# A Hedge Fund as a Function Frank Long · AI Engineer NYC 2026 · October 13 [Watch recording](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24466s) [06:47:47](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24467s) Hello [06:47:47](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24467s) Exciting to be here. My name is Frank. I work at Coatue which is a technology investment firm across public and private markets [06:47:54](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24474s) So, in our business, what you really need to do is be able to take high conviction bets [06:48:00](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24480s) on big themes and watch those play out. [06:48:03](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24483s) I think, honestly, a career is the same way. [06:48:06](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24486s) And last year when I joined Coatue, the bet that I wanted to make was I think that agents [06:48:11](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24491s) will have an even bigger impact in investing than they did at coding. [06:48:15](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24495s) I know that's a big statement. [06:48:16](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24496s) We're all here because of the transformative impact that AI has had in coding. [06:48:21](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24501s) But I really believe it strongly. [06:48:23](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24503s) And today I want to talk a little bit about the framework that we use to actually realize [06:48:27](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24507s) that impact. [06:48:29](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24509s) So, at a high level, you can think about a hedge fund as a function that takes in two [06:48:34](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24514s) parameters, context external and context internal, and spits out one tuple. [06:48:40](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24520s) Selection, size, timing. [06:48:44](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24524s) stock, how much, when. [06:48:47](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24527s) And I think this is a good framework to actually ground our work, which is we are trying to [06:48:54](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24534s) solve this compression problem of how do you take all the information in the world and [06:48:58](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24538s) spit out this one simple tuple and the product that we provide is essentially a time series [06:49:05](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24545s) of this tuple that constructs a portfolio. [06:49:08](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24548s) let's start from the top, which is the inputs. [06:49:11](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24551s) simple philosophy. [06:49:13](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24553s) We use our screens, we look at stuff, we do work. [06:49:17](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24557s) Every single thing that a human can see in pixels on the screen, an agent in your building [06:49:21](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24561s) should be able to see via API. [06:49:23](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24563s) The analogy I use is imagine you brought the smartest people ever, the greatest investors [06:49:29](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24569s) alive and you brought them into your company but you gave them a computer without logins [06:49:34](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24574s) to 75% of your services. [06:49:36](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24576s) It's almost immaterial how smart this system is if it doesn't have the tools to actually [06:49:41](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24581s) action that. [06:49:42](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24582s) So the two parts, external context, no surprises here, think the news, prices, filings, all [06:49:51](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24591s) of that should be easily accessible and not via dragging and dropping, copy and pasting, [06:49:56](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24596s) like directly via API. [06:49:59](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24599s) Internal context is just as important, right? [06:50:03](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24603s) So this is your own firm situation. [06:50:06](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24606s) So that might be our current positions, potential liquidity and risk requirements we have around [06:50:10](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24610s) certain names, and past case studies or history with a certain sector or stock or management [06:50:19](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24619s) team. [06:50:20](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24620s) That is all super important. [06:50:21](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24621s) So this is the foundation, and I think, no credit to me, Coatue has been an amazing leader [06:50:26](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24626s) on this for over a decade. [06:50:29](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24629s) It is both technological and cultural. [06:50:31](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24631s) We have an amazing data science team, amazing processes, amazing systems that bring all [06:50:36](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24636s) this in, and just as importantly, we have a culture that is super, super rigorous and [06:50:42](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24642s) built from the ground up to capture all this insight. [06:50:45](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24645s) For decades, we filed every single note about every single trade, about every single decision [06:50:50](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24650s) we made or didn't make about what went well and what didn't go well, and that has poised [06:50:54](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24654s) us to expose us to AI. [06:50:57](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24657s) If you think about it, I probably can't just take all of those things, jam them into a [06:51:03](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24663s) context window and get my magic three numbers and go home. [06:51:07](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24667s) It can't be that easy. [06:51:09](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24669s) You probably have to break down this function a bit further. [06:51:14](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24674s) So maybe the first step is I take in that external and internal context and I gather [06:51:20](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24680s) it. [06:51:21](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24681s) I want to gather it, I want to filter out what is relevant and not relevant. [06:51:25](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24685s) Okay, well, I pulled some news articles about the MBA that probably isn't material to stocks, [06:51:30](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24690s) but this is related to local permitting that might impact data centers. [06:51:36](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24696s) There is a lot of nuance in this gather step. [06:51:41](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24701s) you will see probably the next step is I need to analyze that information and form [06:51:44](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24704s) a view. [06:51:46](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24706s) I need to think about what that means for the markets, public and private, and then [06:51:54](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24714s) I need to take that view and turn it into a decision. [06:51:57](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24717s) A decision object has these three parameters, these three sub pieces. [06:52:03](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24723s) You can see every single step has an input and an output, and the input for the subsequent [06:52:10](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24730s) step is the output that the previous function generated. [06:52:14](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24734s) I think as an engineer, this is a way for me to conceptualize the problem in a way that [06:52:20](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24740s) makes a lot of sense. [06:52:21](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24741s) But even this, it really can't be as simple as writing three functions. [06:52:26](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24746s) So let's just, to really appreciate the scope and understand the full complexity of the [06:52:33](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24753s) problem, let's look into the first step, a tiny sub piece of a tiny sub piece of just [06:52:38](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24758s) gathering information and let's do a case study on interpreting the news. [06:52:45](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24765s) So it's probably no secret to you guys that news from Asia and Korean media outlets has [06:52:52](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24772s) been super important this year because semis, AI, have been a huge, huge theme in the market. [06:52:59](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24779s) Huge theme. [06:53:00](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24780s) And I think this is an area where AI is poised to have a lot of impact. [06:53:05](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24785s) Reading Korean language news better than our completely non-Korean speaking staff. [06:53:12](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24792s) I think I have a good thesis that this might be a good use case. [06:53:15](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24795s) So you can imagine the first step is I have articles. [06:53:20](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24800s) I need to collect the articles. [06:53:23](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24803s) This [06:53:23](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24803s) could be an MCP, this could be a CLI, license the data feeds, clean the data feeds. There are a lot of ways to do this and a lot of dimensions of freedom for how you can even just write this simple piece. How do I chunk it when I pass it to the LLM? Do I want to pull the images as images? Do I want to pull the images and then turn the images into text? There are actually a lot of knobs here. [06:53:48](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24828s) I almost think about those knobs as hyperparameters in a traditional machine learning system. [06:53:53](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24833s) I think about each of these knobs as something for me to tune and optimize against. [06:54:00](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24840s) Obviously, filtration, we talked about it. [06:54:04](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24844s) There are probably Korean NBA fans, right? [06:54:06](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24846s) There is probably a bunch of stuff that is not necessarily material to the names and [06:54:14](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24854s) themes and stocks in our portfolio that really, really matter. [06:54:16](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24856s) So I want to make sure that I filter those out. [06:54:22](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24862s) There is this translate slash extract slash should I summarize, should I extract this [06:54:27](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24867s) into some sort of structured form. [06:54:29](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24869s) Again, so, so many ways to do this, so many ways to experiment, how to expose information [06:54:34](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24874s) to the model, and one of the reasons this is hard is it keeps changing. [06:54:39](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24879s) And obviously checking. [06:54:40](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24880s) Checking our own internal notes, our own internal views, potential conversations we have had, [06:54:46](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24886s) meeting recordings against these things. [06:54:48](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24888s) We actually need to cross-reference those things. [06:54:51](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24891s) Actually the idea of pulling information from disparate sources and cross-referencing them, [06:54:56](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24896s) that's super powerful. [06:54:58](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24898s) So you will see on the right here I have code, LLM, LLM, code, person. [06:55:03](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24903s) There's many, many ways to run these functions. [06:55:06](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24906s) I think about, oh, yeah, and finally, update. [06:55:10](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24910s) We update our views. [06:55:13](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24913s) There are many ways to run these functions. [06:55:15](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24915s) I think of each line as having a runner, and the most obvious one is code. [06:55:22](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24922s) Code is very good. [06:55:23](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24923s) It is fast, it is cheap, it is reliable, so for something like if the prices to earnings [06:55:30](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24930s) ratio is less than 15, I should just use deterministic code for that. [06:55:35](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24935s) Runner number two is statistics, classical machine learning, and you can imagine fitting [06:55:42](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24942s) the line to a regression for a specific pattern, this is very powerful because you can model [06:55:47](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24947s) much more complex relationships than just pure deterministic code. [06:55:52](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24952s) And for even more complex relationships, the reason we are all here, LLMs, AI agents, that's [06:55:58](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24958s) another type of runner. [06:55:59](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24959s) So for something like summarization, that's the state of the art. [06:56:04](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24964s) It's probably totally impossible to build a bunch of if statements that can summarize [06:56:10](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24970s) better than a modern LLM. [06:56:12](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24972s) Of course, people. [06:56:15](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24975s) People are important runners in the system. [06:56:18](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24978s) Imagine a function meeting the management team of a company. [06:56:22](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24982s) I probably don't want to send an LLM or an AI agent in to do that. [06:56:28](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24988s) I probably don't want to send an if statement into that meeting. [06:56:31](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24991s) I think there are many cases where because of judgment and the ability to navigate the [06:56:35](https://www.youtube.com/watch?v=akEqiXFo3GM&t=24995s) real world and all of these things, a person is the best runner for a given piece of the [06:56:40](https://www.youtube.com/watch?v=akEqiXFo3GM&t=25000s) function. [06:56:42](https://www.youtube.com/watch?v=akEqiXFo3GM&t=25002s) I think this is a super, super important point. [06:56:45](https://www.youtube.com/watch?v=akEqiXFo3GM&t=25005s) There is no glory in automating things for the sake of automating things. [06:56:49](https://www.youtube.com/watch?v=akEqiXFo3GM&t=25009s) The goal is not to take as many things and make them LLM as possible for us. [06:56:54](https://www.youtube.com/watch?v=akEqiXFo3GM&t=25014s) The goal is for any given piece of this function, what can I do to maximize absolute quality? [06:57:03](https://www.youtube.com/watch?v=akEqiXFo3GM&t=25023s) And if you actually think about the function, it probably looks something like this. [06:57:07](https://www.youtube.com/watch?v=akEqiXFo3GM&t=25027s) It's probably not just a linear statement of a bunch of different things. [06:57:11](https://www.youtube.com/watch?v=akEqiXFo3GM&t=25031s) There's probably all kinds of branching logic and all kinds of ways to actually fork this. [06:57:16](https://www.youtube.com/watch?v=akEqiXFo3GM&t=25036s) So maybe I want to collect the information and then translate it, and my main goal is [06:57:20](https://www.youtube.com/watch?v=akEqiXFo3GM&t=25040s) to extract quotes and check those against sources to make sure that these are actually [06:57:24](https://www.youtube.com/watch?v=akEqiXFo3GM&t=25044s) the things that people said, and that's my approach. [06:57:27](https://www.youtube.com/watch?v=akEqiXFo3GM&t=25047s) But that might be totally wrong. [06:57:28](https://www.youtube.com/watch?v=akEqiXFo3GM&t=25048s) Maybe what I want to do is I want to de-dupe and then I want to filter and then I want [06:57:33](https://www.youtube.com/watch?v=akEqiXFo3GM&t=25053s) to pull external figures or maybe I want to do something totally different. [06:57:36](https://www.youtube.com/watch?v=akEqiXFo3GM&t=25056s) I actually want to, I'm passing in a note first and I want to find other confirmatory [06:57:41](https://www.youtube.com/watch?v=akEqiXFo3GM&t=25061s) data points or maybe I'm trying to red team. [06:57:43](https://www.youtube.com/watch?v=akEqiXFo3GM&t=25063s) There's a million different ways to configure this. [06:57:46](https://www.youtube.com/watch?v=akEqiXFo3GM&t=25066s) So yet again, more hyperparameters within the control flow, right? [06:57:52](https://www.youtube.com/watch?v=akEqiXFo3GM&t=25072s) So many ways to think about just this very simple problem of pulling, filtering, summarizing [06:57:58](https://www.youtube.com/watch?v=akEqiXFo3GM&t=25078s) aggregating the news for LLMs. [06:58:02](https://www.youtube.com/watch?v=akEqiXFo3GM&t=25082s) And each of these boxes probably has an optimal runner, right? [06:58:09](https://www.youtube.com/watch?v=akEqiXFo3GM&t=25089s) So the optimal runner might be deterministic code for collection, right? [06:58:15](https://www.youtube.com/watch?v=akEqiXFo3GM&t=25095s) I'm pulling last 24 hours for a given source. [06:58:19](https://www.youtube.com/watch?v=akEqiXFo3GM&t=25099s) That's deterministic code. [06:58:21](https://www.youtube.com/watch?v=akEqiXFo3GM&t=25101s) Filtration? [06:58:23](https://www.youtube.com/watch?v=akEqiXFo3GM&t=25103s) models maybe. [06:58:25](https://www.youtube.com/watch?v=akEqiXFo3GM&t=25105s) LLMs? [06:58:25](https://www.youtube.com/watch?v=akEqiXFo3GM&t=25105s) statistical methods, a combination, right? [06:58:29](https://www.youtube.com/watch?v=akEqiXFo3GM&t=25109s) So this probably understates the complexity. [06:58:32](https://www.youtube.com/watch?v=akEqiXFo3GM&t=25112s) The reality is most things are multi-runner. [06:58:34](https://www.youtube.com/watch?v=akEqiXFo3GM&t=25114s) Most things have deterministic code plus LLM plus statistical method or human with LLM [06:58:39](https://www.youtube.com/watch?v=akEqiXFo3GM&t=25119s) in conjunction. [06:58:41](https://www.youtube.com/watch?v=akEqiXFo3GM&t=25121s) reality is it's probably combinatorics between my four buckets and there's even more [06:58:46](https://www.youtube.com/watch?v=akEqiXFo3GM&t=25126s) buckets than that. [06:58:48](https://www.youtube.com/watch?v=akEqiXFo3GM&t=25128s) Because within the LLM category I can tune the model. [06:58:52](https://www.youtube.com/watch?v=akEqiXFo3GM&t=25132s) I can choose big model, small model, medium model. [06:58:55](https://www.youtube.com/watch?v=akEqiXFo3GM&t=25135s) I can tune thinking effort. X high, high, medium, I can tune harness. I can choose the harness. I can [06:59:02](https://www.youtube.com/watch?v=akEqiXFo3GM&t=25142s) choose which skills load with the harness. I can choose which hooks load with the skills. [06:59:05](https://www.youtube.com/watch?v=akEqiXFo3GM&t=25145s) So yet again, more hyperparameters within my search space. I have an absolutely enormous [06:59:10](https://www.youtube.com/watch?v=akEqiXFo3GM&t=25150s) amount of hyperparameters within my search space. And this is just for this one tiny problem, right, [06:59:15](https://www.youtube.com/watch?v=akEqiXFo3GM&t=25155s) of just filtering the news. So the way to think about breaking down these problems is what can I [06:59:22](https://www.youtube.com/watch?v=akEqiXFo3GM&t=25162s) actually compartmentalize in a way that we can understand and evaluate. [06:59:26](https://www.youtube.com/watch?v=akEqiXFo3GM&t=25166s) That's what I'm hill climbing against. [06:59:28](https://www.youtube.com/watch?v=akEqiXFo3GM&t=25168s) Otherwise I'm just tuning these things. [06:59:30](https://www.youtube.com/watch?v=akEqiXFo3GM&t=25170s) I'm okay, that should actually be a machine learning system instead of an LLM. [06:59:34](https://www.youtube.com/watch?v=akEqiXFo3GM&t=25174s) That should be a little model that we train. [06:59:36](https://www.youtube.com/watch?v=akEqiXFo3GM&t=25176s) Oh, no, that shouldn't be a model at all. [06:59:38](https://www.youtube.com/watch?v=akEqiXFo3GM&t=25178s) That should be an if statement. [06:59:39](https://www.youtube.com/watch?v=akEqiXFo3GM&t=25179s) Or, oh, man, LLMs are really, really bad at translating that to portfolio impact. [06:59:44](https://www.youtube.com/watch?v=akEqiXFo3GM&t=25184s) It should really be a human on that step, but they're using a deterministic system to [06:59:48](https://www.youtube.com/watch?v=akEqiXFo3GM&t=25188s) help them search and pull the information. [06:59:51](https://www.youtube.com/watch?v=akEqiXFo3GM&t=25191s) It could be all different things. [06:59:52](https://www.youtube.com/watch?v=akEqiXFo3GM&t=25192s) And if you don't have a north star to hill climb against, you're just kind of, you're [06:59:57](https://www.youtube.com/watch?v=akEqiXFo3GM&t=25197s) just fidgeting around. [06:59:59](https://www.youtube.com/watch?v=akEqiXFo3GM&t=25199s) And I think I cannot convey enough that this is just one tiny thing we do. [07:00:04](https://www.youtube.com/watch?v=akEqiXFo3GM&t=25204s) Imagine every single data source that we pull in, there's probably more than just three [07:00:09](https://www.youtube.com/watch?v=akEqiXFo3GM&t=25209s) transformations to get them to the intermediate output. [07:00:11](https://www.youtube.com/watch?v=akEqiXFo3GM&t=25211s) So the actual graph of Coatue and funds like Coatue, it actually probably looks, oh, not [07:00:18](https://www.youtube.com/watch?v=akEqiXFo3GM&t=25218s) like this. [07:00:20](https://www.youtube.com/watch?v=akEqiXFo3GM&t=25220s) is my screen saver. [07:00:22](https://www.youtube.com/watch?v=akEqiXFo3GM&t=25222s) The actual graph probably looks something like this. [07:00:25](https://www.youtube.com/watch?v=akEqiXFo3GM&t=25225s) It is probably massively complex. [07:00:28](https://www.youtube.com/watch?v=akEqiXFo3GM&t=25228s) In each of those nodes, we just talked about at least ten hyperparameters with at least [07:00:34](https://www.youtube.com/watch?v=akEqiXFo3GM&t=25234s) 20 different values for them. [07:00:36](https://www.youtube.com/watch?v=akEqiXFo3GM&t=25236s) So the search space is basically infinite. [07:00:39](https://www.youtube.com/watch?v=akEqiXFo3GM&t=25239s) And figuring out how to actually lay this out in a way that you can comprehend both from [07:00:45](https://www.youtube.com/watch?v=akEqiXFo3GM&t=25245s) technical perspective, but also absorb culturally, organizationally. [07:00:49](https://www.youtube.com/watch?v=akEqiXFo3GM&t=25249s) That is the great challenge of applying AI in investing. [07:00:54](https://www.youtube.com/watch?v=akEqiXFo3GM&t=25254s) And beyond investing, what we're really talking about is creating a state machine that can [07:01:00](https://www.youtube.com/watch?v=akEqiXFo3GM&t=25260s) take information from the world and understand it well enough to make coherent decisions. [07:01:05](https://www.youtube.com/watch?v=akEqiXFo3GM&t=25265s) There's use cases for this far and above, wider than just investment firms, in policy, [07:01:12](https://www.youtube.com/watch?v=akEqiXFo3GM&t=25272s) in non-profits, in drug discovery, in corporate strategy. [07:01:16](https://www.youtube.com/watch?v=akEqiXFo3GM&t=25276s) I think one of the most exciting things for me [07:01:19](https://www.youtube.com/watch?v=akEqiXFo3GM&t=25279s) about being in forums like this [07:01:20](https://www.youtube.com/watch?v=akEqiXFo3GM&t=25280s) is talking to other people in different spaces [07:01:23](https://www.youtube.com/watch?v=akEqiXFo3GM&t=25283s) with different problems that kind of congeal [07:01:25](https://www.youtube.com/watch?v=akEqiXFo3GM&t=25285s) on the same problem space. [07:01:27](https://www.youtube.com/watch?v=akEqiXFo3GM&t=25287s) I'd love to keep talking about this. [07:01:29](https://www.youtube.com/watch?v=akEqiXFo3GM&t=25289s) I'll hang out in the hallway afterwards. [07:01:30](https://www.youtube.com/watch?v=akEqiXFo3GM&t=25290s) Here are my socials. [07:01:31](https://www.youtube.com/watch?v=akEqiXFo3GM&t=25291s) And thank you, everyone, for your time. [07:01:34](https://www.youtube.com/watch?v=akEqiXFo3GM&t=25294s) Thank you.