00:16:04 · Brett Caughran
[00:16:04] S1: What would be like one or two examples of that sort of
00:16:08 · Ying Hua
[00:16:08] S0: I'll give you on the boring one because I think no one actually out there is actually care about it. So I don't disclose anything that our client tells us. I used to cover insurance companies, car insurance companies. And car insurance company, you estimate loss ratio a lot of times kind of figuring the wind you know like oh the weather is about the same a little colder so I'm going to bump up a few basis points in loss ratio. When I was a band one of the fundamental thing we did was we, actually, build a tracker or build a scrape for seven of the top is like highway patrol websites. And we pulled data around accidents. So for every major city, major like biggest, seven of the top 10 markets which I think is more than like 70% of market share, We actually know exactly how many car accidents are happening. And then we look at the duration of the car accident, when did it clean up as a proxy for severity. Then we actually have an index, It just runs in the background. Have an index on is this should be a higher loss ratio or not. And then we combine that with domain knowledge. Right think about, we know that every third quarter, insurance companies do a reserve study. Do we think they're going to release reserve? If they do we can add that to whatever the data we have. So that's how I think about kind of data that's only by the way there are only like three auto insurance companies you trade that data for. But that's my point, like who's going to build that data? It's not an easy data to build. Who's going to build it? So that's an example of I think a useful data set. I think that's going to become increasingly more important because those are data sets that fall in the crack of quantum and touch it, there's not enough data set backtest to do it, right? You may have thirty years of data four quarters, that's 120 data points times three. Like that's not enough data to do a backtest. But fundamental people can use that as part of the mosaic.
00:18:11 · Brett Caughran, Ying Hua
[00:18:11] S1: And to do that by hand for a fundamental investor, enough there's probably of an ROI to get five bps more accurate, but combining many of those signals together into a forecasting mechanism sort of could scale that precision and forecasting in a... [00:18:30] S0: Yes. Now it's a... Think about every part of that process a lot easier. Like, it took us a few months to build it ourselves, but now scraping is a lot easier, doing analysis is a lot easier. I can schedule run it a lot easier. I don't even have to go like before there's also technical problems of like we run a python program, it spits out a number. How do you pull that number into my excel spreadsheet? My model. Like even that was a hurdle right because I basically had to write a script to open my excel sheet, put in a number in there, refresh it, save it. Like a lot of things are a lot easier now. So I think my hunch is more and more these like alpha edge, things that will be nice to do before. Now you actually have time to do it and it's a lot easier to do it. I think it's going to become increasingly more important.
00:18:30 · Ying Hua, Khe Hy
[00:18:30] S0: Yes. Now it's a... Think about every part of that process a lot easier. Like, it took us a few months to build it ourselves, but now scraping is a lot easier, doing analysis is a lot easier. I can schedule run it a lot easier. I don't even have to go like before there's also technical problems of like we run a python program, it spits out a number. How do you pull that number into my excel spreadsheet? My model. Like even that was a hurdle right because I basically had to write a script to open my excel sheet, put in a number in there, refresh it, save it. Like a lot of things are a lot easier now. So I think my hunch is more and more these like alpha edge, things that will be nice to do before. Now you actually have time to do it and it's a lot easier to do it. I think it's going to become increasingly more important. [00:19:19] S2: Yeah. Can I just ask on that data scraping question? Is that an AI specific skill in the sense that AI is going out to do it? Because that's what strikes me as, like you said, a Python deterministic process that, you know, you go find the site, you know, you clean the data and you put it in a different place. And so can you clarify, like, the AI role in that? Is it that it's just easier to code that into existence? Or is there an actual, like, native AI activity in what you just described?
00:19:52 · Ying Hua, Khe Hy
[00:19:52] S0: Yeah. So I'll split it two parts. I think one, it's easier on the data collection side. So get the raw data. We're probably one of the earliest ones to have an agent that does scraping. It does most of the scraping, human in the loop, we clean it. There is a second part is, it's very similar to how everyone reads 10 ks a little bit differently. Everyone should process data slightly differently. And from that point it's no longer deterministic. So gathering data is just deterministic, cleaning data is deterministic, there is a right answer. But how each team use that data, how each team... What insight are they trying to get from that data is non deterministic. But that part is also I think accelerated by AI. And this is especially important for... We have a lot of even teams inside of large firms even though they have a lot of resources technically but there are still a lot of things you're like it will take me too long to explain it to the data team or by the time I explain to the data team the catalyst would have gone. Now they can just do It takes a few minutes. They can take a first crack and does it worth more time invested? Then if it does, then let me do a deep dive and none I can pass. So I think that part is the AI native part. [00:21:08] S2: And just on that, do you think that Brett and I have had this conversation, like the prevalence of Claude code for non coders? [00:21:15] S0: Yeah.
00:21:15 · Ying Hua, Khe Hy
[00:21:15] S0: Yeah. [00:21:16] S2: Do you, you being closer to it and us being kind of the Claude, the non coder coder types, is realistic that a non coding analyst can go in and be like, Okay, Cloud Code, help me generate this scraper that pulls this deterministic data and so on? Or is that a bit of a pipe dream that is gonna collide really harshly with reality?