00:16:14 · Brett Caughran, Khe Hy
[00:16:14] S0: Yeah. I love this question because I think like, you know, as a Wall Street user, we stare at this blinking context window and it's actually like harder than it would seem to figure out to enact that context window to be helpful in your investment process. Worse than that, I think the knee jerk reaction of most investors is to use it the wrong way. Is this sort of, hey, I used to read the 10 ks, let me do a 10 ks summary now and speed read through that, the artisanal process of comprehending the company. And then sort of, you know, or the expression layer, like here's a bunch of notes, create a bold thesis for me, right? It's sort of an AI slop thesis. So I think I don't know what it is about AI, but it causes probably late, Fred, like it helps elevate the sort of lazy predispositions in all of us that sort of just want to like sit on the couch and eat Cheetos versus apply real rigor. And it makes makes the sort of transmission mechanism to laziness much easier because it sort of looks good enough. [00:17:19] S1: Yeah, yeah.
00:17:20 · Brett Caughran
[00:17:20] S0: That to me is a very important force to fight against. Yeah, yeah. So I think it like we think about it as sort of five layers of like the triage layer, the execution layer, the decision decision, decision support layer, the sort of rigor enforcement layer, and then the idea tracking layer. So like, it really is to me like building a system around it. To your specific question, the decision support layer, you know, something like that where you could build a headless RMS to sort of go into all of the prior operating memorandum, all of the prior deals, pattern recognize the sort of the best deals, the worst deals, errors you've made in the past, and have that in a dashboard while you're sort of doing the evaluation of a new operating memorandum. That to me is like that decision support layer is the real... It's not some AI slop summary, right? It's, hey, go and actually run and analyze, you know, sort of a real institutional grade process on this OM. And part of it is, you know, pulling in the lessons of history. I think to me what has been so exciting, and I think we're at a real inflection point on these tools on Wall Street is the fable and soul class of models have gone from answer engines, prior LMs were like Q and A, give me an answer, to actual work mechanisms. To me, you can build with Fable and Soul is so much more impressive than what you could build with Opus or GPT 4.5, etcetera. So the ability to actually, you know, ramble Max into a context window and have it build the first version of, you know, some architectural thing is I think that we were sort of going through this diffusion mechanism where more people are realizing how powerful this latest class of models and probably then sort of looping that back into the work you're doing with Codex. Like you probably just ramble into Codex and that decision support layer on the OM analysis in a way that's like much more frictionless than it was six months ago.
00:19:36 · Khe Hy
[00:19:36] S1: Yeah, yeah, absolutely. And I I think part of this question, though, it comes back to context and data, right? And and again, you know, I've been spending a little bit more time with private equity, but they have mountains of data from their portfolio companies, right? And this is, again, I think this is a lot of that data falls into the category of, if a human were paid 6 figures to analyze this, we wouldn't do it. But now we can pay $100. What's the question you want to ask, right? Are you doing a cohort analysis that didn't make sense to do because it would have taken someone hours to do? And now maybe it's not as good as a human, but let's say it's 90% there, and it's gonna cost you $20, right? And I think that, you know, and people really people are really struggling with this. Think I'm surprised and I'm not surprised. I just take you on a side tangent here. It's like I was an early user of Notion. And Notion was like, Oh, you could do a database. You could do this. You could do a spreadsheet. You could do a to do. You know, like, it was a little bit like AI, but then you open it and it was just like... It looked like a Google Doc. You're You're like, Wait a minute. You're telling me I could categorize, like, all the Pokemon in the world, but all I see is a flashing cursor. And, like, Notion still struggles that people can't see all of the tools that they're giving you and then map it to how they see the world and the problems they're trying to solve and do it. So they still hire Notion consultant, right? They hire these intermediary that acts as this like, bridge between them. And I've been wondering I've been trying to, like, ask my you know, prompt my clients in different ways. Like, you know, even things Like, you know, if you hired not even a new analyst, but you hired someone senior like yourself, like, what would you have them do? Right? And sometimes they don't... They have to, like, step back and like, Well, I'm not exactly sure. And like, Oh, well... You know? And so if you're not sure what you would do with another person, you're gonna be even less sure what you're gonna tell a powerful AI to do. And so there is that process that, again, I don't know.
00:22:06 · Khe Hy, Brett Caughran
I haven't found an easy way to get around it other than spending time with people who truly understand what they want out of an investment process, like what the end game is, and that are willing to, like, be weird about it, right? Yeah. [00:22:23] S0: I think you nailed it. And I think certainly for public markets investors, the challenge in AI adoption has been the heterogeneity of what investors want, right? You could have two hedge funds. One will make investment because they like the management team and like the product. Another will invest solely on revisions to revenue guidance over the next three quarters, right? And so that has been very difficult for both internal builders and product builders for public market AI tools because it's been very hard to design something that crosses the bar, passes the bar for all investors. That's why this sort of like mass personalization engine that's possible now with these tools is like so exciting. The ability to actually spin that up and to some degree, maybe make customizations of the harness, but certainly customizations to the skills layer and the context layer and the dashboard layer, that's now becoming very possible. And so I think very, like I've been more and more bullish on the application layer for a number of reasons, but I think you're going to very soon, all you have to do is sort of go to a fund, give them logins and you pre configured a dashboard to them very specific to their investment process. It's going to be really sort of a wow factor. And the skills as well, right? How have you
00:24:00 · Khe Hy
[00:24:00] S1: been finding? Are your clients finding it easy to, like creating a skill is relatively easy, but it's imagining what skills you need and how they might fit together. Like the kind what we've been discussing before, art of the possible. How has skill building evolved or shown up in your work?
00:24:24 · Brett Caughran
[00:24:24] S0: Yeah, it's constantly changing like everything, right? And I'll give you sort of one example. I had a deep dive skill that created 100 page primer. And you have to sort of the prior model iterations, you sort of go through like, hey, these are the 12 analyses I want to do on this company, right? It would run the same 12 analyses on every company. It would adhere to those skills. One of those analyses may be balance sheet and liquidity, which could be important for a leveraged company, but not that important for a company that has net cash. But in sifting through that primer, I'm still reading those sections for every company. So there's sort of like the concision dimension was just not that satisfying, right? This is one of the things that unlocked with Fable and Soul is sort of ability to sort of rethink that operating system. One shift I've made to that sort of simple primer skill is this concept of, I call it twenty-sixty-twenty, where it's like, it's 20 pages of standardized sort of business overview, setup overview. But then like a good senior analyst, that front end is actually trying to understand what the key debates are in this situation and spending the meat of the primer actually going in and doing analyses, right? And so it's identifying what matters in the first 20 pages, and then it's going to give like a deep, you know, custom bespoke analysis of each of those factors. And this has been like a complete unlock for me terms of the quality of the primer that you can create with this sort of skill. But it also creates a totally different system. Have to create the pool of deep of key driver workflows that the digital analyst can call from. So it's like, I'm kind of like decommissioning the old structure to create a new structure that that really leans on the new capability, the sort of more end to end capability. The exciting thing is you're getting to the point with Excel fluency. These new models are much more Excel fluent, Where after you've identified the key driver, say looking at Brunswick that has a boat business and I want to spin up a price like an aftermarket pricing, secondhand pricing tracker to go scrape boat trader, etcetera, you can start to sort of build into the system, Hey, this aftermarket boat pricing is a key driver because it sort of impacts a price gap of new, etcetera.