I have a very close friend He probably manages around 15 billion Probably able to achieve the top 10 quantified domestically Liang Wenfeng from Deepseek Previously, when mainly managing magic square quantification They also often go to communicate about certain things As for me, I can't waste such an amazing friend Regularly schedule a conversation with this friend It's not so much about chatting It's more like I'm going to learn something He wants me to join his team I said I haven't played enough yet The cryptocurrency circle is so fun Your things are too serious and old-fashioned It's meaningless. Wait until I have played enough or really go bankrupt You can also be considered as the last line of defense to help me withstand risks Hahaha Joking aside Let's talk about the main topic we discussed Quantify without bull or bear. In addition, after the AI era, the advantages of quantitative companies like them have been reduced because the threshold has been greatly lowered 1. Abstracting market volatility into variables such as price, trading volume, volatility, correlation, spread, liquidity, etc 2/Compressing variables into signals: whether there is a slight shift in the distribution of future returns after a certain market state occurs 3/Use mathematics to verify the advantage of probability: see if it has a positive expectation and can be stably established in a large number of repeated bets 4/Use violent testing to torture signals: changing cycles, changing markets, adding costs, adding slippage points, adding delays, running outside of samples, running extreme market trends. The parameter possibilities included in a sample test may be in the billions 5/Combine the low correlation signals that have survived into the system, and then use execution, risk control, and iteration to turn small edges into long-term compound interest Before AI reached its current level of maturity, the threshold for this thing was actually very high, with abstract fluctuations and variable compression into signals, all of which required at least a PhD in mathematics The verification of probability advantage and brute force testing require an extremely large system, and the computational resources required for data processing are also enormous. It is impossible for the public to build this thing, including further machine learning and the like But in the current state of AI, the mathematical threshold has been completely broken, and various computing power resources for brute force testing can be solved by buying tokens. The construction of this system's building AI can also be quickly set up After AI, the gap between people has actually become even greater Before the AI era, if you had ideas, insights, and knew how to organize this system, but you didn't have the corresponding resources and execution capabilities, then you were no different from ordinary people and couldn't do anything. Everyone is just an ordinary person Now it's different. With ideas, insights, and knowledge of how to organize this system, there's no need for any other resources or execution capabilities. If you leave these tasks to AI, then you're not just an ordinary person. Behind you, there's almost all of humanity's technology and various theoretical sciences supporting AI working At this point, in this field, you are standing shoulder to shoulder with the world's top people That being said OKX Onchain OS is considered the best on chain AI trading infrastructure in the industry Enabling complex transaction logic such as combination and filtering to run on the chain, while greatly improving the efficiency of a large number of basic on chain interactions npx skills add okx/onchainos-skills， The thing is here. If you're willing to study it, just tidy it up I am also using AI to build some quantitative trading tools, and I am still in the stage of playing. This is a strategy I have been playing with recently, which is to make a profit and run, and wait for the next excellent opening point to go up