Gabe Stengel - Building Investing Superintelligence

Patrick O'Shaughnessy (0:01)

Ramp is the only platform built to make your finance team leaner, faster and better, saving businesses 5% annually on average, so you can stay focused on growth. Ramp just opened for business in the United Kingdom. So if you're running a business in the UK, you can now use Ramp's AI powered finance platform to manage cards, expenses, bills, approvals and accounting all in one place. I run my business on Ramp and so should you. Learn more at ramp.com/invest.

OpenAI, Cursor, Anthropic, Perplexity and Vercel all have something in common. They all use Work OS. And here's why.

To achieve enterprise adoption at scale, you have to deliver on core capabilities like SSO, SCIM, RBAC and audit logs. That's where Work OS comes in. Instead of spending months building these mission critical capabilities yourself, you can just use Work OS APIs to gain all of them on day zero. That's why so many of the top AI teams you hear about already run on Work OS. Work OS is the fastest way to become enterprise ready and stay focused on what matters most, your product. Visit workos.com to get started.

You and I have talked many times about this basic question that I'll start with over the last couple of years. You are effectively trying to build investing super intelligence tools to help investors do their job much faster, better, cheaper, easier, higher quality. But it's starting to feel like, wow, we're really eating a lot of the core functions that even a very smart analyst or even portfolio manager was doing a couple of years ago. How do you think about that trajectory as you've seen it and lived it so far and where it's going over the next two years?

Gabe Stengel (1:42)

I think two years is actually easier to reason about than 10 years or 20 years. Because in two years, the best investors are going to be figuring out how to reinvent their own firms and reinvent themselves. And if you look at what happened to market making and on trading, James Street took 15 years to build the dominant franchise. And the world's best investors today are going to spend the next two to five years figuring out how to integrate AI into what they do.

And Dario has the great line about everyone's going to have a data center full of geniuses or a country full of geniuses in the data center. What would Goldman do? What would Millennium do? What would Citadel do if they had a country full of geniuses show up? It would probably take them a while to figure out how to change the way they work, how to take advantage of that, how to integrate it into their system. I think figuring out how to fly AI into the investment life cycle is the biggest challenge over the next five years for every great investor.

Patrick O'Shaughnessy (2:32)

There's been many companies on this trajectory where the product was, Cognition is very famous for like literally their ads. Now say, remember Devin, like it's good now.

So lots of now clearly great companies with great products had a stage of an AI business where the product stunk. And now it's excellent. If you think about the couple increases in capability that we've seen just from the raw models, could you do the same thing for the eras of Rogo and its product? Like you pick how many eras it is. I don't know how you frame it up, but like what it could do at each level up to and including today.

Gabe Stengel (3:08)

Yeah. I actually tried starting Rogo two times before we got started. So in high school, I had a friend whose dad was an investment banker who wanted an app for trying to track the basically equity exchange rate of two public companies as they were merging. So we tried using really old AI techniques to do that. Terrible. Then in college, before GBD3 came out, we published a paper on AI assistance for econometrics and financial econometrics and we tried commercializing at the time and nothing worked at all. Then when we actually started the business, it was when GBD3 came out, pre-Chat, GBT.

All the early days of Rogo, it was clear how magical it was. You could demo things that were cool, nothing worked at all. I mean, I would say since things actually started to work, the eras are very tied to the model eras, right? It was O1 Pro and then probably Opus 4.5.

O1 Pro was the first time you got enough reliability where it was a good search tool at the very least. You could say, help me calculate this financial metric for this business over the last 12 quarters, and it could do it reliably enough where it wasn't so annoying that you would just do it yourself. And then with Opus 4.5 and the end of last year, the end of 2025, end of the beginning of this year, the models just became capable of basically anything a junior investment professional or junior banker was doing, as long as you gave it the right instructions and context. I mean, I think there was a first movers disadvantage for a lot of applied AI companies because you thought you knew where the world was going and you wanted to build a product for it, but the models weren't quite there. And so people would try to go, this is terrible, this is garbage. You have a disadvantage.