Fast Is Not Enough: Enzo Coglitore on Turning GTE Into an AI Trading Co-Pilot

0:00Josh Han Welcome back to the Content Suite in Singapore. Today, we have Enzo from GTE. Before we get started about GTE's new products, do you want to introduce yourself, Enzo? Sure. Happy to.

0:11Enzo Coglitore I'm Enzo, the founder of GTE, where we built the only trading platform that gets better with you. It learns based off of you, your actions, and your behaviors. As you take trades with GTE, our shared AI co-pilot starts to suggest trades, set-ups, recommendations for you to the next trade that You have.

0:30Josh Han Before we get into GTE and this expansion towards AI, maybe we can go into how you got into crypto and what led you into Building GTE.

0:40Enzo Coglitore Yeah, I'm happy to. My background to get into crypto is a bit atypical, but very typical for where I'm from. I grew up in the Bay Area and I grew up in Palo Alto. My dad worked in tech for a long time. He was one of the original pioneers of the data center. He built a lot of the first data centers for Google, Amazon, eBay, PayPal, Yahoo, YouTube, Napster, you name it. I took that company public when I was a kid and then worked at Facebook as the senior director of hardware engineering in 2007, 2008 To 2015. I grew up in this super nerdy household where the internet and everything was very native to me. And I was very into video games, in particular this game Minecraft.

1:29Enzo Coglitore And I realized when I was about nine or 10 years old, like if you could inject JavaScript into like the Minecraft client itself, you could turn off certain blocks and you could see diamonds. And I realized like if I played on online servers, I could monopolize resources. And I became like a trillionaire by using a voice changer and like being a part of this faction in this clan in Minecraft. And I would sell TNT for like in-game currencies. And that's where I kind of got into the world of virtual marketplaces, digital marketplaces. And I kept doing this. I started expanding to World of Warcraft, Guild Wars 2, selling gold, buying gold, counter-strike, counter-strike global offensive. Then you really start to understand middlemen, trust, Online marketplaces, everything else.

2:14Enzo Coglitore And that led me to after I graduated college, I ended up in Eastern Europe with the Ethereum Foundation where they were focused on scaling Ethereum. This is about the beginning of 2023. So I was living with Vitalik in Montenegro for about a month. And that's where I met my co-founder and the rest is kind of history for us. Nice.

2:33Josh Han When did you meet your co-founder and when did GTE start?

2:35Enzo Coglitore I met my co-founder in the middle of 2023 and GTE started right at the kind of end of 2023, going into 2024.

2:44Josh Han And correct me if I'm wrong, but when GTE started, you wanted to work on the biggest marketplace in crypto, which I think is perps. Tell me about the story of like going from perps to kind of the pivot that you're going towards with AI. 100%.

2:57Enzo Coglitore So there's kind of two great like mega trends that are happening right now, right? There's like the mega trend of crypto and tokenization of everything you can trade, like quite literally anything, anywhere at any time. Every exchange is the everything exchange. Hyperliquid is at $96 at the time of this taping. You can trade anything you want, right? You can trade Bitcoin. You can use Bitcoin as collateral to get a loan on Hyperliquid and open up a space expert. And that's kind of always been like the holy grail of finance, right? Is the ability for people all over the world to take a sovereign kind of asset, Bitcoin or a stablecoin and use that to Get exposure to Western capital markets. When we were starting GTE, this was not the case, right? Like a majority of things that were traded on chain were meme coins and governance tokens. And we thought, okay, this is a very similar parallel and analog to the internet in which as you start to scale up a lot Of the base underlying infrastructure, a lot of value starts to move to applications that have real utility for the end users.

3:55Enzo Coglitore And so if the internet's constraint was compute at the time during the first dot com bubble, and then subsequently, the excess compute enabled applications Like YouTube or Facebook, we thought a similar sort of phase shift will happen where the constraint during the last bubble was block space was Very expensive to do transaction on Ethereum. Solana was very unstable. It's hard to get price discovery on chain. When that gets solved, all of the value will move to applications that are able to actually bend these assets. And so when we started GTE, our focus was on how do we contribute scaling technology that actually enables 24 seven permissionless markets around the World. I think we're there. And I think in the other side of that mega trend, you know, is the AI vacation of everything AI in the past, like three To four years has gotten quite literally smarter like than any human that has ever existed ever.

4:46Enzo Coglitore And the rate of change is an exponential. And so we were having this conversation last year, the best AI model app was like probably Claude Sonnet for like GPT five 365 days Later, those models look like toys like they're like, you wouldn't even like want to like use it in any capacity because what it can Do now. Right now, you open AI today put out like I think it was like 106 different like mathematical advancements that they made with their new Model that's unreleased. And that is like an example of like what exists today. Our belief is that by this point next year, we're going to be talking about use cases that AI enables for finance and on chain Activity, that our mind will things that we couldn't have even thought of. And so when we saw, okay, we now have very liquid exchanges, Hyperliquid, Lighter, Solana, the next kind of crash course is that AI Is going to see this marketplace and start to absorb that people will then start to enter the marketplace with agents, where the agents able To read across all the different chains and make an objective decision.

5:48Enzo Coglitore And we're the best places to execute that trade. Maybe you want to trade Bitcoin perps for you, that could be Hyperliquid. If you have capital on chain, maybe you want to trade meme coins for some people, that's Robinhood Chain, the AI has no opinion, it Will just execute what the demand of the user is. We're very focused on making that next leg of access possible for the end consumer.

6:10Josh Han That makes a lot of sense as perp exchanges meme coins kind of decentralized and a lot of the old pains that we had in Like 2021 have gotten solved by the new exchanges. You guys want to give the ability for everyone to utilize AI. I think you're correct in that the equivalent of a PhD is in our pocket now through Astra. And you would have to previously pay an employee or a quant, let's say a six figure salary, but you can do that now with Like a $10 subscription. Where do you see that enables GTE? What kind of trading or different types of strategies are possible now that weren't let's say a year ago, 100% like one of our

6:49Enzo Coglitore Taglines is Wall Street intelligence in the palm of your hand, where you're able to quite literally have, you know, I would say it's actually Smarter than a PhD, right? Like I would say like the smartest person in the world is probably Terence Tao by raw IQ. And AI is doing math that like Terence Tao cannot do anymore. And even he's kind of conceded defeat. If you look at like there's like Jelani Nelson, who's like the CS professor at Harvard that went to Berkeley. That was actually like Jeff's like teacher for when Jeff was like a TA. I don't know how goes class even he's joined Anthropic, right? It's like, okay, like, yeah, academia has like basically conceded defeat to AI, right? And what does that mean in practice? What it means is that people who have some semblance of Edge are going to have an even bigger Edge and find the limits of Their Edge much faster in terms of relation to the markets and trading.

7:38Enzo Coglitore So if you, you know, we call this kind of like a one man firm have the ability to quantify information and stack that with Qualitative beliefs, and then use that to make a trade, you're able to do that much faster with an agentic tool suite. And so the way that we think about our product is it's not a product where you can just tell the AI to make money. It's almost the inverse. The AI is waiting for you to tell it what your ideas are, to make money and validate it for you. Use back testing, use simulations, track flows across finance, buy-in and Hyperliquid, be able to traverse the world of on and off chain Finance and help you make a decision with confidence much faster. So what this means is that people that previously would have had to set up entire desks, quantitative developers, infrastructure teams, execution engineers can now Rely on a suite of agents through a singular agent that we call Edge to define and determine what their Edge is in relation to The rest of the market.

8:32Enzo Coglitore As they start to trade more and more Edge starts to remember who you are and what you did. So let's say we made a trade where we made a lot of money. I like Meta Muse as an example, I bought Meta Muse or I bought Meta stock when when Muse came out. The AI remembers that now. So it's going to ask me, why did you buy Meta? Like what did you learn about the world that you were right when the market opened on Monday? If you bought it on a Saturday, you got gifted with like a God candle. And the next trade that you make, it's going to remember that you've made that trade. And then by the 10th or 11th trade, it's going to find like the next setup for you. Because what we've done not only in building this harness for the AI is we've also given it a computer. So it has the ability to be omnipresent. It's looking and searching for things while you're asleep.

9:19Enzo Coglitore And it never stops. And so we believe that just today, those are a couple like example use cases by next year, it's going to be able to surface Insights better than any human ever could have and do research better than any human ever could have.

9:32Josh Han Got it. So using GTE's new product Edge, you have the equivalent of having a CFO analyst, on-chain analyst, or really whoever you want on your Team managing your finances, or at least your portfolio. Tell me a little bit more about how you are using it. Am I using this as a new harness or am I bringing it into my harness? What is the first product suite for Edge look like? So we built our own custom harness.

9:59Enzo Coglitore And the way it works is it's built inside of GTE. And what we do is obviously we're this Labs Corporation and we license that product to GTE Company. And on our front end, now we have a representation of the harness. And so what you're able to do now is you're able to set it up with your past. Give us your wallets, give us your past brokerages. Let us read and create a trader profile for you and create a universe where it's uniquely you and the AI. So the AI goes in and says, Hey, you bought Hyperliquid at $4. What were you thinking? Most people did not treat that as a consensus bet, obviously, because Hyperliquid ripped to $96. So you had alpha in relation to the rest of the market. Why? Tell us why. As we start to build this world with you and the AI, the AI starts to get better as it starts to understand what you Like to look for, what your alpha is, what your Edge is.

10:49Enzo Coglitore And so the reason why we had to build our own harness is because general purpose models and general purpose harnesses do not really apply Well to domain specific industries like finance. The reason why AI is so good at math and science is because there's a verifiable reward. There is an outcome that's deterministic where after smashing 10,000 agents, it's able to solve a millennium problem. Finance is a bit more like time dated. You make a trade, you are either right or wrong in that trade, but there are a series of time has to happen before you Get rewarded. If, for example, you could buy hype at $4, you only get rewarded the $4 to $96 move by holding it the entire time. What was your conviction? What was your belief? And so, but as the AI starts to collect more and more of the data in that time series, it starts to get better operating Within that.

11:37Enzo Coglitore But the only solution that we found was to create a harness and subsequently what we call an ontology, which is the relation of objects That the AI has the ability to understand. It's like a term for metaphysics essentially means, You know, the AI knows what a tool is, what a sandbox is, what a back testing is, what simulation is, how to search Twitter, How to know when to call what tool for each specific function. So if a user says the word backtest, it calls our custom backtesting tool, and it knows the ability to do that. If a user says the word perps, it determines whether or not they're saying finance or Hyperliquid. That is like where a lot of the value lies. If you ask ChatGPT or Claude, it gets you pretty close. But it eventually suffers under the noise.

12:22Enzo Coglitore And this is why companies like Palantir exist. That's where my co-founder worked. You have the ability to actually surround the model with what you just called a harness and really exemplify its abilities for the specific domain That it exists in.

12:36Josh Han That makes a lot of sense. I've trained my OpenClaw and Hermes agent now to do specific crypto things, but it still hasn't reached the level that I want. Another concern that I have as a potential user is really just, how do I know my funds are going to be compromised? If this agent has my private keys, I could wonder if it could get injected, prompt injected, and send my keys or like my funds to someone else. What kind of security measures are you guys thinking about as trading becomes more agentic? A hundred percent.

13:08Enzo Coglitore So the first is, we try and keep it extremely locked down. We were exploring things like TEEs where we don't even see the inference that you have in relation with the AI. I think Muse is a perfect example of this where you're able to cross the chasm, make it so that even if it has sensitive Information, A human still has to approve all of those different actions. So even if somebody tries to prompt inject it in relation, that already means they have access to your device, which means that you've been Compromised pretty hard. And so for things that are API based, things that somebody, if they're able to get access into the system, we try and build as Many guardrails as possible that require the agent to actually have multiple levels of approval. This is where this ontology comes inside of. Eventually at some point, there's these benchmarks where the agent is supposed to make money.

13:57Enzo Coglitore It always ends up like, boom. It's because the agent, in isolation, it just eventually suffers from noise. And what we've done is we spent a lot of time making sure the agent isn't just going to eat your balance into a perp And in position because it thinks it's a good idea and it just gets wrecked. Because the worst thing that can happen here is not just like, oh, you have the wrong response. The worst thing that happens is Ken Griffin takes all your money. And you really don't want that to happen. That's even worse in some capacity than having databases deleted or having the AI send a text to your ex-girlfriend. This is the next level of that. So we've spent a lot of time applying, what does AI look like in enterprise? What does AI look like in terms of getting traders to trust it?

14:43Enzo Coglitore If we're able to get a trader to trust it, and these traders are what I call security aware. Then we're able to probably cross that chasm for everybody else.

14:54Josh Han Very nice. I have so many questions because you've used this more than anyone else now. Whether you call it an agent or Edge and how much Edge it actually gives you, you only know. What are some of the surprises that you found testing Edge that maybe people should look forward to or I could try once it comes Live?

15:16Enzo Coglitore Yeah, so Edge is really good at crossing strategies. AI is really good at copying. So let's say, for example, you have some sort of macro trader that you love, Stanley Druckenmiller. It's like, okay, Kevin Warsh worked for Stanley Druckenmiller. He's now the new Fed chair. You tell the AI, okay, go use yt-dlp. This is like the Python YouTube downloader. Download all the Druckenmiller's transcripts for everything he's talked about. Download all of the Kevin Warsh when he was getting selected by Trump to be Fed chair post. Understand, like, okay, in what world did he actually decide to raise rates? What world do we live in right now based off what these guys have said this surpass people?

16:03Enzo Coglitore This is macro. Now, let's find assets that are correlated to this. So, okay, like if they're raising rates right now, but ultimately in the long term, like Druckenmiller and Warsh, they want more volatility. What assets are going to benefit from that? And then from within that basket of assets, what assets do I know? Crypto, stocks, tech stocks, et cetera. When do I know assets are mispriced relative to what other people think? Then how am I able to then start to apply more quantitative thinking to when's the proper time to buy? Do I have another trader that I know that times trades really well that are shorter term variations? Say I'm long term bullish on tech, but I'm looking for an entry. When is an entry? What qualifies an entry? Well, there's another trader. Let's say I like Ansem as an example. Ansem, by the way, has great trading knowledge from like 2021 before the meme coin shit started, like really smart.

16:51Enzo Coglitore You can take some of those where they're talking about specific entries, then apply that to the thematic sector. What I've done now is I've just taken information from the Fed chair, the best macro investor of all time, and I've combined that with Like these crypto guys that are really good at timing flows and in and outs and momentum plays. That gives me the ability to now see both sides of the game. I don't have to just buy Nvidia and chill. I can actually accumulate properly based off what I think is the best entry point relative to my greater grand thesis. That is where I see Edge and AI really starting to like kind of [unclear] users' behaviors. It would have taken you a whole team of analysts to study just Druckenmiller, all team of analysts just study momentum trading, whole team Of analysts just study execution on top of that, break thematic sectors, et cetera.

17:37Enzo Coglitore But like an agent can do this all in one run. It's able to do this and then remember it. Then if you make the trade and if you were right, agent knows you have alpha relative to everybody else. Example of this is like I did a trade where I bought Intel right after Claude Code came out because I was like, okay, CPUs Are going to rip. The agent helped me because you have a lot of emotions as a trader. Where you lose money is you either overtrade or you get flooded out of the position and then it rips. It's like the classical story of like it's like a dollar away from my liquidation price. I close the position and then like Zcash ripped like $1,000 and I would have made a million dollars. As a human, you're subject to these biases. AI does not suffer from these same biases. What it's really good at is coaching you to stay calm. When I bought Intel, it was at like I think $40 and then it went back down to like $25 and I was like, should I close this position?

18:29Enzo Coglitore Am I wrong? It was like, tell me your fundamental thesis. What did you think here? Why did you think that? Well, if you're right, then the risk to reward calculation is greater than the drawdown another $10 if you're playing a long term game. It like basically soothes my emotional state. Three months later, it ripped to like $20. It's like, oh my God, I will never not use AI to structure my trades and think through my trades with me because it's going To catch my faults as a trader and it's going to magnify my wins and magnify the things that are good about me. And that creates a really good co-pilot to exist within a financial market.

19:07Josh Han I think co-pilot is a really good way to describe this potential product, mostly because trading is a solo game. It's more similar to golf than it is a tennis or like game with other teammates. And now you have the equivalent of like talking to other agents in use. Do you imagine this to be some kind of like UI where I have my Druckenmiller agent and then my portfolio agent and then my On-chain agent? Or do you kind of imagine it to be one agent that kind of has context to everything?

19:38Enzo Coglitore Yeah, it's a good point and a good question. I actually am not bullish on like the [unclear] setup. I think it's just too hard to manage. Like, you know, you have your chief of staff agent or like my manager agent. I'm much more bullish on like what we've done, create a singular agent with the ability to like nest sub-agents but hide that from You where you have just your Edge, you can name it, and then you can interact with it. But it has the ability to spawn up sub-agents and orchestrate different workflows on demand, but that gets all abstracted away to the end User. This seems to be the next frontier of agent experiences. You can see everybody kind of going for this, OpenAI with [unclear], Meta with Muse. And so our belief is that this is going to continue to increase and we want to specifically create that agent for finance.

20:27Enzo Coglitore What we don't want is to show you, hey, like, here's like a base straight agent, or here's like a Druckenmiller agent. Like, that is just like noise in relation to like actually working with an agent that's doing research that only you know is real. That only you know is good information because you took the trade and you got that data yourself. What we don't want to be is like telling you like download GTE and like use our agents and like we're going to make you Money. Like it's actually the complete opposite, right? It's like download GTE, use our agents. If you make money, you're going to make way more than you would make just trading on [unclear] or the same thing on trading On Robinhood Chain on Solana or anything else.

21:06Josh Han I mean, from everything that Enzo has told us, this does seem like the most comprehensive agent solution for not only on-chain trading, but Really portfolio diversification and management. When can people expect and try it themselves?

21:20Enzo Coglitore Yeah, so probably at the time of like this recording coming out, I expect like we rolled it out to like 20 to 50 kind Of VIP traders. Think like your favorite names on CT or like people like head of trading at like major HFTs that are just like trading on their PA. We want to roll this out cohort by cohort. We're very interested in collecting data and see what people use for what we call financial intelligence. So what are they asking the AI to do? What questions are they asking it to actually do? Is it sandbox? Writing code? Is it back testing? Is it research? Is it looking at Twitter? Is it checking flows across exchanges? All these different things. Then we want to scale that out. The thing about this product is like pretty compute intensive. And so we don't view ourselves as going for the mass market here.

22:09Enzo Coglitore Actually, I'll be very opinionated in saying that. Like if you want to just like monkey see, monkey do, there are really good social trading platforms for you to trade on. We are much more interested in the person that does $10 billion worth of volume that wants to $100 billion worth of volume while retaining Their alpha and their P&L gets magnified in exponential for that. If that's you, then I think you're going to love using GTE and you can find me. Nice.

22:35Josh Han It does seem like you're building the product for yourself as well as like the team members that I've met from GTE for the people And fans of GTE. Thus far, is there anything else that you want to share or let know what's happening?

22:46Enzo Coglitore Yeah, I think that what I would say is it's taken us a long time to get to market. And there's a good reason. We only wanted to put a product out there that was fundamentally Accessible only on GTE. And for the longest time, I think GTE was very focused on creating something but faster. Solana, but faster. Hyperliquid, but faster. And what we realized is like speed is only a single like lever that we can pull in that, you know, Hyperliquid is pretty Fast. Solana is pretty fast. It can definitely get faster, but fast is not enough. What we need now is intelligence. And what we've built is a product that you can only access on GTE. And that's why we're excited to roll this one out now because we've been using it internally. We've turned over a couple hundred million dollars worth of volume just on our PAs in the last couple of weeks.

23:36Enzo Coglitore We've been making like money doing this. And what we think is going to happen is that this is going to be kind of like a genie out of the bottle. When people start doing this, it's going to exponentiate and magnify volumes where we're on top of any chain or any exchange. We're able to go and talk to the most sophisticated traders that have setups there and work with them to magnify their setups. Our main focus is to really exponentiate volume across the board. We're big believers in the perpification of everything, the tokenification of everything. We think AI is going to love crypto if it doesn't already already. And we think that this is going to be the next leg to turn the volumes from like one to two trill to like 10 Trill. Because when that gets combined with RWAs and tokenized assets, by the end of the decade, these agents will be helping you process trillions of Dollars worth of volume.

24:25Josh Han Very nice. Very exciting updates from GTE. Enzo, thank you very much for your time and we'll wrap it up. Thank you. Yeah. Nice, man. Thank you.