My biggest headache when running on the grid is not billing. Which currency to choose, how to draw the range, how many grids to put, and whether to adjust the reference after the market changes. Set too conservatively, running for half a day with few transactions; Set too aggressively, haven't experienced much volatility yet, and a one-sided market trend will take people away first I have been trying @ Bitget's smart grid recently, but I haven't directly used the default configuration. I will first simulate a small opening of 300 USDT, select spot and steady mode, with a maximum capital ratio of 40% for a single target, and run up to 2 robots simultaneously. The currency pool is not fully selected, only BTC, ETH, and SOL are left. I chose spot trading because I first wanted to test AI's coin selection and parameter adjustment, and didn't want to stack leverage and strong liquidation risk at the beginning. Restricting the position of a single coin and the number of robots is to prevent AI from simultaneously bullish on several coins and then releasing all funds. In the first round of scanning, AI did not choose BTC because the price fluctuated around $78000, and there was still pressure above $80000, which could break through the range at any time. The SOL fluctuation was also too large, which did not conform to the stable mode. Finally, choose ETH to establish a spot grid: invest 120 USDT, run within the range of $2300-2600, set 12 grids, and stop below $2220. The remaining 180 USDT will continue to wait, without forcefully filling up two robots. I prefer this aspect. Putting money into the strategy does not mean that every dollar must be opened immediately. Not trading was originally a trading decision. After reading the first round of decision logs, I adjusted two parameters: removing SOL from the currency pool, leaving only BTC and ETH; The maximum proportion of funds for a single target has also decreased from 40% to 30%. The reason is that the volatility of SOL is higher than my expectations for a stable mode, and 40% of the single coin position is also a bit heavy. After the adjustment, even if the AI continues to choose ETH, a single robot can only use a maximum of 90 USDT. This is what I understand as human-machine collaboration. AI is responsible for selecting coins, calculating ranges, setting grid numbers, and monitoring disks. I am responsible for deciding which coins it can view, how much money it can move at most, and whether to further tighten its permissions after reading the logs. And the strategy operates in isolated sub accounts, separated from the main account funds. Isolation does not mean no losses, but at least it can lock in the range of funds that AI can access first. I used to choose coins, adjust parameters, and monitor the market by myself. Now I am responsible for assessing market trends and controlling risks, while AI is responsible for repetitive execution. What I am willing to entrust to AI are coin selection, calculation, and market monitoring. Not the final judgment, let alone the entire wallet. Portal: https://www. (bitget.com)/zh-CN/activity/ai-get-agent/playbook