Mycroft: An Open Source Educational Experiment Using AI to Invest in AI | Building to Learn

The video introduces Mycroft, an open-source, multi-agent framework that uses AI to analyze investments in the AI economy. It is presented as an educational experiment: the project builds in order to learn which approaches actually help with processing information, spotting patterns and keeping an investment strategy disciplined, rather than claiming a finished answer. The name comes from Sherlock Holmes's elder brother, described as an analyst whose powers may have exceeded the detective's own.

Why better models do not automatically mean better decisions

The video opens with the observation that the intelligence revolution is no longer something people are waiting for. Current frontier models can reason across complex problems, write and debug software, analyze documents and images, conduct research and carry out workflows using digital tools. The bigger shift, the narrator argues, is the rise of AI agents: systems that plan, investigate, compare evidence, use tools and coordinate with other specialized agents.

Greater intelligence still does not produce better decisions by itself. Financial markets stay noisy, uncertain and deeply human, AI companies change quickly, valuations can move faster than fundamentals, and yesterday's market leader can turn into tomorrow's cautionary tale. That is the case the video makes for tools built specifically for investing in this sector.

Research and verification agents

Mycroft's research agents generate reports on companies across the AI ecosystem, from semiconductor makers and cloud platforms to foundation model developers, enterprise applications, robotics, cybersecurity, data centers and energy infrastructure.

Reports are not the end of the process. Verification agents independently examine important claims, compare multiple sources, identify conflicting evidence and separate established facts from assumptions and speculation. Comparative analysis agents then evaluate companies within clearly defined sub-sectors, looking at competitive advantages, financial performance, technical capabilities, strategic partnerships, market positioning, vulnerabilities and risks.

Portfolio, diversification and risk agents

Portfolio agents turn the research into disciplined investment strategies. Diversification agents monitor exposure across companies, industries, technologies and economic scenarios. Risk management agents evaluate concentration, volatility, liquidity, drawdowns and position limits so that an idea does not grow into excessive risk.

Advisors and intelligence agents

Conversational advisors adapt to the user's level of knowledge, whether that person is an experienced investor, a researcher or a beginner, and translate technical and financial concepts into accessible guidance. Intelligence agents continuously monitor market news, regulatory developments, company disclosures, technical research, competitive announcements and shifts in public sentiment.

The orchestration layer

At the center is the orchestration layer, the system that coordinates the specialized agents, records the evidence, challenges conclusions and combines the agents' work. The video's claim is that the combined result is greater than the sum of its parts.

What Mycroft does not claim

The video is explicit about limits. Mycroft does not predict the future with certainty. Instead it helps a person investigate possibilities, question assumptions, understand uncertainty and make more informed decisions. It is meant to amplify human judgment, not replace it, bringing research and analytical capabilities to individual investors while leaving accountability and final decisions with people. The project is described as an open-source framework for transparent, evidence-driven investment research.

Key takeaways

• Mycroft is an open-source, multi-agent framework for researching investments in the AI economy, built as an educational experiment.
• Specialized agents split the work: research, verification, comparative analysis, portfolio construction, diversification, risk management, advice and news monitoring.
• Verification agents check claims against multiple sources and separate facts from assumptions and speculation.
• Risk agents look at concentration, volatility, liquidity, drawdowns and position limits.
• An orchestration layer coordinates the agents, records the evidence and challenges conclusions.
• The project does not promise certainty about the future; it aims to help people question assumptions and understand uncertainty.
• Final decisions and accountability stay with the human investor.

Try it yourself

Visit the Mycroft repository on GitHub, listed in the video description, and read how the agent roles are organized. Then pick one AI company and sketch which of the roles above (research, verification, risk) you would want to check its claims, so you can see where an agent adds value and where you would still decide for yourself.