Even if the strategic logic (trend following, statistical arbitrage) is the same, differences in AI application methods can completely dismantle performance correlation. Traditional quantitative trading is a "competition on the same track," whereas in the AI era, quantitative trading is a "contest across different dimensions." Under the gap of these dimensions, the results naturally vary greatly—one making money, the other losing money—a divergence situation.

Essentially, its AI application has formed a closed-loop ecosystem of "data - algorithms - computing power":
Data side: Integrates over 120 alternative data sources (including satellite, sentiment, on-chain data), which is 6 times that of traditional institutions;
Algorithm side: Uses a "Large Language Model (LLM) + Reinforcement Learning" dual-engine, rather than a single machine learning model;
Computing power side: Self-built GPU clusters, with computing capacity more than 10 times that of medium-sized institutions.
In contrast, most small and medium-sized institutions are still stuck in the fragmented application stage of "buy data + rent computing power + tune models"—this "ecosystem gap" is the ultimate reason for performance differentiation.

The performance data of the quantitative industry in 2025 directly confirms the impact of AI application differences:
Differences in AI application and strategy crowding
Time (veterans with over ten years of experience vs. newcomers), key areas (signal research / portfolio construction / trade execution), and methods