Perfecting the Investing Craft with Caro-Kann’s Artem Fokin

Episode Summary

Executive Summary: Andrew Walker and Arden Focan discuss investing as a craft, emphasizing iterative process improvement, better customer diligence, and how AI/expert calls may reshape research. They debate track-record interpretation, base rates, convexity, and how to avoid outsourcing conviction or drifting into groupthink. The core message: use tools to augment judgment, not replace it.

Main Topics: Investing as a craft and process improvement (Priority: 5/5): The conversation frames investing as a skill that improves through iteration, feedback loops, mentorship, and exposure to other investors' process ideas rather than just their stock picks. How to interpret investor track records (Priority: 5/5): They debate whether strong returns prove skill, noting statistical significance problems, style differences, and the possibility that some outsized returns reflect a narrow sample or luck rather than repeatable edge. AI, expert calls, and the future of research (Priority: 5/5): They discuss how AI and expert-call libraries can democratize access to information, but also risk making market participants more consensus-driven if they outsource thinking to the same tools. Customer diligence as a major process upgrade (Priority: 4/5): Arden says he now puts much more weight on understanding customers, based partly on a past mistake where he underestimated the difference between satisfied customers and true raving fans. Convexity, base rates, and compounders (Priority: 4/5): They argue that investing in compounders means betting against base rates and that success requires both improving probabilities and having enough upside convexity to make outlier outcomes matter. Avoiding groupthink and outsourcing conviction (Priority: 4/5): They stress that ideas from respected peers should be welcomed into the research pipeline, but conviction must be built independently through the investor's own process. Sizing, IRR, leverage, and risk adjustment (Priority: 4/5): They explore how ranking ideas by expected IRR can inadvertently overweight levered names, and how hard it is to create a universal ranking system that properly adjusts for risk.

Key Arguments: Investing is a craft that improves via iteration, feedback, and copying better process ideas, not just by chasing performance. Strong track records are informative but not always conclusive; style, sample size, and base rates matter. In compounder investing, investors are statistically betting on outliers, so they need both better screening and upside convexity. AI will likely make research more consensus-driven unless investors use it as a supplement rather than a substitute for judgment. Expert-call libraries and AI together can speed up research, especially where written information is sparse or non-existent. Understanding the customer is one of the best ways to improve investing process because it reveals real product value and switching behavior. Ideas from peers should raise research priority, not lower the investment bar or replace independent conviction. Expected-IRR ranking can unintentionally push portfolios toward highly levered names, so risk adjustments and qualitative judgment remain necessary.

Data Points: Podcast guest count: 7 or 8 appearances - Andrew says Arden Focan is one of the podcast's most frequent guests. Small-cap community size: about 50 people - Andrew estimates the core group of similar small-cap value managers is around 50. Bitcoin example appreciation: 200 to 115,000 - Used as a hypothetical illustration of a 10-year hold from 2012 to present. Bitcoin return multiple: 500+X - Andrew describes a buyer from 10 years ago as potentially having 500+ times returns. Research habit frequency: once a week - Andrew says he forces himself to do one expert call per week. AlphaSense/Tegus expert call library usage: 5 expert calls - Andrew describes a workflow of tossing five expert calls into an LLM for summarization. Customer-call threshold: 5 calls - Arden cites five customer expert calls as a basic threshold, though he notes it is low if more are available. Reading load example: 3 years - Arden suggests some investment processes require reading three years of earnings calls and presentations. Financial time horizon: 10 years - Used repeatedly in examples about compounders, track records, and process learning. Levered company example: 6x EBITDA / 4x debt / 2x equity - Arden contrasts two otherwise identical IRR opportunities to show why leverage matters. Market-cap example: $2 billion vs $5 million - Illustrates how access to expert calls used to favor larger funds before tools became cheaper.

Pivotal Quotes: "Investing is a craft." — Arden Focan: Arden defines the podcast's theme as improving through repetition, feedback, and process refinement. "Don't outsource conviction, that's a dumb idea, but make idea your own." — Arden Focan: He explains how to handle investment ideas sourced from respected peers without falling into groupthink. "AI will likely do is that different competitive advantages or competitive strengths would be a better term that people have may become obsolete, and then you need to reinvent yourself and figure out your new competitive strengths." — Arden Focan: Arden describes how AI could compress old research edges and force investors to adapt.

Implications: Listeners should treat AI and expert calls as accelerants, not substitutes, and build independent conviction through deeper customer, industry, and process work. The best edge may shift from information access to original insight generation.