开源 5806 个量化策略，用 Claude Code 搭建、Jev 做实时决策
作者给出完整工作流：策略文件定义逻辑，Claude Code 搭数据管道和执行引擎，Jev 接实时行情在 100ms 内评估买卖持，再由独立风控层把关下单。
this repo just open-sourced 5,806 quant trading strategies for $0. not prompts. not strategy names. the actual code, parameters and math behind spot, perps, grid, statistical arbitrage and market-making systems. the interesting part is what happens when you connect the vault to AI: → the strategy file defines the trading logic → Claude Code builds the system around it → Jev evaluates the live market → a separate risk layer controls execution here’s the workflow: 1. clone the vault and choose one of the 5,806 strategies. 2. open the strategy file. the entries, exits, parameters and underlying math are already inside. 3. give the file to Claude Code and use it as the source of truth for the build. 4. let the agent create the data pipeline, execution engine, monitoring and risk controls. 5. connect Jev to the live market feed so it can evaluate buy, sell or hold decisions in under 100ms. 6. convert its confidence score into position size using liquidity, volatility and your predefined limits. 7. pass every order through a hard risk layer before anything reaches the market. 8. replay historical data and run the system in paper mode before enabling live execution. the useful part: each component has one specific job. Claude handles the deeper engineering and improves the system from previous trades. Jev reacts to the market while the setup is still relevant. one model builds and reviews. the other evaluates and reacts. and a strategy that used to sit inside a file can now be tested, monitored and improved 24/7. the full 5,806-strategy vault and codebase are in the original post below. save this for your next trading agent build.
这条是引用着另一条写的：x.com/sector_sol/status/2093769463065702779