Invest with AI
00:33:53 · Brett Caughran, Khe Hy, Doug Clinton
[00:33:53] S2: Yeah. Okay. How have you thought about this? I mean, you've spent a lot of time with Claude specific training, Claude coworking in particular, which has been sort of the default UI for, you know, mid to small size funds. Super powerful, particularly when you sort of set it up with the right connectors and the right skills. If funds have adopted Cloud Desktop, Cloud Cowork Code, etcetera, and you see this model evolution, what's next? What's the counter punch that evolution? [00:34:27] S1: Folks are scratching their heads right now. I think that the smaller funds that I work with are they're not ready to go to the open source. I mean, a lot of these funds use managed, you know, managed IT services. Like they don't even have a dedicated IT person in house, let alone a software engineer in house. So I think the open source question is not for the kind of the long tail of smaller funds without the technical expertise. I think that there is a huge pull into Claude. I think some folks are still using ChatGPT web because they haven't like Codex was very scary, you know, just the name intimidated people. But Codex is having its co work moment. I mean, it had it and it's very, very good. The new Codex or ChatGPT work, whatever. I don't know. I don't even know what to call it anymore. Yeah. They need to [00:35:26] S0: fix that. Got a branding problem. You're right?
00:35:28 · Khe Hy
[00:35:28] S1: They've got big branding problem. I still call it Codex just so people know what I'm talking about. And I think some folks are starting, the folks that kept their ChatGPT subscription open are starting to dabble. I've found that in the folks I talk to, intelligence, so much of the way LMs are being used is like the end to end workflow or as much of the workflow that you can capture. So yes, obviously you want the model that is 2%, 3% better. But if the harness isn't there, then that raw intelligence doesn't really matter to my clients. And so they are kind of poking around on Codex. I think that until they fix the branding problem, though, no one's really pulled the trigger because no one understands what Codex is, but it's really co work. And I think that's the problem that people haven't had. And I think that just the issue of cost, I think one of the things that's really interesting about the chat GB, the codex versus co work conversation is that, as Doug said, the 5.6 model series is very good. It's very good at agentic work. I can't speak at the nuances of financial analysis, the edges of financial analysis, but it's also significantly cheaper. And I just as a tangent, I've got a $200 Claude Max Plan that I personally that I get close to using up regularly. And I've got a $20 GPT plan. And they're generous on their resets, but it takes a long time for me to hit the $20 limit on the GPT plan. So I think that folks are trying to figure it out. I think eventually they'll get to a dual place, but that's really like an IT and a mind share question until there's like a 20% delta in performance that's not there at the moment.
00:37:32 · Brett Caughran
[00:37:32] S2: Yeah. Question for you, Doug, perspective on this two ks. Like in other areas, we've seen the sort of wrapper businesses become ascending again, like Harvey and Ligora. It sort of solves a lot of these, like, you know, model router and data wholesale aggregation, you know, commercial questions. You know, we've seen that in certain areas of finance, investment banking with Rogo really sort of ascendant. We haven't seen that yet in investing yet. Is that due to just the heterogeneity of our craft? Or do you think that's just a timing issue that we'll start to see that moment where you see some leaders, know, AlphaSense is sort of most adjacent to that, FactSet's building out an AI workspace, etcetera. Bloomberg's sort of perpetually eighteen months behind the frontier. But I'm curious if you've heard any interesting vendor stories you think we'll start to see know, sort of the right you... Your workspace being this finance specific workspace. I
00:38:36 · Doug Clinton
[00:38:36] S0: haven't heard any sort of under the radar companies or workspaces yet, but I wouldn't be surprised if somebody does try something here, and maybe it is an AlphaSense or or or somebody like that. It would sort of make sense they might try. We've talked to... And and I know, Kaye, you've talked to a ton of people in this space too. We've talked to a handful of pretty big asset managers over the past year just about, you know, other ways we could help them solve problems. And and one of the things that always is an issue or a sticking point with us working with them in any capacity is how do you get through compliance related issues? How do you get through sort of data sovereignty related issues and making sure that, you know, if they're sharing data with us in some way, our systems aren't using that data. That data's not available to any of their competitors maybe that are using the same system. And so these are all, I think, solvable problems, but I think they're they're easier to solve technically than they are to to solve organizationally, if that makes sense. And I wonder if that level of headache is the thing that's maybe holding us back more than, you know, just the reality that the technology is probably ready. It's, you know, are the institutions ready and and are very large financial institutions, which which I would argue and I hope you guys would agree are are maybe not the most aggressive as it pertains to, being innovative are just really slow to try to figure it out too.
00:40:05 · Khe Hy, Doug Clinton
[00:40:05] S1: Yeah. I I would say that I have seen a small resurgence of kind of like the wrapper vertical tools, this first pass tool or something that just gives me all the investor transcripts and so on, like cleans it up. You know, the three steps that you had to do in Claude to get to sanitize and get that data into the working format. And people are more open to those. But it's a little bit aside of your outside of your question, Brett. In terms of the platform, I think that people want to... People kind of want the iPhone of all this. It's like you push a button. Yeah. And there's a recognition that the minute it becomes the iPhone, then you've flattened out the edge into the product. Right? And so you need to have an iPhone that's more like an Android where you can customize this widget and this widget and this widget and this data source. But then you're kind of back at the same starting problem. And so I think that people are it's a big change management question. And I think one of the questions that I think larger firms will ask is like, what will get people to use the thing? Right. And if it's giving them an iPhone, then give them a damn iPhone. But I think that there are always going to be firms that are going to say like, I want you to have the iPhone and I want you to have Chrome because I want you to take that widget because we never do the widget that way. Yeah. [00:41:36] S0: Yeah.
00:41:36 · Doug Clinton, Brett Caughran, Khe Hy
[00:41:36] S0: Yeah. [00:41:37] S2: Yeah. And then that intersects the sort of obsolescence or the bitter lesson dynamic. Like, I look back to our fall twenty five curriculum where we were teaching people about chat GPT projects and downloading documents, uploading them, creating system prompts, all of this, which is, like, completely obsolete now. Yeah. So how do you deal with the change management problem while the sort of like the baseline of what needs to be changed is constantly evolving and continually up for debate? A hairy Lots [00:42:10] S1: of work for people like us. [00:42:11] S2: It's a hairy, it's a hairy problem.