Key Takeaways

  • Barney Hussey-Yeo, founder of AI financial assistant Cleo, once funded his entire university education playing online poker, before complex Game Theory Optimal (GTO) solvers became mainstream.
  • Today, Hussey-Yeo has a personal AI agent that plays low-stakes online poker for him, generating a passive income of approximately $200,000 per year.
  • The advent of GTO solvers around 2016-17 transformed online poker from a game of human intuition and strategy into a largely "solved game" where optimal algorithmic play often prevails.
  • This highlights a sharp insight for founders: identify problems that are, or are becoming, algorithmically solvable, then automate their execution to capture value and free up human bandwidth.

The Method

Barney Hussey-Yeo wasn't just good at online poker; he was good enough to pay for university. This was in an era when human ingenuity and psychological reads held sway. Then, the landscape shifted dramatically. Around 2016 or 2017, Game Theory Optimal (GTO) solvers arrived. As John Collison clarified on the podcast, GTO is about “doing the exact right thing that the probability tables would tell you that you should do, right?”

These solvers effectively codified the optimal strategy for nearly every poker scenario, turning a game of calculated risk and human reads into a computational exercise. Hussey-Yeo describes it starkly: “we actually learned how you could build beat the game. You it's still right at the, you know, the edge... but it's basically a solved game, which means online play is kind of is dead now, but you should play tournaments.”

But rather than mourn the loss of human dominance, Hussey-Yeo saw an opportunity. For him, the game wasn't truly dead; it was just solved. His solution? Build an AI agent to play for him. It's not a grand, sentient AI, but a focused algorithm. “I've just built an AI agent uh that now plays for me and earns rake back,” Hussey-Yeo revealed, adding, “So I make about I shouldn't say this publicly, but I earn about 200k every year now just through my little agent that plays and barely beats the low stakes.” This agent exploits the now-known optimal strategies, consistently extracting small but cumulative wins at tables where human players still make mistakes. It's arbitrage, but with algorithms.

His method is deceptively simple: find a domain where the "optimal" strategy is largely known or can be computed, and then deploy an automated agent to execute it relentlessly, capturing value that humans either miss or can't execute consistently. It's about letting machines do what they do best – tireless, emotionless, optimal execution – in a well-defined environment.

Where This Breaks Down

Hussey-Yeo's poker AI is a powerful illustration, but it's not a universal blueprint. This method hinges entirely on the "solved game" premise, and not every domain is neatly solvable by current AI. Poker, at low stakes, became predictable enough. High-stakes poker, however, still demands human adaptability, reading opponents, and handling unexpected variables that even GTO solvers can't fully account for when other players are also optimizing. The game continually evolves, and human intuition at the bleeding edge remains critical.

This approach also stumbles in highly regulated industries or those demanding deep empathy. While Cleo applies AI to finance, the human element of understanding complex customer needs or navigating nuanced regulatory hurdles (which Hussey-Yeo discussed in other parts of the episode, highlighting UK vs. US financial regulation) often prevents full automation. You can't just deploy a bot to handle every aspect of a startup that requires trust, human interaction, or creative problem-solving in truly novel situations. And, of course, there's the ethical question: if you're not transparent about an AI playing, is it fair play? For this method to work, the "rules" of the game must be stable, the stakes appropriately low, and the environment largely unreactive to advanced AI deployment.

What to Do With This

Founders in their 20s and 30s should absorb Hussey-Yeo's mindset. Stop thinking about AI as a distant, general intelligence that will solve everything. Instead, look for "solved games" within your own business. What repetitive tasks, decision-heavy workflows, or data analysis points are currently handled by humans, but could be reduced to a set of optimal, algorithmic rules?

This week, pick one internal process. Maybe it's customer support triage, initial sales outreach qualification, or even onboarding new employees. Map out the decisions made at each step. Are those decisions based on clear parameters, even if complex for a human to execute consistently? If so, you're looking at a solvable problem. Don't aim for AGI; aim for algorithmic arbitrage. Build a simple agent – even a script or a basic rule-based AI – to handle the predictable 80% of that task. This isn't about eliminating jobs, it's about shifting your team's energy from repetitive execution to the truly creative, human-centric problems only they can solve. Follow Hussey-Yeo's lead: free your human capital by automating the "solved." The passive income might surprise you, not just in cash, but in reclaimed time and focus.