Generative Now | AI Builders on Creating the Future
00:10:28 · Mike Mignano, Gabe Stengel
[00:10:28] S0: So these are your own models. You're not using, you know, GPT? [00:10:33] S1: Well, we build on top of those. So, it's like a lot of these platforms now offer fine tuning on top of their models. So, you can fine tune 4.0. You can fine tune 2.5 Pro. You can also distill these models into open source and then run reinforcement learning. So we're not doing any pre training. But we are then, you know, doing post training on these models. [00:10:51] S0: Yeah. And you mentioned that you're you're sort of you're fine tuning them to to work with the tools that, you know, people in this industry are working with. Talk a little bit about that. Like, do you mean actually sort of agentically using these tools?
00:10:51 · Mike Mignano, Gabe Stengel
[00:10:51] S0: Yeah. And you mentioned that you're you're sort of you're fine tuning them to to work with the tools that, you know, people in this industry are working with. Talk a little bit about that. Like, do you mean actually sort of agentically using these tools? [00:11:02] S1: Oh, Yeah, yeah, If you were an analyst at Goldman and you started on Monday and someone said, hey, find the precedent transactions for the Microsoft Activision deal, you would probably go to EDGAR's, which is the SEC's, you know, list of filings and start looking. Microsoft Activision. Look for the DEF 14 a. Okay. I found it. Now I'm gonna control f through the DEF 14 a for the precedent transactions. Okay. Great. There they are. That's not a tool that an agent has access to. Think we more and more see agents that have browser capabilities and so could That's do that a lot less efficient than just giving an agent access to a filing search tool where it can put in the ticker of a company, filter by a filing type and a date range, and then control F through those documents. And so we both build the tools and then we teach the model how to use it. And so for each different type of tool, we actually can have specialized models that query those documents more effectively. Right? They do the right keywords. Or a specialized tool for, you know, formatting Excels. And so we have to teach the models how to use these in a much lower latency, much higher accuracy sort of way.
00:12:06 · Mike Mignano
[00:12:06] S0: Yeah. I'm sure you're asked this all the time. You know, common question for, I feel like, any AI startup. You know, how do you think about the role of Vrogo in using all these tools and accessing all these data sources that we talked about versus the role of an OpenAI, an Anthropic, you know, an x x AI that, you know, probably want access to many of the same data sources and, you know, maybe over time have the capability to to do, you know, perform many of the same agentic tasks.
00:12:33 · Gabe Stengel
[00:12:33] S1: The better the next generation of models is, the better ROGO will be. Right? We can very quickly take those models and take that intelligence and and have it build into our system almost immediately based on the way that we architect. So it's great for us. And the difference will be we are spending all of our time on financial tools and financial workflows. And so when GPT-five comes out and we already have access to all this tool data and how they're used most effectively and evals over how to do these financial workflows, we're gonna be able to more quickly distill that into an end to end system that can, you know, generate a SIM, generate a PIB, analyze an S1, then they will. And they'll have to collect those evals and create those tools over time. And, you know, finance is a big enough market where they will likely start to focus on it. But there's been more focus on sort of generic consumer applications competing with the Microsoft Office suite, coding. And I think finance is a domain where, you know, there's a lot more intricacy from the product layer and the workflow layer than people expect. And so I think we're gonna have, you know, a a distinct advantage for a while. And ideally, the point where they're turning their attention on finance, we have a of actual user data and evals and sense of what the workflows are such that, you know, we'll always be a
00:13:44 · Mike Mignano
[00:13:44] S0: step ahead. Yeah, that seems to be the big debate, right? You know, where obviously the word moat, you know, we talk about so much in AI and it's, you know, it's constantly this question of, oh, you know, do you want to be the model provider? Do you want to be at the application layer? And I think, you know, the logic that I hear often, and I frankly, I agree with them, it sounds like it's similar to yours. It's like, you need to specialize in something and you need to run really, really far ahead. Right? And as you said, unlock the proprietary data and, you know, lock in the different workflows that are going to be really, really hard for someone else who's very horizontal to specialize in. Is that kind of how you guys think about it?
00:14:22 · Gabe Stengel
[00:14:22] S1: Yeah, a 100%. I mean, I actually think more and more so the value is in the application layer. Even if I look at my own usage of ChatGPT versus Gemini versus Claude. At this point, because I have a history of threads on ChatGPT, it responds in a more personalized way that's really helpful. And I don't need to go in and explain to Gemini, hey, I'm Gabe Stengel. I have a company called Rogo. This is what we do. Use that context to answer the question. ChatchBT does a very good job of of personalizing responses, recording what I'm doing, you know, having a good sense of of the types of answers I like. And now it's kind of a pain for me to switch. Totally. And that even if the model is slightly worse, I mean, I think that's a that's an advantage of application and product layer. At the model layer, I mean, I'm not an expert on these models at all. But my sense is, you know, no one's ever going to be more than six months ahead of anyone else. And so the distinct advantage you'll have as a model provider is just, you know, how low are your costs, how reliable are your services, how low latency is, kind of thing.