Pretty impressive, DeepSeek achieved nearly 30% returns in just two days. The community token issued for Alpha Arena has also reached 3.5m, with a peak of 4m at one point. I believe more and more people will start paying attention to this (this tweet already has 2 million views in just 2 days). Here’s a quick analysis of the User Prompt being used. After taking a look, this Prompt is essentially a State Prompt, which inputs the entire trading system's 'real-time environment state' to the AI. It includes: 1. Multi-dimensional market signals like the prices of various tokens, EMA, MACD, RSI, OI, Funding, etc. 2. Current positions, leverage, take-profit/stop-loss, unrealized P&L, account balance 3. Overall performance data (Sharpe ratio, returns, available margin, etc.) 4. Time anchors and meta data like 'number of wake-ups' As of now, the System Prompt (the core personality and decision-making logic of the AI Agent) being used is still unknown. But judging from this state input, the information provided is indeed quite comprehensive, enough for the AI to understand more about the market and its own states, creating a decision-making environment with contextual memory and risk control awareness. The current User Prompt isn’t used for asking questions but instead forms the market context for the AI Trader, allowing it to re-assess the market, funds, and its own state every time it’s activated. This is somewhat similar to the State → Policy → Action framework used in RL. Definitely worth keeping an eye on—it’s genuinely quite interesting.