wtf is ai-berkshire?

AI Berkshire is a collection of skills for Claude Code and Codex that turns an AI coding agent into a stock research team. It packages the investing methods of four well known value investors, Warren Buffett, Charlie Munger, Duan Yongping, and Li Lu, into structured, repeatable research workflows a single person can run. The author argues that simply asking an AI chatbot to analyze a stock produces vague, wishy washy answers that hedge every point and are not useful for making a decision. AI Berkshire instead forces a clear verdict, such as pass, fail, or gray area, along with specific price ranges and position sizing suggestions. Its flagship feature runs four separate AI agents in parallel, each one reasoning from a different investor's point of view, so their conclusions can genuinely disagree with each other rather than blending into one bland summary. The README shows a real example where the four perspectives on a company reach different scores and different conclusions, which the author treats as useful tension rather than a bug. The framework also builds in checks meant to catch overconfident or sloppy reasoning: rating how much reliable information is actually available before trusting a conclusion, deliberately searching for reasons a company might be a bad investment, a short list of automatic disqualifiers such as management integrity problems, and financial calculations done with exact decimal math instead of rounded estimates, cross checked against at least two independent data sources. The repository ships about 20 separate skills covering deep company research, earnings report analysis, industry wide screening, portfolio management, and tracking whether an original investment thesis still holds over time. It works with both the Claude Code and OpenAI Codex command line tools, and the author reports two consecutive years of real, verified investment returns using this framework, alongside the caveat that past performance does not guarantee future results.