Does the right tail require a different lens?

Yesterday in Holding two ideas in tension, I wrote about how hard it was to believe in AI and, at the same time, apply a traditional investment framework that justified current market prices given the assumptions driving the underwriting.

Today, Patrick O’Shaughnessy released a podcast interviewing D1 founder and CEO Dan Sundheim and he summarized the tension perfectly:

“Everything about AI’s impact on the economy is inherently low conviction. Because I think everyone is likely underestimating how much these models are going to improve. And to really think about what’s going to happen, you have to almost not think like an investor, you have to think like somebody who’s into science fiction.”

I think he is pointing to a nuance I underweighted yesterday. The “base-rate” lens is still necessary, but it might be insufficient if you are underwriting a general-purpose technology. In that world, the right question is not “can I model this confidently?” It is “what is the skew, and what are the few variables that matter most?”

So much about AI right now sits on top of the paradigm change it may unlock across industries. You need a lot of mental flexibility to imagine a distant future and accept that the way you thought about businesses, moats, and margins might be rewritten.

If the models keep improving and the TAM is as broad as it seems, then a lot of what looks like over-optimism might just be a different underwriting style: accept low conviction on the path, but take the bet because the right tail is so extreme. The discipline shifts from forecasting to position sizing, survivability, and being explicit about what has to be true for the upside to compound.