你刚刷到这条消息，本来准备顺手划走，但又怕自己错过了真正会影响下一步判断的那一点。
最容易做错的，是xbtlin / ai-berkshire；代价往往是如果只盯表面热闹，你很容易在错误方向上花掉时间、预算和注意力。；我先给一个保守判断：AI投研最值钱的能力是否决，而非发现。
My takeaway is simple: the most valuable capability in AI investing research is veto, not discovery. The interesting part of AI Berkshire is not that it helps you find more ideas. It tries to stop you from buying bad ones with 8 red lines.
The proof is in the decision format. The README forces three states: pass, gray zone, or fail. The checklist docs add a mirror test: if you cannot explain the business, defensibility, management, valuation, and downside in five sentences, do not buy.
That is the part I would copy first. A tool update is worth watching only if it changes your next decision, not if it just adds more surface area. For most teams, a reusable veto pattern is more valuable than another model that can produce longer analysis.
My boundary is narrow: this read comes only from the current main-branch README and checklist docs, not audited returns, a tagged release, or live portfolio data. So I would copy the decision pattern first, then ask for performance evidence. If you build AI for any high-stakes 工作流程（workflow）, share this with the person who still thinks the model's job is to generate more analysis. What are your own veto rules before a model gets to recommend anything?
#AIAgents #Investing #DecisionMaking #OpenSource
真正该讨论的是：xbtlin / ai-berkshire