Hudson River Trading is a quantitative trading  firm. In order to sort of explain why was HRT  

founded, what were the founders ultimately trying  to do, it's important to set the scene for what  

trading looked like at the time. Trading was  automated to some extent in 2002, but there were  

humans involved. So although a lot of the trades  that happened were on exchange, you imagine like  

sort of pit traders yelling at each other, having  a human in the loop clicking a button to ensure a  

trade gets executed. At the same time, there were  more fully automated marketplaces, Island, ARCA,  

and other smaller electronic exchanges known as  ECN's that were growing in volume. So HRT was  

founded with the idea of like, okay, there's  really some automation that can happen here.  

As you might expect, predicting the price of  anything tomorrow is really, really hard, right?  

And so these models are certainly much smarter  than I am or any single human is at predicting  

markets. But you shouldn't imagine them as being  like clairvoyant. They can't like predict say, you  

know, the price of Tesla will be this tomorrow.  What they can do is systematically, you know, make  

better than random predictions at massive scale on  any instrument you can think of. That is sort of  

the the sort of the core engine of our trading is  being able to give a a better better estimate or  

at least our best guess of the price of any stock  future bond in like a minute, an hour, a day. And  

that is driven by an AI model of our own creation.  The impact that HRT has had on these markets over  

the years has has been pretty noticeable. Not  just HRT, but but all of these quantitative  

trading firms. Retail investors are basically  paying zero to trade into anything. And that's  

very different than how it was back in 2002.  And a lot of that can be attributed to automated  

trading and the service that quantitative  trading firms such as HRT play in the world.

HRT uses C++ fairly extensively at least in  various performance sensitive contexts. So  

that would be our live trading environment and  also various parts of our research environment.  

So why does performance matter? There are two  really important facets of performance. One is  

this throughput or scale and the other is latency.  Talking about latency first, latency is how long  

it takes to get a particular message through the  system. Latency is really important in trading  

because ultimately when when you have trading  strategies, packets can only move at the speed  

of light. They're always looking into the past  in terms of they're reacting to some market event  

that already happened. Microseconds matter in this  space and C++ is a language that lets you achieve  

that latency without without really needing  to micro optimize every line of code. These  

days market data scale is about 20 terabytes of  streaming market data every day that HRT consumes  

that's very high throughput. It's also latency  sensitive. And so these market data parsers sit  

between the exchange and HRT and are translating  market data from this external format to something  

a little bit more standardized that HRT can then  consume. Some people are concerned by the lack of  

memory safety that C++ provides. HRT isn't very  concerned by that. I think that comes with the  

territory is how I would frame it. When you want  to use a language that is as highly performant  

as C++ is, there's some amount of safety that you  are signing yourselves up for in order to achieve  

that level of performance. How do we get over  that hump? Well, we hire strong developers who  

ultimately know how computers work and can map  the language, C++, to what is happening under  

the hood and write safe and performant code at the  same time. HRT has really grown and grown up with  

C++ over the years. If you look at our impact on  the market right now, US equities in particular,  

HRT is roughly 15% of the market. And I truly  don't think HRT could have done it without C++.