Using AI to Invest in AI: Building and Learning Together
Named after Sherlock Holmes's enigmatic elder brother—a man of superior analytical abilities who preferred to orchestrate from behind the scenes—the Mycroft framework represents an educational experiment in AI-powered investment intelligence. With its tagline "Using AI to Invest in AI," this open-source platform explores how a ecosystem of specialized agents might work in concert to analyze the rapidly evolving artificial intelligence sector while implementing disciplined investment strategies. As an experimental project, Mycroft emphasizes learning by building, inviting contributors to discover what approaches actually work in practice.
At the foundation of the Mycroft experiment lies its analytical division—agents designed to test different methods of gathering, processing, and interpreting vast amounts of information about AI companies and their competitive landscape.
Research Agents function as experimental information gatherers, generating reports on companies in the AI sector. These agents experiment with different approaches to combing through financial statements, earnings calls, patent filings, and technical documentation to construct profiles of target companies. The goal is to discover which analytical approaches yield the most useful insights.
"The research agents are designed as a testbed for different analytical methodologies," explains Professor Nik Bear Brown, PhD, MBA. "We're building to learn which frameworks actually capture the nuances of AI companies. This is very much about discovery through experimentation."
Verification Agents serve as the framework's built-in skeptics, testing approaches to independently validating claims and checking data points from research reports against multiple sources. This experimental layer explores how different verification methodologies might reduce confirmation bias and enhance analytical reliability.
"In traditional investment research, verification often happens haphazardly, if at all," notes Professor Nik Bear Brown, PhD, MBA. "Mycroft's explicit verification layer lets us experiment with structured approaches to research integrity."
Comparative Analysis Agents explore how to effectively evaluate companies within specific AI subsectors such as semiconductors, cloud infrastructure, model developers, enterprise applications, and vertical solutions. These experimental agents test methods for identifying patterns across competitors that might escape detection in isolated company analyses.
While the analytical agents experiment with intelligence gathering, Mycroft's portfolio agents explore how this knowledge might be transformed into actionable investment strategies.
Diversification Agents test approaches to balancing a portfolio across multiple sectors. These experimental components explore different applications of portfolio theory and risk modeling, seeking to discover which methods actually work for maintaining exposure to AI's growth potential while buffering against sector-specific volatility.
"The diversification agents let us experiment with different mathematical approaches," says Professor Nik Bear Brown, PhD, MBA. "We're building various simulations and stress tests to learn which methods actually identify vulnerabilities in a portfolio."
Risk Management Agents implement experimental safeguards at both the individual position and portfolio levels. These agents test different approaches to position limits, stop-loss strategies, hedging opportunities, and correlation monitoring.
Rebalancing Agents explore different methodologies for assessing whether a portfolio allocation remains optimal given changing market conditions. Rather than assuming rigid rebalancing schedules are best, these experimental agents test dynamic approaches to discover what actually works in practice.
Mycroft's experimental approach extends to human interaction. The advisory agents explore different methods of communication between AI systems and human investors.
Conversational Financial Advisors test various approaches to creating natural dialogue about investment goals, risk tolerance, time horizons, and portfolio construction. These experimental interfaces adapt their communication style based on user expertise, allowing us to discover which communication approaches are most effective.
"The advisory agents represent our commitment to learning through building," emphasizes Professor Nik Bear Brown, PhD, MBA. "We're experimenting with different ways to make complex investment concepts accessible while avoiding oversimplification."
Goal-Setting Agents explore methodologies for establishing concrete investment objectives and translating abstract aspirations into quantifiable targets. This experimental component helps us discover effective approaches to explaining trade-offs between risk and return.
Educational Agents test different approaches to ongoing learning. These experimental components identify knowledge gaps and explain investment concepts in context, allowing us to learn which educational methodologies actually enhance financial literacy.
To ensure Mycroft's experimental analyses remain current, we're testing a network of intelligence agents that monitor developments across the AI landscape.
News Monitoring Agents explore different approaches to tracking reporting from technology publications, financial news sources, academic journals, and corporate announcements. These experimental components test methodologies for filtering the overwhelming volume of daily news.
Social Sentiment Agents experiment with analyzing discussions across social media platforms, developer forums, and industry communities. Unlike simplistic sentiment analysis tools, these experimental agents test approaches to contextualizing discussions within broader technical and industry frameworks.
Financial Report Agents explore methodologies for processing quarterly earnings releases, SEC filings, investor presentations, and other formal disclosures. These experimental components test approaches to identifying subtle shifts in language that might signal changing business conditions.
Regulatory Monitoring Agents experiment with tracking evolving policy landscapes around AI governance, data privacy, competition, and sector-specific regulations.
At the heart of the framework sits the experimental Mycroft layer—testing approaches to coordinating the activities of specialized agents, resolving conflicts, prioritizing actions, and maintaining overall system coherence.
"The Mycroft layer is where we experiment with coordination itself," explains Professor Nik Bear Brown, PhD, MBA. "Individual agents are narrowly focused by design. The Mycroft layer lets us test methods for synthesizing insights and resolving contradictions."
This experimental orchestration layer explores several key mechanisms:
Cross-Agent Validation tests approaches to identifying when different agents reach contradictory conclusions. Rather than simply averaging or voting, the experimental Mycroft layer traces reasoning paths to discover effective methods for resolving analytical conflicts.
Dynamic Task Allocation explores approaches to distributing computational resources based on changing priorities. When significant market developments occur, we test how quickly and effectively analytical capacity can be redirected.
Pattern Recognition experiments with identifying connections across seemingly unrelated developments. Like its literary namesake who could deduce far-reaching consequences from minor diplomatic incidents, the experimental Mycroft layer tests approaches to synthesizing disparate signals into coherent narratives.
Decision Optimization explores methodologies for translating insights into appropriate portfolio actions. The experimental Mycroft layer tests approaches to balancing competing recommendations while considering practical constraints that affect implementation.
Continuous Learning tests how the entire framework might improve over time. The experimental Mycroft layer tracks prediction accuracy, investment performance, and user satisfaction to discover which refinement methodologies actually enhance system performance.
The Mycroft framework was designed with a clear educational philosophy: to build systems that help us learn what actually works in AI-powered investment analysis. The project explicitly embraces experimentation and discovery rather than claiming to have definitive solutions.
"We're building to learn," emphasizes Professor Nik Bear Brown, PhD, MBA. "This open-source experiment is about discovering which approaches actually help process information effectively, identify meaningful patterns, and maintain disciplined investment strategies. We don't have all the answers—that's precisely why we're building."
This approach distinguishes Mycroft from commercial "black box" systems that make bold claims about performance. The framework's transparency allows contributors to understand the reasoning behind each component, challenge assumptions, and discover through experimentation which approaches yield the most valuable insights.
As artificial intelligence continues to transform industries worldwide, the Mycroft educational experiment offers a compelling example of how open-source development can contribute to our collective understanding of AI applications. By building and testing a recursive intelligence that helps investors navigate the very technological revolution it embodies, we're creating an educational platform that discovers what actually works.
Whether some of Mycroft's experimental approaches will prove effective remains to be seen—that's the nature of educational experimentation. But in combining specialized AI agents under the coordination of a sophisticated orchestration layer, the framework represents a significant opportunity for collaborative learning—one that could help individual investors have a better understand the AI revolution reshaping our world.