A 20-year market-beating investment framework lived in one person's head and a book. WolfWorks codified the knowledge, making it accessible to anyone.

Carl Wolf is not a professional investor. No Bloomberg terminal, no team of analysts, no institutional backing - just a disciplined, rules-based framework for finding growth stocks at reasonable prices, developed and refined over decades of market cycles. Over 20 years, his portfolio averaged 22% annual returns, versus 11% for the S&P 500 and 16% for the NASDAQ - through the dot-com crash, the 2008 financial crisis, and the COVID collapse. His edge was never information. It was discipline. In 2026, he published that framework in a book: A Wolf Among the Bulls and Bears: How An Individual Investor Can Beat the Market.

For decades, Carl Wolf has been the person his family and close friends turn to with a stock question. His framework is detailed, nuanced, and deeply practical: 44 rules, 5 core criteria, a proprietary valuation model, conviction levels, and position sizing logic. It is not something anyone can hold in their head during a live stock analysis, even people who have read the book.

The book preserves the framework. But a book is static. It cannot run a company through Carl's criteria. It cannot tell you whether today's price clears his valuation model, what his conviction level would be, or what Carl himself would do. For the people who wanted to apply his methodology, the options were: call Carl, try to remember what the book said, or go without.

None of those scaled, and the framework deserved better than that.

The instinct with a project like this would have been to build a financial data tool first. Connect to a market data feed, build a screener, let users run filters. That would have missed the point entirely.

A screener can tell you a stock's PEG ratio. Carl can tell you whether that PEG ratio, in the context of this business's secular growth story, at this point in the interest rate cycle, with this analyst coverage profile, is a reason to buy or a reason to wait. The difference between those two things is judgment. And judgment cannot be retrieved. It has to be encoded.

The insight was that the right foundation for Wolf's Way was not data architecture: it was knowledge architecture. Carl's complete framework needed to be encoded in a way that an AI could not just retrieve but genuinely apply: working through the criteria in the right order, applying the right thresholds, surfacing the right rules, and reaching a conviction level that Carl himself would recognize as his own.

That required a different kind of design work. Encoding Carl's framework meant capturing not just his rules but his reasoning: the mechanical thought process, the accumulated expertise, and the years of lived market experience that inform how he weighs criteria, handles edge cases, and knows when to make exceptions. It meant building a valuation model that produces a specific fair value estimate from live market data. The result behaves less like a search tool and more like a conversation with the person who built the framework.

WolfWorks began with the knowledge architecture. Before any platform was designed or any line of code was written, the work was to understand Carl's framework completely, not as a set of rules to retrieve, but as a system of judgment to encode. That meant working through all 44 Wolf Rules, the 5 core GARP criteria, the Wolf Fed Model valuation formula, conviction levels, position sizing logic, and the nuanced application guidance that makes the difference between a rule and a judgment call.

With that foundation in place, WolfWorks built Wolf's Way: a production web application where users interact with Carl through a conversational interface. A user names a company or ticker, and Carl runs it through his complete framework: gathering live market data, scoring each criterion, producing a fair value estimate, assigning a conviction level, and delivering a bottom line in Carl's own voice. Each assessment covers 11 structured sections, from business overview through covered call opportunity, with every financial figure sourced and cited inline.

What makes that output possible is how the system's components work together. The knowledge framework - Carl's rules, criteria, valuation model, and judgment - is encoded at the foundation. The book's content is chunked and indexed so that the relevant passages are retrieved and present in every assessment and every conceptual question. Carl's voice, boundaries, and domain are defined precisely enough that the system knows what Carl would engage with, what he would redirect, and how he would say it either way. Claude's reasoning capability operates within all of that context simultaneously, not retrieving answers, but applying Carl's framework to new information the way Carl would. The result is something that could not be achieved by any one of those components alone.