity and inherent unpredictability. He knew that no model was omnipotent—black swan events always arrived unannounced. His models were merely attempting to build relatively sturdy sandcastles on the beach of probability, striving to endure a little longer under the wash of the tides (market fluctuations), to capture that tiny, yet cumulatively substantial, "edge."
This reverence stemmed from a grander ideal buried deep within him. He was no mere profit‑seeker. His earlier studies and work had shown him how international financial capital could, with ease, sway a nation's economy, harvesting the wealth created by the hard work of its people. He saw the immense, almost blind power of capital—how it could drive innovation and prosperity, yet also create crises and suffering.
A conviction gradually crystallized and grew firm within him—**to use capital to restrain capital**.
He could not change the profit‑driven nature of capital, but he could try to harness it, steer it, employing a more advanced, reason‑and‑algorithm‑based force of capital to counter those primitive, destructive speculative flows. He dreamed of constructing a more stable, more efficient financial system—one that truly served the real economy and humanity's long‑term well‑being. His trading was not merely about profit; it was a practice, an exploration and foundation‑laying for an ideal financial order. The capital he accumulated would be the ammunition for greater aspirations in the future.
On the screen, the [Oscillation Model] quietly executed several trades. When the gold price touched the preset upper boundary of the range, it established a small short position; when the price retreated, hitting the preset profit target, it automatically closed the position. The entire process was calm, swift—the profits slender, but as precise as clockwork. These minuscule gains, amidst the vast undulations of capital, were like scooping a single spoonful of water from a raging sea.
Mozi glanced at the performance report automatically generated in the account backend. Sharpe ratio, maximum drawdown, win rate… a series of cold metrics measuring the "blind man's" efficiency on this "descent." He gave a slight nod—satisfied enough. The model was performing well under the current market conditions.
He stood up and walked to the enormous floor‑to‑ceiling window. Below, the Huangpu River lay like a grey‑yellow ribbon; the historic buildings of the Bund and the skyscrapers of Lujiazui faced each other across the water, silently narrating the shifts of history and the power of capital. His gaze traveled beyond this prosperity, toward a more distant, unseen place.
He recalled a brief news item he had seen a few days earlier, about a domestic R&D team making progress in DUV light‑source technology for lithography machines. He thought of Yue'er in Princeton, discussing mathematical certainty with him in the deep of night. They—each on different battlefields, with different methods—were all wrestling with the world's complexity and disorder. Xiuxiu sought to harness light, carving out crystals of order; Yue'er sought to unveil the universe's underlying order through the pure logic of mathematics; and he, Mozi, tried to find those fleeting, quantifiable fragments of order within the seemingly random, emotion‑driven tides of capital.
Their paths appeared divergent, yet in their pursuit of "order" and "certainty," they resonated in a curious harmony.
Capital, in a sense, was also a kind of "light." It flowed, it empowered, it shaped reality. But like untamed light, it could scatter, burn, be misused. His work was to attempt crafting a precise system of "mirrors" and "lenses"—his quantitative models—for this "capital‑light," so it could be focused more accurately, utilized more efficiently, to illuminate places that truly needed illumination—such as the hard‑technology fields Xiuxiu engaged in, which would determine a nation's future competitiveness.
This thought sent a surge of warmth through him, alleviating the mental fatigue accumulated from long hours facing the screens. His ideal was no mere castle in the sky. It was built upon a profound understanding of market principles, skilled application of mathematical tools, and this day‑after‑day, monotonous practice of seeking tiny certainties amidst the market's tempestuous waves.
He returned to his seat. The curves on the screen still danced erratically. The chaotic nature of the market remained unchanged. Yet his [Oscillation Model] continued to run steadily; the "blind man" still tirelessly, step by step, explored and advanced on the mathematical "hillside" he had constructed.
Order was not the world's default state, but a precious thing that must be diligently discovered, constructed, and maintained. In this digital labyrinth of information and capital, Mozi—this modern "architect of order"—was carefully, brick by brick, laying the foundations of his ideal edifice with his code as the cornerstone. He knew it would not be easy; the road ahead would be fraught with unknown risks and challenges. But at this moment, listening to the low, steady hum of the server cluster and watching the algorithms execute his will with precision, he felt a quiet, creative, and controlling strength.
This was his battlefield, his mission, the way he chose to confront chaos.