Hey, What's up, party people? My name is Sunny Ray. This is the Sunny Ray Show. Back at it again. I've got another awesome guest. His name is Steve Lacy. He's the founder and chief executive of Coats AI, a Troy, Michigan company, building an engine that decomposes financial markets into the underlying waves driving price and then reads the direction of those waves. I mean, when I heard this, my mind almost blew up. Steve, what's up?
Oh, not much. How's it going, man?
So you're building coats AI. What is it? Why should people care? I guess like what's the problem you're solving?
Sweet for now, but we'll go Yeah, we'll go deeper into it later on as well. Go ahead, Steve. Sorry.
Yeah. Oh, no problem. the problem we're solving is that retail traders are not given a lot of good tools to work with. And when I started looking at the traditional technical indicators that they're given like MACD RSI, they all lag very far behind the price chart. and you know tens of minutes you know to 20 minutes and Citadel in the time that you click to initiate a trade and then you get the response could make like something like three or four million trading decisions on their FPGAAS. So how are you going to compete against that if you're lagging 10 to 15 minutes behind? And so I kind of just came at it from a low-level signal processing standpoint to try to figure out how you could be a lot faster with one minute or five minute data.
Wers that super interesting for exits across biotech, fintech, and retail. like what was the through line in the kind of problem you kept picking?
I guess I just sort of fell out I fell into it, right? I dropped out of Stanford before it was cool to drop out of Stanford. who knew that would become a trend. I dropped out in like 92. I did a quarter where I took a bunch of signal processing in 4A transform classes and then I just went and worked in the valley and I just kind of got lucky, right? Like one in 10 startups exit. I started at AF Max. they got bought by like Glaco client Smith and then I went to Afftric. I think they went public and got bought. and then they spawned a sister company called Simics that went public and got bought. and so I grew up in a house where my dad was a high energy particle physicist at NASA and was inventing things in the garage. And so I guess it was just normal to me. And so I went to these companies when I was like 22, 23 and I just kind of started inventing things and I didn't even realize I was doing it in total honesty. and so I think what why I started coach is because I realized that I can invent stuff and I'm pretty good at it and so why not go to where the money is and I did have quite a bit of interest and I did some day trading on the side and stuff like that and so I said why not take this and see what I can do. and the rest is history I guess or maybe it will become history.
Whoa. so signal processing and financial markets are two very different worlds like what put them in the same room for you in the first place.
I started looking at the stock charts, right? And if you grew up when I did and there was no internet and you had to do control system theory where you have a pencil and a piece of paper and you have to derive an equation that takes two or three pages and like by the end of one and you've got like 20 to do your hand hurts and you're pissed off and then that's all in the time domain and they force you to do that and then later on in the semester they're like oh and by the way if you do it in the frequency domain it's two lines and it turns into algebra from calculus. and I so when I started looking and I wish I could show you behind me my normal background I have a Ethereum breakout decomposed spectrally. It's a control system response just like your air conditioner or your cruise control. And so if you've looked at that and done way too much homework on it, it's like I recognize that. And so the rest, you know, I kind of just started doing what I did in the homework and it panned out. Yeah. Yep. I spent eight years in real-time control systems. My dad's also an instrumentation engineer. So, this is all super fun stuff. I you So, you build in Troy, Michigan rather than, you know, New York or let's say the Bay Area. So, what has a Detroit metro given you that you know a coastal address would not?
Nothing I think is the answer to that. I just happen to be here, right? and I think Mike and I have talked about it quite a bit. We're both very open to relocating to San Francisco or New York probably makes more sense, right? but yeah, at this point you know, we're raising money and where we end up is, I guess a roll of the dice at this point, but I think that KOSA would probably make more sense.
And you, before codes, right? So spectral trading was pointed at the stock market through strike stock AI. Like what did what did that first run teach you?
It taught me to believe in it. you know at the beginning I approached the whole thing kind of like Charles Darwin rolling up to the shore of the Gopagos Islands. I didn't talk to anybody in quant finance sort of specifically so that I could just look at it and do what I do and you know as soon as I started decomposing things in equities I started to see some trends and then I went back to first principles engineering where you look for you know say a perfect breakout which our president was happy to provide with a few based on these tweets and so like on April 9th I looked at the breakout that happened after the today's a good day to buy stock tweet and I saw phase lock about 45 minutes before the actual breakout and it stayed locked until about an hour after the break lock after the breakout and I was kind of like okay this is really cool and then I took it to crypto and prediction markets and but it all started with equities like you said and just looking at the traditional stock.
Coat says markets are made of hidden waves. Like when did you first believe that price is a signal rather than noise?
I think it started with the first principles breakout that I just mentioned. But I mean if you think about say a tidal wave or you know let's imagine you're a surfer, you're sitting on the ocean, you're looking for the dream wave, right? But you don't know it until it rises up out of the ocean. And that doesn't happen until it gets close to the shore. If you could strip away, if you could go out deeper in the ocean and strip away all the noisy waves, you would see that big wave coming and you get your surfboard ready and you could ride it. and so, my personal belief is that, the high frequency trading firms in Citadel create these giant waves. We just don't see them until they're erupted. And unless you know it's coming and can catch it as it happens, you can't surf it. And so that's kind of the whole idea here is you're not really predicting anything. You're just seeing you're stripping away all the noisy waves and seeing the big one that's going to move price. This is insane. yeah, this is so cool. Okay. So, the engine was trained and tested on 7,654 markets. So, what did that number I guess? Yeah. Have to reach before you trusted it yourself.
It's actually now we that was when I first contacted you. we're over 12,000 now. And so, we're doing BTC 15-minute updowns on KHI. And when Poly Market becomes legal in the US, we'll be on Poly Market 2. But we've recorded since April. And so we have over 12,000 sessions. which is really good from a statistical standpoint, right? Because nobody believes this, right? You talk to a PhD quant, they'll laugh at you. So having a test bed where it it's really fast and you get a lot of sessions to give you statistical significance is kind of critical. and it also is a rugged place for us to, you know, sort of shake out the infrastructure and everything. So, it's been great.
You know, prediction markets are far newer than equities. So, why start there rather than the stock market you already knew how to sell into? when I so I had a couple products released in equities and started talking to like you know investors and quants and they wanted three month live results. All this is going to come back to why we're doing live results now. But the they wanted three months live results on a 50k bank that I provided myself and then maybe they would invest in me, right? which to me felt like a enormous gamble. and then prediction markets sprung up and they were all the rage and still are all the rage and it seemed like a more investable thing. And then also as a father of five and my youngest is 18 and on call she it felt right to provide more institutional level risk control to help you know younger people who are betting you know on these zero sum you know binary markets to give them a better chance to help them with risk control and things like that. So, we pivoted to Koshi. but we're ultimately going to pivot back to equities as well because that's where all the big money still moves.
Can you give me spectralbot in one minute? Like what goes in, what comes out and like what does a trader actually do with it?
What goes in is BTC from Coinbase. we aggregate our own bars. So, we take trade ticks, aggregate them into 5-second bars, roll those up to five different time scales up to like 15 minutes bars, and then feed it into ML, which sorts through about 800 or 900 spectral fields across all the time scales, builds a model, and effectively gives us up or down with a confidence And so the from a user's perspective, you just look at the BTC 15-minute as it goes and at some point we may or we may not say up and we think it's about 85% chance. I think it's a 90% chance. like that.
Yeah. And you are I guess in early access now like who were the first people to actually find Spectral Bot and like how did they find you? we were on Telegram for a little while because we thought that's where the crypto traders were and I think they are but like we had a lot of people not have Telegram. So this point we just had you go to Code.AI it's a web app like any other web app runs on your phone runs on your desktop. The desktop is much richer. It's like a full studio. and on your phone it's a sleeker you know phone app. and at this point available to customers is just the signal engine. which in all honesty has been interesting because most of the people that are on call sheet are looking to buy a lottery ticket that wins 100% of the time. And what we have is more like a wannabe hedge fund grade signal where if you trade every signal for 24 hours, you're going to make like 3 or 4% a day over the whole thing, right? or 3 4% per trade. and so part of why we've sort of knuckled down on just doing the live trading record is to show that works because if you trade it like if you come in and trade it five times even if the win rate is 85% you're going to lose once and it's going to look bad, right? You have to So the that's sort of the evolution I think. Hey, you say plainly that you're not an adviser and make no profit claims. Like h how do you sell a real w how do you sell a real edge.
Like without promising one?
We're we publish a signal like we never handle anybody's money, right? we are working we have full auto trading under the hood that we use internally but nobody we don't sell that. I think you could sell it on a small trial. I think the CFTC and the SEC allow like little trials of like 10 or 15 people like friends and family kind of thing, but at this point it's just a signal based on historical data and you know there's it's not financial advice in the sense of we're not saying to buy it or do anything with it. It's just based on historical data. It's like a we think the plane's going to you know leave the ground because of the physics. We think the plane's coming down because of the physics. but you know historical results don't necessarily promise future results or whatever the exact phrase is.
Got it. What had to be true in the next like 90 days for you to call early access finished and kind of like open the doors? I think we we've set we didn't settle on a set strategy and execution setup until this past weekend. I think I would want to see our bank grow for a couple weeks to a month on a fixed constant setup. We've basically been building it and modifying it as we go. So we have a live trading record, but it's like every day we've changed and tweaked parameters. We've deployed different models. We've changed the execution rules. And so at this point we want a clean track record where we have changed nothing and just let it run with the current model for two weeks to four weeks and feel really groovy that it's doing well. And then I, you know, you don't want to, you certainly don't want to give access to something that's gonna lose anybody's money. That's for sure.
Yeah, that makes sense. You know, building this out properly takes capital and people I guess like what's the big yes you have to like give and who else is in the room when you give it?
I'm not sure I quite understand.
Like are you guys like is there fun you guys have you guys fundraised in the past already? We we're actively talking to investors. we've raised a little bit of money on a safe. part of doing the live record is that maybe we don't need to raise money. We're hoping that maybe that's the case. Although on the bank that we're starting with even if we had crazy returns, you're it's it probably makes sense for us to raise. So, we're definitely talking to investors and could massively accelerate. This point, it's basically just me and Mike. two guys working 18 hours a day coding their butts off.
It's super interesting. I love how like the insights you bring in. It's like, just makes sense just the waves and how you talk about it, right? And markets and related. Hey, so I guess where do people learn more about Coats and then, where do people learn more about you? I don't know you.
Website maybe.aii. Is it.
Codesai? Yeah, codes.ai.
Okay, cool. And then are you on LinkedIn or are you where do you live? Like if people want to get in touch with you and see what you're tweeting about or blogging about or whatever.
We're on LinkedIn. we're on Instagram. We're on X.
We've tried Tik Tok, but we probably need somebody younger than me to help with that. yeah the standard st and fa we're on Facebook too.
Cool. Perfect. Well, thank you so much for your time and with that I'll bring this one to an end. Thank you all for listening live as well and yeah, hang back one second. Steve.
