Build a Claude AI Trading Assistant — Step-by-Step Guide & Transcript

[00:03] completely wrong. They think it's about prediction, getting AI to tell them where a stock is going. Or building a bot to trade for them. It's not. The real edge is operational. How quickly you prepare, how clean your

[00:17] process is, and how effectively you can execute on your ideas. And until recently, that kind of infrastructure was only accessible to firms with teams behind them. That's what's changed.

[00:30] Every trader on our desk is spending hours working with AI every day. But the best traders aren't looking for cheat codes. They're building tools. Not one tool, a system. An AI trading assistant that can support

[00:46] any part of their workflow. Prep, research, execution, review. So we sat down with a trader on our desk, Justin Speiro, a teammate and a friend, who's been deep in this process. And we pushed him on it. Not just what

[01:01] he's doing, but how to actually approach this from scratch. How to use tools like Claude. How to think about building your own assistant, and how to turn AI into something practical inside your trading without needing a coding background.

[01:17] What he laid out is a very real, very accessible blueprint for how traders can start doing this themselves. If you've been curious about AI, but haven't known where to start, this is a conversation you want to hear.

[01:31] If you want to compete, you need to understand how the best traders are leveraging AI to build their own edge. And this is where it starts. Here's our conversation with Justin. Welcome to the Trading Floor Podcast

[01:44] Welcome to the Trading Floor Podcast episode 15. Should I Should I announce our our guest, Tim? Should we just start there? Okay. So Justin is a trader on the New York City desk.

[01:56] He's a veteran. he's a mentor, he's also a teammate, and he's a tremendous guy, actually. Like really really great demeanor on the desk, really great character. And one of the things I enjoy about

[02:10] Justin is that he's always willing to share. Um and that comes down to market views or information or trades or anything. to have you. Thanks for being

[02:23] >> so much for having me, guys. Appreciate the kind words, Garrett. Uh been a pleasure being your teammate for the last what, 7 years? We started We started >> He's been around for like 4 or 5.

[02:37] Little behind you. >> Yeah, and I'm always willing to share as long as there's plenty of liquidity for everybody. >> there is, right? >> Yeah.

[02:50] >> Yeah. Yeah. So, all right. So, so look. About a month ago, I think it was, we started getting this email in our our inboxes generated by you,

[03:04] Justin, that um summarized like all of our research and and pretty much like put into perspective the stocks that are probably the most poignant to watch and what's going on that might move the market on

[03:17] any given day. And it's been really really useful. And this is something that you created generated by AI. And you have pretty much no coding experience.

[03:31] Right? >> I have little little coding experience. Um when I first joined SMB, everybody. And for some people, it really clicked.

[03:46] For some of the traders, they did some of that in college. Um I had no experience with Python. I took a course on Python, but even after that I was like, "Eh, this this just isn't really for me."

[04:01] >> For somebody who has very little coding experience, um, I I just think it's interesting how quickly you you got this together as soon as you

[04:14] a project. What were the first things that went through your head when you started with Claude Code and realized that you could build something like this, which clearly is is

[04:30] not even close to like something that you could have built before? >> Yeah, it would have been impossible for me to build something like this. And I'm still trying to think about before I joined the podcast, like,

[04:44] how did this even happen? And I think that we got an email in our inboxes like a couple months ago. And it had something to do with Claude. Um, it was from one of our bosses and

[04:57] they're like, "How are you using this new technology?" And I'm like, "You know what? I feel like I could be using it way more. I could be using it way better." And one of the first things I thought about was

[05:10] of the first things I thought about was all the emails and the research that we get in our inboxes. And after just, you know, talking to a with not just Claude Code, but Claude

[05:25] I was like, "Wow, some traders are building these dashboards that organize their playbooks. Some people are building dashboards that

[05:37] TradeerView." And this idea just clicked into my head, well, if I could get this AI to parse through my emails in the morning, um I could be extracting a lot more information than I already do.

[05:55] Uh and that's sort of where it all began. what are what are we talking about? What is the project? >> All right. So, should we take a or can we do a little look here? Can I share my

[06:08] >> Yeah. >> Would you like to gain the biggest edge a retail trader can get? All of our daily and weekly in-house trader meetings are now available to you. Just head over to

[06:21] smbtrainingfloor.com to find out more. >> Here is what I built. Um and it's still you know, it's still an ongoing process, trying to make it better every day,

[06:35] seeing where it trips up, but I feel like it's in a really good spot. Um it gives you a macro rundown. Starts it gives you a macro rundown. Starts with what the AI thinks is the most

[06:48] with what the AI thinks is the most important overnight development. Gives you an economic calendar rated by importance. So, 7:00 a.m. MBA mortgage applications, you know, low importance.

[07:03] events. Tax Day, historical retail buy the dip dynamic. Uh talks about all the earnings today that came out already in the pre-market when I ran this.

[07:18] And then I just let the AI do its thing, and I ask it for what it thinks is going to move the most today. So, it gives me a couple ideas, and then also, you know, why it will move on this

[07:33] catalyst. And you know, something that I'm keep adding to this, and I think that's what makes it kind of fun. what makes it kind of fun. If a company gets a contract,

[07:47] you could ask it, you know, what percent of its yearly revenue is this contract worth? And if it's like 50 or 100 or even more than 100, you know it's super significant. Uh and that could help.

[08:03] >> You could even do it by market cap also. >> Sure.

[08:16] names it picks up from those emails, it will just put the ticker and a quick Uh and then a calendar for the week ahead. And this was ahead. And this was like my first project and it took me

[08:31] maybe like 20 hours, but this is like the first big thing that I over, it could probably take an hour,

[08:44] maybe even less. You just have to know how to prompt the >> So, clearly with building something like this, the data that you have access to is really important and I want to touch on that in a second, but first

[08:58] you're scanning some emails that have really good information and I want to make sure that everyone watching this um who doesn't necessarily have access same thing out of what we're talking about here today,

[09:12] um which is really like how to get going and how to turn in turn it like what is in your head into the actual tool that we're building with the AI. So, like when you sat down to do this,

[09:25] what like what did you open up Claude? There's no like instruction manual. So, like for somebody who like has never done this before, like what would you if they're like, "Hey, like I got 2 hours." Pretend you're talking to my dad cuz

[09:39] he's he's actually a trader by the way. He know he understands market regimes. He understands um sectors and rotations and different asset classes, but like he can barely like get into his email, all right? Like he's not a tech-savvy

[09:53] person. So, what would you tell him what would you tell him to do for the first 2 hours of trying to tackle this and create some kind of AI assistant for trading?

[10:05] >> That's a great question. So, the first thing that I would tell your dad is think of a daily process that you have that you're doing manually right now. And

[10:20] And that's what you want to try to automate. that's what you want to try to automate. So, you know, let's say your dad wants to build

[10:34] a trade tracker. A dashboard where he could upload his trades at the end of the day that looks like TraderView um and gives him all different statistics. So, the first thing that he would do is he would open

[10:50] would do is he would open Claude Code Work a dashboard that tracks my trades. I want to be able to upload my trade data

[11:04] at the end of the day and then I want you to take that data and show me what my win rate is, my um win-loss ratio, average losing trade, average winning

[11:18] trade, uh you name it. Whatever stat that he could think of." And the first thing is like frame what it should look like and what exactly you want it to do.

[11:32] telling Claude exactly what you're already doing in your process. That's kind of like your starting point. >> That's the starting point. And the more articulate you could be, the better.

[11:47] So, I've often used Chat GPT to help me better express exactly what I want. So, it's like you're using one AI that's good at one right? A large language model that's good at one thing, answering questions,

[12:03] to you into Claude. daily and weekly in-house trader meetings? There's no bigger edge retail smbtradingfloor.com to learn more.

[12:17] >> Okay, so this touches on a really important aspect of this, and something that I think that I mean, a lot of us understand, but maybe somebody like my dad, again, who's like a very savvy trader,

[12:30] like he knows what he's doing, um might not get the big shift that's just with AI. Like, why are we even talking about cuz like some people might be you talking about AI?" It's like, well, okay, yeah. Um

[12:46] AI's been around for a bit now. We've had Chat GPT. But like, we've spent a lot of time just just talking to these language models back and forth and getting advice and getting information. But

[12:58] up until very recently, like this has not been agentic, meaning an actual agent on your computer doing stuff. And like, that's the big inflection that has catalyzed all of these things and everything that

[13:13] we're asking you about. It's like, how to get the agent, the AI agent, Claude Code or Claude Co-work, to actually build these things for you and to manipulate your computer and to do these things

[13:26] for you. Like, that's that's the whole reason why this is starting to explode. And that Tim, when did that When did that start? >> A couple months ago. Cuz honestly, yeah, before it you were just using AI every

[13:38] time you had a hang up on your own manual project that you were doing. bug. I'm going to like ask ChatGPT how to debug it and then you go in and you just actually code yourself and you fix it.

[13:51] you're doing is you're using ChatGPT in like the old-school way to really fine-tune the language and the prompts and like how you're communicating with Claude Code or

[14:05] Co-work, I assume the two that you're using, that then become the agent for the task that you're trying to complete. >> Exactly. So, I don't know if you guys have heard of the term vibe coding.

[14:20] >> Oh, yeah. It's all over Twitter. >> I have not How have I not heard of that? >> It's kind of like when you just open up Claude Code or Claude Co-work and you start, "Hey, like build this dashboard, do this." And I recommend everyone

[14:33] starts off that way. But you're going to inevitably hit a and Claude's just not going to be doing the right thing. the right thing. And that's when I recommend you go to

[14:46] ChatGPT, just the regular, you know, LLM, that's how I use it at least, and start asking it questions. And then you're going to start to learn And then you're going to start to learn how you should actually be prompting

[15:00] Claude. Like what an actual good prompt looks like. And you're going to just start to think differently after that. Um and you're going to start to say, "Oh, this is how I need to frame this. I need

[15:13] it to be a little more specific. It doesn't understand what I'm saying um because I worded it wrong. And this is how I need to word these prompts moving forward. So, I recommend like you just play around with it, tell

[15:29] it what you want to build. Um and then once you hit a bump, Um and then once you hit a bump, use Chat GPT and you'll start to learn. >> Can you use the the language model in Claude to do the same thing? Like you

[15:42] code is like a whole different conversation than than the basic Claude? >> Well, you don't need both. I just like Chat GPT, plus I pay the money for it.

[15:57] something. >> That's That's fair. prompting because honestly like that dashboard is sick. Like the end product is awesome, but you're never going to get there

[16:11] without good prompts like you're saying. What I do with Claude code is there is a plan mode. So, before like it starts building anything, you switch over you toggle over to plan mode where you're just brainstorming, coming up with the

[16:23] project brief. And you can ask Claude to say, all right, like we're just brainstorming here and trying to figure out what the game plan of attack is. And brief, you like how you prompted it, it looks like it knows what you want to do

[16:38] and accomplish, then you switch over to build mode. But I think the science and the magic is in the prompting like you're saying. And for anyone getting started, you really you want to spend as much time as

[16:51] ahead. Although like the shiny object, which is the dashboard, is great. Like you said, experiment, play around. >> Something that's been on my mind as I dive into this world as well. And I mean that to my like you and I have been

[17:05] using I mean we all three of us have been using like AI for you know, ever since Chat GPT came along, but clearly this this new frontier is moving very And we're we're all doing our best to keep which I think you have to be doing.

[17:19] and one of the things that crosses my mind when I'm tackling these things is that like you can have it build anything. So, it could end up being very generic if you don't teach it about yourself.

[17:35] I'm doing these. I'm like well, I need to do a really good job of teaching it to do a really good job of teaching it about myself, so it's personalized to me. So, the way that I trade, like my playbook, my checks in favor, my rules,

[17:51] the way I grade these things, the things that actually matter to me in in my own trading. So, again, like we you know, take your pre-market example. Like it's very different, in my opinion,

[18:06] explore. I'm interested to hear what you guys have to say about this, but it's very different to say here are a here's a bunch of data, here a bunch of of emails. I want you to like scan all of these and then summarize them every

[18:20] me versus okay, I have all these emails. There's tons of data that are briefing that it's a brief about every single morning. And I'm going to be giving you these, but

[18:33] also what I'm really interested in doing what I'm really interested in doing is taking this information and trading breakouts and trading mean reversion setups and trading catalyst plays. And

[18:46] so, here are my playbooks. And here are all the checks in favor and all the rules and all the past examples of really good trades. So, now you know, as a trader a little bit more about like what's

[19:01] out of this brief. You're like teaching it your business and your system. What do you guys What do you guys think about that? >> I think that's big. I think that goes in like we said with the prompting or the

[19:14] more detailed just makes the end product even better. And I've done very similar. >> If you do that, Garin, I'd be really curious to see what the outcome of that is and what I've learned is you could take this

[19:29] as far as you want. Um >> [clears throat] >> but it doesn't need to be something super complex and extravagant like that. Like you could even do something as

[19:41] simple as build me a small risk widget that sits on my screen and I could put my daily stop in there and it reminds me what stop in there and it reminds me what size A+ is, what size A is, what size B

[19:56] size A+ is, what size A is, what size B is, what size C is. that sits there. >> Not to deviate too much from the combo, um but on the personalization, what I've done too with like this

[20:12] automated trading analysis process that Garin and I had is it's took about 4 to 5 hours of me going line by line through this very large script that I have, which is how we analyze, and literally going line by line to tell it why every

[20:27] line is there. And then it knows that script like the back of my hand. It knows every single parameter that goes into it. And without it, I think it would have probably taken 10x the amount of time after. Where if

[20:39] beginning, they end up being bugs down the road. So, I would be curious, Justin. Yeah, I know with this um premarket research report, like you said, it's ongoing. Yeah, I would be curious if you keep

[20:56] fine-tuning and tweaking like what specific with it, if that would be a separate section, too. Maybe there's a area of actually looking at stuff that's consolidating, you know, in range and

[21:11] compressed. Or I don't know. What what what is that Are there any um I guess improvements on the top of your mind that you're working on with it? >> That is a really great question and the

[21:24] next step would be to get like a an API key from Polygon, which is now massive. So, it has real-time market data and it could look at where the stock actually is

[21:39] on the daily chart. It could see where the market is and how far the market's ran or how far how long the market's been consolidating. And it could probably, if you feed it playbooks like Garrett suggested,

[21:52] it could definitely start to see, only if it has live market data, like oh, this is set up for one of Garrett's favorite breakout trades. This

[22:04] uh ticker that popped up in this email today that has this news of a massive contract with Meta, it's set up for a breakout and I know that Garrett likes breakout trades. It could probably feed

[22:18] you that. Um definitely more complex, definitely going to take a lot of time, >> How would you take what you've done, which I mean frankly is is taking a bunch of um great proprietary like research emails

[22:34] >> that we we all get. I mean, I'm always joking about how like if you send me an we get we get like a hundred research emails to them. Um like not everybody has that, right?

[22:48] And that's a really good data. And so, some of this is like down to it's only that's that's kind of always been the case, right? That's why Renaissance Capital like crushed it back in the days cuz they had I think they were the first

[23:01] bar data and stuff like that in their back testing. >> Here's a mind-bending stat for you. SMB traders have 20x to 40x higher odds of success than independent traders. Find out why and how to greatly increase

[23:18] smbtradingfloor.com. >> So, for somebody for like a retail trader who wants to do what you've done, they might have to take I think it's totally possible, but you might have to take a slightly different

[23:33] approach and cater it again to like your world. So, I I mean, things on my mind like one, you know, news source, right? you probably don't need an AI to summarize just one news source because

[23:48] they're probably using AI to summarize all that to you, right? So, you've got one news source. So, how can we find good data as a retail trader, um

[24:00] and and and be able to get some kind of morning report that will add value, um if you're not getting hundreds of emails in your in your inbox?

[24:12] emails in your in your inbox? >> Fair enough. Fair enough. And we do get sell-side research from the street that a retail trader would not have access With that said, there are plenty of research emails that

[24:27] there are plenty of research emails that we get that are just popular we get that are just popular amongst not just traders, um anyone in the finance industry. Right? So, we get something from

[24:39] Bloomberg. We get something from Wall Street Journal. We get something from Vital Knowledge. Vital Knowledge is probably, if you go around the desk, how many traders are in New York? 30?

[24:53] all what your favorite morning newsletter is, I guarantee you half of them are going to say Final knowledge. And that is anybody can sign up for that Um Hammerstone reports is another one that

[25:09] Hammerstone reports is another one that anybody could get. The point is MarketWatch. Um and I know there's so many other ones >> Trade the news. >> just get Trade the news. Boom. If you

[25:24] just get six of them Maybe I'm a slow reader, but you'd have to wake up at 7:30 in the morning and start reading. Especially if you wanted to like synthesize all that information you'll be reading till 9:30. So, even

[25:39] though we get more newsletters and special research, whatever you want to call it I would still only read like three or four of these emails. I would just have to pick what my favorite ones are.

[25:55] Right? Cuz we're so limited on time. Um so, I would say even for a retail trader, maybe they're not getting 25, but if they're getting six, they could still save themselves an hour if they do this. And I think that's the

[26:09] whole point. >> Yeah. I've I've got I've got some ideas I mean, you might as well just sign up for 20 newsletters. If there's one thing for someone listening to this to do, it's just sign up for

[26:24] more than one newsletter. Cuz like you're right, there's a ton out there. So, what about like the 8Ks? Which are the the That's That's where these these news

[26:38] services will get the earnings report information, right? Am I getting that Something. There's something that's available to the public. So, you can have it grab those.

[26:51] Right? You could You could grab all the earnings call transcripts from all the Um all the company websites. So, like in terms of earnings reports, you could have it do its own

[27:05] do its own thing. >> It's pretty easy to find out who's going to be reporting the next day before the open. And you could I mean, that's a great idea, Garrett. Like See, this is you just have

[27:18] to have an imagination. >> A lot of this information is is available to the public that is just disseminated by these news sources because they are spending a lot of time

[27:31] gathering all this information and then feeding it in bite-sized forms to us, but that data comes out um first, and it's it's all available to the public. And so, that's that's one idea. Another

[27:46] idea would be what about like you got some market filters on TradingView or something like that that are like your favorite sort of pre-market scanners?

[27:59] Could you Could you incorporate that into the morning report as well? Like, hey, look at my pre-market scanners as well, and like cross-check that with the >> It's pretty interesting, like gap up scanners, like volume overnight

[28:15] scanners. Yeah. Like if you give it access to look at your screen every morning from 8:30 [clears throat] to 9:00 and just watch this screen with my scanners as you're getting the news.

[28:28] Um I think you might be able to do that. And scanner's just like a whole 'nother Like my my trading station, my screens, my screen real estate, it just looks totally different than it did 2 months

[28:42] ago because now a scanner that would take me a month to build or I wouldn't even be able to build it. Um, you could build it in 10-20 minutes. And that's build it in 10-20 minutes. And that's for any trading software that you use.

[28:56] It's huge, cuz like Tim and I have wasted so much time. I wasted is a strong word, but we've spent so much we've spent so much time coding because, you know, like all this time before AI just you and you're just

[29:12] I mean, I'm losing years off my life just sitting here till 8:00 8:00 p.m. in the dark, like last one to leave. The only light on in the office is my computer screen, and it's just all Python, and I'm just trying to figure

[29:26] in our >> to take us weeks to and like a week and a half to find one bug to debug. >> Yeah, it's just it's insane. And now, you know, you don't even have to know how to code to to whip this stuff into

[29:41] shape like pretty quickly. So, it's it's it's it's powerful, man. Like, it is unbelievable that this is possible now. So, like I'm kind of jealous now of all the people who like don't know how to code.

[29:55] Because you never had to feel that pain that we had to feel. >> [laughter] >> But I'm sure with that knowledge, you could build something pretty cool

[30:08] >> Yeah. Well, what's So, what's your experience, Tim, having hit this pretty hard? And also having that background of like kind of doing this a lot of this stuff, trying to do it manually, like building

[30:22] automated models, algorithms, scripts, market views, like all from scratch, all from writing Python to now using Cloud Code in your work. Like, how's that been? >> Yeah, I honestly think the past

[30:37] experience expedites the work I'm doing now. Where like Justin said like same with my screens like so many more scanners and alerts that used to take you know, a week to build now you can do like he

[30:51] said in 15 minutes. What I think is the biggest benefit and anyone could do this and this is more on the alert side the scanner side and actual automated models. But once you have any template for a script like

[31:06] you've an automated model once you write one that's a template and you can then give it to Claude. So it's not just going from scratch but it's very much working off of what you already built. So just to give an idea like last Friday

[31:19] I was having a ball but you know, me and Garrett like we used to I don't know how least a week for us to write a model. Probably even like longer. I wrote 10 on Friday for us. Oh my god. Yeah.

[31:34] But what helped was that I had a template where I was like all of the script is mainly just features and data that we want to collect. So don't touch control center where this is the only thing you want to change where it's just

[31:49] a bunch of different steps to enter steps to exit and like that's your playground to work with Claude. And then you give it very specific again accomplish like what's your entry signal what's your symbol universe what's your

[32:03] exit and then it's able to do it super quick. So you're giving it our old scripts to >> Yeah. to kind of learn like here's our process here's here's how we get them to work. And then you're isolating the part

[32:16] that just changes the trade logic to change the strategy while keeping all of the data collection part the same. Did I just get that right? >> Exactly. Exactly. And I will say like

[32:29] and I'm I have a question for you after to get back to the premarket stuff Justin. But But have noticed like even with as great as Claude is, like it doesn't get everything right the first go. And that

[32:43] has happened especially on the automated trading front. And it's good at debugging, um but certain things I've noticed too, I have to go in and be like, all right, this is how we do this. Like for

[32:55] example, like it would try to do its own way of collecting minute bars. And I'm like, no, I have a really good way how I collect minute bars and I use list through those bars. And other debugs, sometimes I like to

[33:09] and look at it myself. So I would say that's another area where like some necessary, you know? >> I think that's one of the fun things about it though, right? Is when you tell it to do something

[33:23] and it gives you something back and it's not quite right and it's like, oh, it's interesting that it did that. And then you correct it and spits something else back out. I don't know. I find that part of

[33:35] the process like enjoyable. Like it's interesting to me. I don't know how you feel about that. >> No, I love that. I mean, I was talking with a friend recently too and there is an aspect of AI if you let it that takes

[33:49] away the satisfaction that you get from working on projects cuz it like eliminates like the creative thinking, the like grit work, the problem solving if you let it. But like you're saying, it is the fun part. Like seeing where

[34:03] did you go wrong? Like what were you thinking compared to how I think? And I don't know if you did this with the pre-market report, but you can ask it, And then it'll give you a response, too. So you actually see like the inner

[34:16] So you actually see like the inner workings. Do you ever ask it like if if if you were me, how would you be doing this differently? Like see if it just has any ideas for what to do.

[34:31] >> After it knows a little bit about you. >> Yeah, you were you know, like maybe it of. I mean, that's something I would always do with the language model with ChatGPT. Because like one of the ways that I used

[34:44] to use it, well, I still do, but you know, just the language part. I Tim and I would be working on a strategy or we'd be trying to come up with like a market filter. Um so, we're trying to get some signal going.

[34:57] get some signal going. And so, you have to define these things that are happening in the market and that kind of forces you to like use math and come up with like an indicator that describes whatever that dynamic is.

[35:10] And I'd be like racking my brain trying to figure out like oh gosh, like how do how do we express how do we get a value that expresses this, right? That whatever thing we're trying to look at. And

[35:22] as I'm like trying to figure that out and basically like telling the language model what I'm doing and like here's here's the trade. This is kind of what I want to express. How can I get like the most robust sort of result with some

[35:35] formula? Sometimes I would just stop and be like, all right, let's just zoom out for a second. Forget everything that I just said and just think about the goal here. Like how

[35:47] would you be attacking this? Like are there other things other avenues that like I that am I just driving us into a hole right now? Like like like and just totally like crashing the plane? Is there a much easier way to be doing

[36:01] this? And sometimes like it'll list a bunch of you know, BS, but sometimes there's like a good idea in there. Which is so interesting to me. Coming from artificial intelligence. I mean, this thing's wild, man. I don't

[36:13] I don't know. I don't know how they do it. It's a new frontier. >> Sometimes you actually have to tell it to slow down. Like hey, pump the brakes. >> Yeah. >> starts going on a tangent sometimes.

[36:26] >> Yeah. You got to train it. >> Have you heard this study of tens of millions of traders which shows that only 1% of retail traders actually make Don't be a statistic. Visit smbtradingfloor.com

[36:41] >> All right, so getting back to the pre-market research, like if we were a retail trader starting today, sign up for the newsletters tell it what you're trying to accomplish. In your daily process, what

[36:56] do you do and what are you trying to automate? I'm curious, you do those two steps for you, were there any like major pain in the early days of working on this project that you remember? Where or

[37:09] something you had in there that you had to remove that just wasn't working? it grabbed the VIX from one of my emails.

[37:22] And it had something cool in the corner, like it would tell me the VIX level. It would like give me its opinion about the volatility in the market. And I was like, oh, I told it like, oh, I really like that. Keep that and every

[37:36] one of the emails mentions where the VIX is, show me that. >> [laughter] >> that gave me the VIX level again. So,

[37:48] the next day it was like 25 and then it was like the next day lot and it was still the same VIX level and I'm like, wait, there's just no way

[38:01] that the VIX is still at 22. It was at 19. So, like you have to just you have to tell it, if that doesn't exist in the emails that you parsed, you

[38:13] have to exclude it. And then you might just want to get rid helpful. So, I mean, it trips up, right? And this is like version three of this dashboard. So,

[38:29] on this dashboard, the earliest entry is March 16th. I've been doing this since I don't know. I guess the end of February, early March.

[38:44] And you know, it just takes a while to get it right. And you know, every morning in the beginning, you would find like, I don't have time to do this now, but in the afternoon

[38:56] fix this. And I think that I enjoyed that part of the process. Like that's what made it fun for me. Like I >> And [clears throat] that's also where you can't just like take it at face

[39:10] value. Where like it makes sense for a pre-market report that like you have to be careful of bad data. Where it's getting information or if the Vix. So, that seems like the one yeah, thing

[39:23] to focus. >> Yeah, and it's not the end all be all. Like sometimes it'll think a top idea is to trade Snapchat, but Snapchat is blacklisted for me. I will never trade

[39:36] ever [laughter] traded it? >> You should tell Claude. never just go into and say, oh, it's top three ideas are are

[39:50] these stocks, but so that's what I'm going to trade today. It's just nice to have an idea of what it thinks might move the most. And there have also been move the most. And there have also been times that I've doubted those top ideas

[40:02] runners of the day. So, >> That's a pretty interesting point, too. How often do you see that this pre-market report leads to really good opportunities for your

[40:16] trading? If you had to like estimate it? >> Maybe three times a week I'll get a really great idea. And it worked

[40:28] though. That's a big That's a big deal. >> Yeah, that's a big deal. And other times it's just made me aware of news that you know, I wouldn't have known.

[40:42] Like today for instance, it grabbed something from one of the emails. you know, it could be something that I got from WSJ, it could be something that

[40:54] MarketWatch, it could be something that I got from Final Knowledge, Hammerstone. >> [clears throat] >> and it said that yesterday when Trump was talking about the war in the Middle East, he was referring to it in the past

[41:09] tense. As if it's like already over to him. And I thought that was just interesting. >> And it's helped me save myself time. like under value. You know, it's like it it's not always that like oh man, like

[41:24] this is this has got so much edge and this is helping me like trades and all that kind of stuff. But like if I can just

[41:40] um pretty thorough in the morning into like one hard-hitting piece of like document or piece of information or email like in your case like you send us

[41:54] this email. You're basically like distilling everything into like here's what's important and if it just takes me 20 minutes to read that rather than spend an hour and a half like going

[42:06] through like all of this other stuff and trying to figure it out for myself, that time I'm just looking at chart blowing through charts and like preparing like that I'm looking at. So I think like anything that I can do to

[42:23] save myself time is like huge because I mean worst case scenario we just make to like hang out with our loved ones more. >> That's like that you know what I mean? Like it's it's great.

[42:37] >> Or just work on another project. >> Worst case scenario I think is you save >> Worst case scenario I think is you save yourself an hour a day somehow. >> Yeah. Right? You get to you get to hit the gym more. You get to like Tim says

[42:49] get maybe it saves time in the morning so you can spend some extra time you know building that automated model on that

[43:02] that playbook that you're trying to add to your quiver or something like that. So >> Yeah and I would also just like one thing that I definitely want to mention is

[43:14] always get questions from the community about ARVOL. Like we don't have the same ARVOL that you guys have on your platform. you guys have on your platform. Um

[43:30] using TradingView Cloud Code it's amazing at coming up Cloud Code it's amazing at coming up with custom studies.

[43:43] thing. And I know Garrett in the past you've built really cool things with for me. Um I just could never do it.

[43:57] an ARVOL study. Right? Um you know take today's cumulative volume and divide it by the cumulative volume over the last

[44:12] five sessions at that same time. And there you have our ball, right? Um so you could come up with creative custom scripts for like ThinkorSwim and TradingView that I probably think that

[44:29] you know, 80% of its current users haven't been able to do. I mean, I I 80% of users. >> 100% especially with ThinkorSwim cuz it's just horrible. I'm I'm sorry, ThinkorSwim and

[44:42] >> [snorts] >> But it's it is just like that coding language is just so cumbersome and like I I've I kind of started there, so I cut my teeth on it, but I feel like

[44:55] every time I'm actually trying to code something there, it's like a massive It's like, oh, something's not possible. Let me see if I can figure out a workaround to make this happen.

[45:07] And it's like pulling teeth, but the part of that process that I enjoy is the creative part of it and actually getting the result so we can use it. So, if you're thinking like you just had you

[45:20] ball, right? There's a number of different ways to get our ball, but that's one of the ways to do it. And that that's a creative way to think about how to express whether something is doing unusual volume today.

[45:32] Which is awesome. And then getting that result is awesome. Those things those both of those things to me are fun. Going in to the code and trying to figure out some workaround to get ThinkorSwim to think script to kind of

[45:46] like that's the part of the process that I enjoy the least. And that's the part of the process that takes up the most amount of time. Right? So, to me that could save right

[46:00] right there might save like two, three, four, five hours of my life. Like over and over and over. >> On the topic of next projects, do you have anything big in the pipeline,

[46:14] but other than that. >> I came up with a really cool scanner >> [snorts] >> and and I and I'm I'll I'll just share this cuz anyone that's been watching up until now, they deserve to know.

[46:28] >> that's true. >> [gasps] >> Yeah, right. Um so, it's just [clears throat] a low float scanner.

[46:40] It looks at stocks that are at least 20 cents but are below $10. Uh it excludes OTC stocks. Uh it excludes OTC stocks. Uh average cumulative volume needs to be

[46:54] Uh average cumulative volume needs to be at least 100,000 shares. 2%. So, that's like a good low float filter. Uh and then you sort that filter by dollar volume.

[47:10] you know, as we're recording this, it's April 15th. The past 5 days, there have been some wild low float runners. Like today, we had Bird, the IRD. Yesterday, we had Snail. I call it

[47:26] Snail, SNAL. Um they both went nuts. And there have been a couple others throughout the last week. They all had something in common. Um they were trading the highest dollar

[47:42] So, I'm sorting that list by dollar volume. >> [snorts] >> and the leaders on that list are trading like a billion dollars, 4 billion dollars,

[47:57] almost 2 billion dollars as giving me the runner of the day. So, my next project is to back test this

[48:09] >> [snorts] >> Um now, I'm going to need some API for real-time data, and I just want to figure out exactly how accurate is this. you know, if I give it those parameters and I say,

[48:24] um for the highest dollar volume ticker, you know, just what do they close on average over the past month or give me just give me the list uh and give me you know, the percent

[48:39] of all the you know, for the last 100 names that were number one on that list. Or, you know, my wording might be a little off, know, my wording might be a little off, but something like that.

[48:52] be sharing with uh our whole team, team KFit's. I like that avenue with AI, too, where you're basically just getting a ton of different stats for certain setups to then have conviction to trade

[49:06] discretionarily. And I've been working on uh a little gap up, gap down single stock script, where basically if you see the other day B's gapping up and you're curious how B behaves usually on similar ATR gaps, it

[49:21] just like separates where it's like in the past year it's gapped up at least half an ATR 10 times, on average it closed green this amount, closed red this amount, the you know, average move from open to high is this, open to close

[49:36] have conviction to then hit the keys yourself, which I think's a cool avenue. >> Exactly. Like the whole point of the study would be give me that extra conviction to size a

[49:50] >> [snorts] >> Tim, the way that So, the way that you're doing that is essentially creating a back test, right? Like, you're taking a back testing platform and you're quickly creating these

[50:04] scripts that represent the back test that you looking at the the stock that you're >> Yeah. So, basically, like, the way our

[50:17] platform runs will be different um than everyone else's, but the idea is you're looking for any one symbol you want to get the last year's worth of data. And then you're just computing a bunch of stats off of it. So, the whole, like,

[50:32] framework is all based on that last year worth of data for all stocks when it gaps up. So, like, if I was going to redo it now, you would want to see all of the stocks that gap up half an ATR, get a huge CSV of those, and then you

[50:47] have your stuff to work with. Go into the cloud, and then you tell it what stats you want. And then, if you just type in a certain ticker, you'll be able ticker. >> Okay.

[50:59] >> Very doable with just a Excel file, too. Like, you just have your Excel, and >> Yeah, and then it's like an instant instant data. >> With the with the prompting. >> And the idea is, like, if you see BE,

[51:13] and BE was gapping up, like, I think it was, like, 1.6 ATRs, and if you see that happened, I don't know, maybe it happened eight times, and in all eight it closed red, then you're like, all right, this doesn't really behave well

[51:25] hands-off, you know, let it come to me. I'm not going to really try to press anything right off the open. Uh and again, it's just that added conviction. >> Justin, you know what else those tickers

[51:37] had in common? They were all an- They were all animals. >> [laughter] >> God damn. >> I I wonder if they ate wonder if the AI picked up on that one.

[51:52] >> Now I have to ask the AI all animal tickers. Um Well, this has been awesome. Thanks for you on again. This is it's just fun to hang out with you.

[52:05] >> Yeah, of course. Thanks for having me. Yeah, I had a great time. on this because I know this stuff is changing really quickly and we're all we're all using it and I'm you know, the stuff that we end up building is

[52:18] powerful. A because we're using it more and B because the technology keeps getting better. Um but anyone who is using

[52:30] cloud code or co-work or any of these other models um for their trading in interesting ways, like please share it in the comments. We'd love to hear about that. Um and have a great week of trading and

[52:43] Um and have a great week of trading and we'll see you next week.