What this covers

Patrick interviews Graham Duncan, who has spent his career allocating capital to investment managers and now seeds and backs them. The conversation moves from how to read a person to how to structure a deal with one. On the reading side, Duncan argues that interviewers distort what they observe, since a nervous interviewer can make a candidate look nervous, so references matter. He uses Jonathan Haidt's rider and elephant to ask which self a reference is speaking for, and says thorough referencing gives roughly 75 percent signal on behavior. He also stresses how much track records owe to luck and Path Dependence, such as whether a fund's first year goes well.

The middle of the conversation covers the economics of backing talent. Duncan draws on Peter Koenig's idea of source, the person who takes the first risk, a claim he notes is contested. He walks through Pricing Talent across autonomy, transparency, fees and duration, tested by whether the terms would feel exploitative if he were the talent. He flags Adverse Selection in Seeding and prefers to catch people very young. He also describes an easy arbitrage in giving up-and-comers office space, as Richard Rainwater did. Being commercial means creating more value than you capture, which he links to Goldman's "long-term greedy."

The later turns are more personal. Duncan estimates 10 to 30 true master investors exist. He describes the best teams as mixing Harvard-style planners with Chicago-style improvisers, and argues that a good CIO needs a quiet ego. He explains why founders must half-pretend certainty and why leaders hold situations with a grip that is solid but loose, like rowing. He recalls how Stuart Miller made people feel rooted for rather than managed.

Sharpest takeaway

Duncan argues that durable investment returns come from pairing a principal with an agent whose commercial instincts and worldview genuinely overlap with theirs, and that this matching, along with sourcing and pricing the talent correctly, matters more than platform design or conventional hiring processes.

  • Principal-agent relationships fail when the principal treats the agent as a hired hand, because it suppresses the commercial risk-taking that makes the agent valuable
  • Hiring and seeding decisions depend on who is 'source' and on pricing talent at the right level of autonomy, fees, and duration
  • Reference checks and observed behavior over time reveal more signal about a person than interviews, which are distorted by the interviewer's own behavior

The argument · threads2 threads · 84 claims
0.71

Unique capital, not access to consensus deals, is what makes seed investing valuable.

2 pointscentrality 4/5
  • Jesse Beyroutey's framework asks whether a capital provider supplies unique capital that the recipient could not otherwise get, versus winning access to consensus deals where other capital would flow anyway; Duncan's seed deals fall into the former category.

    “whether the kind of capital you want to provide is unique, singular or not”

  • Patrick observes that Founders Fund GPs describe the same dynamic: they want deal heat and leave when the price is wrong, so pricing talent is recursive and applies at every level of the market.

    “It's so interesting how it turtles all the way down.”

0.43

Great CIOs require wanting materially better outcomes and accepting higher ambition and risk.

1 pointcentrality 3/5
  • Most principals default to a conventional advisor or a beta-like outcome from an institutional CIO; finding a great CIO requires wanting something materially better and accepting the trade-off of higher ambition and risk.

    “You have to deeply want something better than that to find that great CIO.”