Edition 5

The premise is that the most valuable data in vertical AI is generated but lost every day. For example, someone overturns a denied prior authorization, corrects a freight dispute, or swaps an equipment order, but the reasoning evaporates the second they move on to the next case.

This is relevant because while agents can already follow rules, pull the right data, and generate documents, they cannot handle the case that does not fit the template. These calls do not live in the ERP or CRM, but rather pile up in the "exception queue" and vanish as fast as humans clear them.

The argument is that over time, enterprise value accrues to whichever software vendor can capture the reasoning behind that judgment.

The caveat on this wedge is that there are two axes: exception density and the cost of failure or delay. High density with low consequence yields a fragile productivity tool; high consequence with low frequency behaves more like a consulting project. Value sits where a costly, recurring queue of judgment calls already exists, e.g., the prior authorizations, freight disputes, permits, and claims that are work "someone is already paid to clear."

Further, what compounds is captured reasoning. That is, does the product sit where decision traces are a natural byproduct of usage? If so, you can build proprietary workflow memory that today mainly lives in domain experts' heads. "The workflow you ship on day one is not the moat. The loop that production usage creates over time is."

The labs may find this hard to follow, as they can keep shipping better models and better general agents, but they do not sit inside a carrier's production workflows long enough to learn why one account was escalated, why one risk was declined, or why an underwriter overrode the appetite guide and was right to do so.

When growth slows in a software business, whose fault is it? Until about 18 months ago, the default answer was sales, because B2B products were mostly static. You added features, caught up to competitors, and shipped a module every few years, so slowing growth was usually an execution problem a great sales leader could fix.

The pace of product development has very obviously changed. Products that were competitive last year are getting lapped by competitors shipping weekly. "Claude is literally 100x better than it was 12 months ago. Cursor went from $100M to $2B ARR in roughly 14 months."

So, back to the question of whose fault is slowing growth, the answer may be flipping. If growth is slowing, it may be less likely that it's "do we have the right VP of Sales?" but rather "is our product still competitive?" Going 6-12 months between meaningful improvements is now a death sentence, and the sales team can feel it before the dashboards do as win rates drop and competitive losses tick up.

Jason's conclusion is "Fire your VP of Sales if you want. But if your product isn't as competitive as it was 6-18 months ago, I doubt it will help... Product velocity is the new growth lever. Get that right, and a great VPS can work their magic on top of it. Get it wrong, and no amount of sales talent saves you."

The Billion Dollar PDF: more and more, a single confident narrative can reorder capital when someone crystallizes the right story at the right time, even if its not fully proven. A viral essay or X thread sets the market's mood, then money follows, in a narrative before returns type scaling.

Jeremy welcomes automation, claiming that anything that can be automated should be automated. Provocatively, he claims every white collar job is "like totally fake and made up" in the sense that these are not contingent for shelter and food and medicine and other necessities. The point being he doesn't think we're at peak jobs, because we'll just make up new jobs to do to meet humanity's unlimited wants and desires.

Certain founders are like modern-day feudal lords, but the grants of generational wealth are allocations of SpaceX, Anthropic, etc. rather than land. Even managers who can sell that access via SPVs, often charging upfront fees and perpetual cuts for little risk, are a new form of rent allocation.

This is his second time on the show, and I actually recommend the first, popular for his points on 1) There is no such thing as general advice. Asking someone for it is simply asking them to make up a general rule based on their own specific experience. Instead, ask people what they did and why, and decide whether and how it applies to your life, yourself. 2) The more advanced the investor is, the simpler his questions are. Anyone can do spreadsheets, but only great investors can do napkin math. It’s the ones that get you excited by their obvious simplicity rather than hidden complexity that are the real gems. 3) If you have to spend countless hours debating an investment idea, you’ve already lost. The best investments should be easy. Where everything lines up, and it all makes sense. Leave the coin flips to the speculators who are forced to deploy capital on a specific timeline.