A Reddit user tested whether artificial intelligence could replicate Warren Buffett's investing strategy by feeding Anthropic's Claude Opus 4.6 model 48 years of Berkshire Hathaway's investor letters. The experiment aimed to extract Buffett's core principles and apply them to stock selection in today's market. The AI analyzed financial data from 50 companies, including Berkshire's actual holdings, value candidates, and speculative stocks.

The AI-generated portfolio achieved approximately 134% returns since 2020 in backtesting, outperforming both the S&P 500 and an equal-weight Buffett portfolio. However, Buffett's actual weighted portfolio, boosted by Berkshire's massive Apple position, still led with about 165% returns. The experiment demonstrates AI's potential in investment analysis while highlighting the limitations of algorithmic approaches compared to human expertise and concentrated positions.

Among the AI's top picks, Alphabet emerged as the largest holding, up 71% over the past year. Coinbase represented the most surprising selection, given Buffett's well-known skepticism toward cryptocurrency. The AI apparently identified Coinbase as a value-style investment based on its 39% profit margin and low debt profile, illustrating how algorithms might interpret financial metrics differently than traditional value investors.