Line Your Own Pockets

Michael: 00:00

Alright, everyone. Welcome to another episode of Line Your Own Pockets. We're gonna talk AI today. Right? It's gonna be something that I think it's unavoidable.

Michael: 00:08

That's gonna come up periodically, but I've been getting into it. I've been getting deep. I've been going I've been going hardcore. So I think it's gonna be a good topic. Might actually end up being a a couple episodes or a series because I now think it's starting to get smart enough and with Claude code and things, ways to do it right.

Michael: 00:28

Whereas before, there was a lot of traders just typing in make me a bunch of money, and then that's not the way to do it. So we're gonna talk about, I think, some good uses that traders can have and and systematic traders can have using AI.

Dave: 00:40

Yeah. So in the green rooms over the course of the last few episodes, you've been filling me in on what you're doing. You've been asking me, okay. How can I get this set up to do some interesting things with Amber broker? You were doing some things with RealTest.

Dave: 00:55

It just sounded really cool. We definitely thought we gotta do a whole episode on this. But and I've been doing some stuff behind the scenes on my end for mainly for programming, though. And it is really incredible. But what you're doing was pretty interesting.

Dave: 01:17

And so why don't you why don't you take us through how you got the idea to start doing it and what you're actually doing right now?

Michael: 01:26

Yeah. So I'll I'll go down the list. So I was using that that OpenClaw Claude bot thing, but I've actually just, over the last week or so, taken that off my computer because I've so so everything I'm talking about now can, as of recording this, be done with clot code. Right? Then before, you know, I I'm sure anyone who was on Twitter for any length of time saw this OpenCloth thing just go completely viral, where it could control your computer and all of that.

Michael: 01:52

But

Dave: 01:52

So why don't you go back and, like, tell us how you started using Cloudbot? You know, of course, that's named renamed itself a couple times.

Michael: 02:02

But Yeah. A million times.

Dave: 02:03

So tell us, like, how did you get to the point where like, tell us what you're doing there. And then what made you make the decision to not use that anymore?

Michael: 02:14

Well, so off the bat so for I could just go all the way back. So if people don't know, it's it was Maltbot, and it was Openclaw, and there's a bunch of names. But think Openclaw is what he what he settled on. And it's essentially an agentic AI that has you can give permissions and controls to use command line interfaces or whatever, like, terminal on on the Mac side and and Linux and all that and web browser access. And the thing that got me right away was there's a couple tasks that I do for my business over at SATSED Trading for clients that are annoying enough that it means, you know, at like 07:00 at night after I put the kids to bed, I gotta come in, sit down, do this very repetitive, boring task that takes like ten minutes.

Michael: 03:00

And it's just it's it's it's it was enough of a pain point. And I was looking at this, was saying, I think I can have this do that for me. And one of the things it would have to do is go in under the command line, import the new data from RealTest, and run orders for the next day. That was kind of the main thing. And then manipulating trading view and substack to, you know, input the symbols into watch lists and stuff like that.

Michael: 03:23

Because that's how, you know, I I I trade them out and and give them out to clients. It was that was the kind of genesis of it is I was saying, oh, well, you know, if if I could do this very simple task, it means that when I'm done work at you know, I chew try to pick up the kids around three or four eastern and then have a nice afternoon with them. That's I think one of the one of the things I get to do because I'm a systematic trader. And it just even though it was like ten minutes of work, it was just always hanging over my head of, oh, I have to go back and do this thing. So that was kind of the genesis of it.

Michael: 03:57

And I got it mostly working. And again, Claude is way better than at all this stuff so far. But just having that simple task, and that started to make me think is, okay, now that I've essentially taught this thing how to use and operate RealTest with the command line interface, that's where everything that we will talk about started to expand out from there. I'm like, oh, so this thing now can, without my input, you know, completely on its own, do use real tests and manipulate real tests. And that just had me think of I sleep for eight hours a day.

Michael: 04:34

I wonder what I could get it up to then.

Dave: 04:36

Yeah. Yeah. So I love this because, you know, just a lot of things have come together now that allow people to do this. So I've been, like, automating stuff like this for two decades more than two decades now. But, you know, I'm doing kinda the hard way, you know, coding stuff, writing the scheduled tasks, writing Perl scripts and Python scripts to do stuff.

Dave: 05:09

And a few months ago, my it was actually a guy at SMB that I'm coaching that I he really insisted that I use Clogcode. And I've been using ChatGPT up until then, and he's like, Dave, you gotta try Clogcode. So I finally did. And holy crap. It's like it has finally gotten to the point where it is literally a better programmer than I'll ever be and a 100 times faster.

Michael: 05:35

Mhmm. And you work while you sleep. Right? That's a

Dave: 05:39

I that's the step I haven't done yet. But that and that's what's so interesting about what you're describing and and how you're using it. But, you know, to the point where you're you know, this is just a very natural progression for what we've been talking about on this podcast for a long time. Like, because we've, you know, set up SOPs for everything we do, it's already kinda teed up for somebody else, like an intern or quad code to take over and look at exactly what you're doing. You don't really have to teach it a whole bunch because you've already kept up your SOPs.

Dave: 06:25

You you you and you've got that all dialed in. That makes this transition much easier and and possible to do where before, like, you know, when you use AIs like this, the the more focus you can give it, the less costly it's gonna be and the more the the quicker it's gonna be able to come to an answer. I mean, the big problem is most people ask too much of it, and it ends up churning and churning using up a lot of tokens.

Michael: 06:50

The

Dave: 06:50

more direct you can you can and the more context you can give it, the better it's gonna be.

Michael: 06:56

Well and that the SOP comment was, you know, really important because that's what I did is I had this kind of written which I actually used AI for. So how how I came up with my SOP because for those who don't know, I think I mentioned it a few times, I have like this rare form of dyslexia in which my brain can't sound out words and it makes it so writing It's why everything I've done has been kind of spoken word and I've and writing and everything is is a bit of a challenge for me. So how I did my SOP is I actually recorded my screen and I spoke through everything that I was doing as I was doing. I'm gonna click here. I'm gonna do this.

Michael: 07:32

I'm gonna do this. And then I I made that a YouTube video, a private one, and that generates a transcript. And I fed that into a different AI, and I said, I want you to watch this video, and I want you to read this transcript. I want you to build and I think I used the words like incredibly super mega ultra detailed SOP. And then I gave that to Claude and said, this is what I want you to do every night.

Michael: 07:55

And the the faster thing is crazy for me. So I said this this job would take, you know, somewhere in the neighborhood of I was just gonna say ten, fifteen minutes every night. Well, it's doing it on average in three. It's and it's constantly as part of the prompt finding faster ways to do it. So eventually, it's only gonna take a minute or so for it to do, and it just pings me.

Michael: 08:14

It says, okay. I'm done. You know, verify my work and and all of this. But having the SOP was everything because all I did is I created a Google document where I had the SOP, and then I put the my phone on speech to text mode, and I walked the doc. The whole time I was just blabbering on stream of thoughts exactly what I wanted to do, why I wanted to do it, all of this thing.

Michael: 08:37

And even though when I started it, Claude Co. Didn't have a scheduling function, but it does now, it because I said it's important that you I don't want to have to be here. Right? I want to like, if I die, I want my business to run for, like, a month without me kinda thing. Yeah.

Michael: 08:53

And the it came up with its own script and and and Windows scheduler and scheduled its own task and and built its own way around it. But you're right. It was having the incredibly detailed step by step instructions, and then my whole word salad of exactly, you know, the point of it and what to do. And I would say I've monitored it for three days to do this simple task just to make sure that it was doing it correctly. And now it's to the point where last night I forgot about it until it it sent me a message and and said it was done.

Michael: 09:24

Yeah.

Dave: 09:24

It's crazy. It's so it reminds me so much of automating a trading strategy.

Michael: 09:30

Right?

Dave: 09:31

Because you you know, a lot of traders are very uneasy with that. You can't just flip your you know, flip a switch and have it work. You kinda ease into it. So you get more confident. You get more confident.

Dave: 09:44

And all of a sudden, it's just it does it for you. Like, it's you know, you have full confidence in it. So yeah. That's really cool. So so you so so you first started with just automating some kind of annoying tasks or or stuff that you knew that you could that were automatable.

Dave: 10:05

So what are the what were the next steps? So you you got started doing that. You you have this process in place. You were like, wow. Okay.

Dave: 10:12

I a bunch more time now. Now tell us how you got to the next phase of what you what you're describing to me.

Michael: 10:21

So, yeah, the next phase was a list that I keep. And and this is it's funny because you're kinda right. This is right. If this was, like, the series finale of of Line Your Your Own Pockets, I think it would make sense because I think everything I'm gonna say is gonna reference probably a prior episode. But we had an episode of idea creation, and I talked about I read a lot of papers and watch videos and podcasts and whatever.

Michael: 10:43

And I have, like, a saved just document of I'll read a paper. Go, that's an interesting concept. Or what you know, and like with anything, most of the content's bullshit. And even the ones that kinda pass your smell test, I'd say most of those are still bullshit as well. But you just haven't tested to see if they are.

Michael: 11:00

So I have this kind of curated thing of every time I see a video that's interesting or listen to a podcast or read a study or something that's interesting, I put it in this document. So the next thing that I told it to do, I said, okay. Well, I'm going to bed. Now that I know you can do this with real tests, I gave it that document. And I said, one by one, I want you to grab the study, do your best to build the real test back test to back test it, and to to just keep going on this recursive process down this list, and run these back tests and make me what I just call a master spreadsheet and made this beautiful spreadsheet of every strategy and the potential results that it came from back testing it and the equity curve and and all of that.

Michael: 11:42

So then I woke up the next morning and it had done like 10 already because, you know, it it finds its own errors and fixes its own errors. So it's it's a bit of a process, but I'm asleep the whole time. What what do I care how long it takes? And, yeah, there was 10 strategies. They all sucked.

Michael: 11:56

But that that's okay. The point is that the the work was done. Again, I would never and I think this is we're gonna highlight this couple times this episode. I'm never gonna take a strategy that it gives me and just trade it. That's not the goal of all this.

Michael: 12:11

The goal of it is to try to get a lot of the crap out of the way that some of the strategies just didn't work. Right? There was one I remember wrote the comment and said, this worked great up until the moment the paper was published, and then it fell apart in the pay like, right after that thing notes like this. So it almost feels like what we talk about. I gave an assistant a work and said, do this thing over the eight hours while I sleep and generate this report.

Michael: 12:36

And now I think that's up to 250 different strategies that it has compiled, and it's given me this report. And I can then filter them and say this is the one I wanna explore today and sit down and really kinda dive deep on those things.

Dave: 12:51

Yeah. It really is almost like an assistant where you're or an intern just doing the the first pass, get to get in a format. It's kind of a grunt work doing doing what it's, you know, what it's doing. The real value is you know, you and I both know the real value is not really the strategy itself. It's like how you optimize the strategy and how you make it your own.

Dave: 13:16

So, yeah, I think that's really cool.

Michael: 13:18

Yeah. But what it's doing for me is it's it's showing me what what is worthwhile optimizing and what and what is not. So so far, the strategies that it's done of the 200, there's probably, like, maybe 10 or 15 that are even gonna be worth my time exploring. And there's one that I'm currently beta testing with a small account. I don't know if we talked about this on air or not, but it's a it was interesting because it has no stop.

Michael: 13:45

And it it's you're simply you're you're kind of selling the first day you're in profit, and that's it. And the act it looks super interesting. And it's existed. It was something that Larry Connors built in, like, the seventies, and it still works today. And I think it works because you don't have a stop.

Michael: 14:00

So it's like one of those you can call it like a penny in front of a steamroller strategy. Every now and then you get slapped, and you have to have the discipline to just continue to do the thing. But it's it's I've taken this from beta testing, and this was a strategy I was thinking of testing forever. It was just kinda low on the list. But now that that list gets cleared out so much faster, it showed me what to then take and then optimize.

Michael: 14:25

Like, I've got, like, 20, and I'm like, okay. These are ones that are gonna be interesting to test myself and go and explore and run through the cruncher and and do all the work for. And the rest of these 200, they can just go away, and I don't have to think about them anymore.

Dave: 14:39

Yeah. So so it's a good point to reference the episode we did on the ICE method for prioritizing strategies. I'm I'm curious now. Did you feed it to it in, like, an order that you'd already given a score to and prioritize based on that method? Or I mean, I could think of two different ways to do this.

Dave: 15:02

One would be it could help you with the scoring. Mhmm. And that's sort of what you're doing. You you could have maybe quickly do take a quick pass and do the scoring and then say, okay. Based on the scoring, I'm gonna I'm just working gonna work on the ones with the highest score first and go down.

Dave: 15:21

How did you do it?

Michael: 15:22

Well, so for that particular so I have two documents. And this so this first one is kind of just like a brain dump. It's just I read this. This is interesting. Let's throw it in there.

Michael: 15:31

Mhmm. I I'll I'll get to it later. And then the other one is the ICE method that we talked about. And these are ones that are are more fully fledged out. So the current project I'm doing with that is actually in that spreadsheet.

Michael: 15:43

I'm attaching a Google Doc in which I'm doing this walk the dog stream of consciousness speak for. So after we're done recording here, I'm gonna go walk the dog, and I've got one just queued up of a strategy because these are strategies that are much they're to make it to the ice list for me, it has to be something I've already formed in my mind. Whereas this other one, this is someone else's work entirely. I just want to see if their work is BS. Right?

Michael: 16:08

Yeah. Where so this is the other one where so for example, I've got a couple intraday strategies that I'm interested in. And what I'm gonna do is once I have this list filled out, that's the next place I'm gonna have it attack and say, so for each of these, I want you to go through in order, and then I want you to read my stream of consciousness, walk the dog, dump. And then from there, I want you to go through and back test. And my hope is is I start this on the weekend because the weekend, I don't really wanna look at my computer too much.

Michael: 16:39

But what I've noticed is that, like we talked about with assistant, it's kind of elevating me to manager. My work now

Dave: 16:45

Yeah.

Michael: 16:45

Is periodically checking on the bot and making sure it's doing okay and and looking at the spreadsheet it's preparing for me and all of this. And then the amount of times where I say, you know, my computer reset, get back to work, or or something like this is is a bit but I'm just way more high level. But, yeah, the ICE document's next. I just need to be able to give it more of a starting place for each one. Because the fear for this, and this was someone I was interested in talking to you about, is one of my fears is that it's missing something.

Michael: 17:15

So when it did those 200 backtests, I'm worried that it either, you know, you screwed up a backtest or, you know, did something wrong, and there might actually still be a gem in there that it kinda passed off as junk, but not because of anything wrong with the strategy, but just because the way it did the testing, I think, is is interesting.

Dave: 17:35

Like, it may have that was gonna be my next question is how do you know that it didn't screw something up or make some assumption that was really dumb or, you know, add some assumption that, like, got rid of, you know, 90% of the trays, and it was operating on the 10%, which, you know, wouldn't make any sense at all. So how how do you I mean, surely, there was a bit of a like you mentioned, the three days with your process for your business. Surely, there's a little bit where you're like, okay. Let me audit what it's doing here to make sure it's not just doing noisy stuff that's costing me money.

Michael: 18:12

Yeah. And I I'm still still doing the auditing, and and the the thing I'm I'm thinking about now is the the process of auditing what it's saying is a good strategy is easy. Right? Because then it's already been deemed worth my time, and I'm I'm reading the code. And this is where I still think even someone like me who just doesn't know how to program, understanding the scripting language of whatever you're using is important because I I'm, you know, auditing it and verifying it for myself and then checking it.

Michael: 18:41

And then I generally run my own back test to verify results, and then I start optimizing. So the ones that it says are good, the process there is pretty simple. Right? You just wanna make sure you just wanna pretend that, like, someone just emailed you the code and says, here, you're gonna risk your money on it. Right?

Michael: 18:58

So you wanna be able to do all these steps yourself. I'm more worried is on, like, the 200 ones that I'm like that are sitting in the trash. The only thing I'm thinking of is periodically throughout time as the AI gets smarter and smarter is make it routine where I say, I want you to go back to the ones that sucked and try again with because the AI is constantly learning as well. Because I'm I'm doing this all in one Claude code thread, which is, I guess, important technical. I guess memory, unlike something like ChatGPT, the memory isn't persistent across different threads.

Michael: 19:32

So you have to make sure when you're doing work like this, you're doing it all contained in one. And it's getting smarter. Like, we'll move over to Amnibroker and and what I'm doing there, which I I think you're gonna find cool in a minute. But Amnibroker, it probably took three, four days of constant nonstop twenty four hour work to even figure out how to do it in Amnibroker. So but with RealTest, what I'm hoping is that it's learning so much that just every couple months, I'm gonna send it back to the list.

Michael: 20:00

And I'm gonna say try again and just see if there's any of these value changes because there's just not physically enough time for me to go through the 200 strategies that it created and try to, like, really hunt for errors. I just can't think of a better way to do it.

Dave: 20:14

Okay. So do you have so back to these 200 ideas. Are these just, like, links to PDF documents and academia worlds? And how much context did you give it other than, hey. Here's the link.

Dave: 20:30

Go go read it. Did you say, okay. Here, I'm thinking you know, I have some thoughts here about this one. Do it in this way. Like or is it just like, hey.

Dave: 20:38

Here's the link. Go at it.

Michael: 20:39

For the so for the ICE method, I'm giving it a ton of detail. But for those, it was simply, like, watch the video, which generally means it goes and grabs the transcript and and reads about it or or the podcast. Like, I had it set the whole chat with traders, you know, every conversation that had there, that kind of stuff. That's just more of a raw dump. It's just see what this person has and then back test it.

Michael: 21:03

That's why I'm more worried. The ICE method ones, I think, will be way more accurate because of the amount of detail I'm giving it. Whereas this other one, it's it's looking at what the person said or wrote and then just trying to reproduce it the the best it can.

Dave: 21:17

So which so which set do you think has provided more value to you right now? The when it's done the first pass on the very sort of generic ideas or your ICE method list where you've given it more details and and of a more fully formed idea?

Michael: 21:36

The way more accurate with the ICE method one, but the issue is I'm the limiting factor there. Yeah. Where with the with the other ones, with with the ones that it's just finding, it's basically like a fishing expedition. It's kinda like throwing a rod in, and most of the time, it's pulling up a boot or an old can or something like that. But every every now and then, it finds something.

Michael: 21:59

And and the way I look at it so I'm able to do this, and I was just looking at my before we got here, my usage for the week, which resets Friday. And I'm on the it's like a $150 a month plan, the of the max plan. I'm only 42% through the usage. I think that's because I use Amnibroker this week, which which again we'll get to. The yeah.

Michael: 22:23

So I want it to work more. It's like because the way I look at it is on Friday, in twenty three hours and thirty one minutes, these tokens reset. I've only used 30%.

Dave: 22:31

So I wanna Yeah.

Michael: 22:32

You know, cram in more. So that's why I have these other fishing expeditions. I think there's just I still have to do a lot of thought about the way to do them correctly because I do think, especially I'm not as worried about this myself, but I'm more thinking about the listeners. If someone has never traded before, I feel like what I'm doing could be a trap for those people. I think whatever you're doing, you have to have a very discerning eye on Yeah.

Michael: 23:01

When it comes back out. But I do think there's value of just saying a lot of people have wrote a lot of stuff out there. Go get it and test it and see if somebody's sitting on something that has, again, uncorrelated edge to what you're what you're currently doing.

Dave: 23:15

Yeah. Think that's a great point that a lot of traders will like, they just ask too much of this. Like, the and the more you understand about it, the better it's gonna be because you can give it more context. You know what it's you know, you're not you're not using it as as like this magic eight ball. You're saying, okay.

Dave: 23:36

Here's what I want you to do. Go do it in a very specific way. So what what when the first efforts you gave it for this, did it churn through a bunch of tokens and not give out much output? And how did you handle that? Or did you did you that that's that's the biggest thing I could I could see happening here.

Dave: 24:00

Like, it just turning through a whole bunch of tokens, like, going on all these tangents, all these wild goose chases. It'll it'll come to the answer eventually, but it'll go to a lot of tokens to do it where you over time, I think you could get a lot better and say, okay. Do it in exactly this way, and it would be way faster. Use a lot less tokens.

Michael: 24:25

Yeah. And I I think it did itself. So way in the beginning, I was using not the the max plan. So for those who don't know, there's like Claude. It's like there's a basic plan, and that's for people who just wanna use it, you know, periodically to, like, help things.

Michael: 24:38

Then there's, I don't know, pro plan or something, and there's the max. And the max has two levels. The first one gives you five times what what what the pro did. And there's also a way that you can say, I'm gonna allow you x amount of dollars more.

Dave: 24:53

So I

Michael: 24:53

gave you, like, an extra $100 a month. It it can it can dip dip into. And in the beginning, I found what happened was it it was very, let's say, Claude intensive because it felt like it was learning real tests, it was learning Amibroker, and it was learning how to do this. And and I was going back and forth with it a lot to try to to get it that way. But like you're mentioning, eventually, I taught it about, you know, base strategies.

Michael: 25:18

I'm like, I want you to allow in as many trades. It kept spitting me back, you know, two to three positions at a time. I'm like, no. I want you to allow thousands positions at a time. I even talked about it.

Michael: 25:27

I even sent it the documents of the cruncher. I said, this is what I'm gonna do with you after Yeah. You're you're done all of this. So just so you know that there's a process on top of this. I don't want so then it started because it was starting to look at things like how much money did the strategy make.

Michael: 25:43

It's the first thing it looked at. Like, I don't care about that. Right? And I actually have it now looking at linear regression of equity curves. So, essentially, there was a whole lot of back and forth with me for probably two, three weeks now, and that was really token intensive because it was doing a lot of work.

Michael: 25:58

But now, especially now that it's on Amnibroker, the reason I think it's done no real no tokens is because it sets a back test and then has to wait for twenty minutes. And then it gets it gets pinged when the results are done and then does another and it it it's constantly working now. So it's more about Amni Broker and my computer than it is about Claude because it wrote one Python script to run 64 back tests for this experiment, which maybe we'll talk about this one, but we're already getting to half hours. So maybe we'll talk about it in the next one. And but that's not on it because it's just waiting for the signal that things are done.

Michael: 26:33

Yeah. So very, very front loaded. So maybe you gotta spend a bunch of money upfront. And then once you have it kinda done the way you want, go from there. And then one thing I just wanted to say while I'm thinking about it so I don't forget, had Claude code, you know, the we had time change.

Michael: 26:48

I I hate time changes. I make you whole, like, rant. But it was working during the time change, and it not only broke Claude code, but it bricked it. So every time I opened it up, it was just a blank screen with nothing in it. So a, be careful of that.

Michael: 27:04

B, how I fixed it was Claude code still existed in the command prompt. So I I talked it there, and I had it fix itself, which is a wild thing of itself. But just be careful with that. And so going forward, now what it does is it it every time I tell it something, it creates a memory document at the same time, and that gets backed up offshore. So if I ever have to restart Claude, I can just paste this document in and say this is where we were.

Michael: 27:32

Now get to work. But just I wanted to say that while I thought about that because that that was a bit of a panic attack. I'm like, this thing's been working for three weeks, and now it's forgotten everything it's done. That Yeah. Terrified me.

Dave: 27:42

So alright. So so what are the actual outputs? Does it if you've got a back test, the CSV file with a bunch of trades in it, you've got probably some column in the spreadsheet that says, okay. This is the average profit per trade or some sort of metrics about, hey, this is what the best I could do or something like that.

Michael: 28:03

Yep. So there's I'm I'm looking at the spreadsheet now for this this thing I've done with I've done with Amibroker, and there there's from column a all the way over to AF for, you know, amount of trades, whether or not it passed or failed, which, again, was something to talk about. Sharp ratio profit factor, average win, average loser, linear regression of the trend, max drawdown, like every metric you could possibly think of. And then inside each of those, it actually links me to the AFL file so that if I ever wanna go in and then test it myself, because what I've been doing is to speed up time is these don't have which I'm gonna need to go back to your website to remind me how to do it. They don't have the Mabe kit column library in there yet.

Michael: 28:57

Because I said just run it raws to run it fast and because I don't need it for everything. The ones that fail, who cares? And then the ones that pass, then I I can go through and I can add those in later. But, yeah, there's a whole document. I'm actually gonna send this to you too so you can see these.

Michael: 29:15

And it created maybe I'll add some of these in the show notes so people could see. It created all these really cool files where it's done, like, these bubble charts and line charts and all of these different things showing the different strategies and and profit, you know, profit rates and and when they make money and then spaghetti charts of equity curves. And, yeah, it's given me a ton of data output so much that I don't really know what I will do with all the data output. That's why and again, maybe in the next episode, we'll talk about I tried to create, like, a pass fail metric Yeah. So that I don't a way to, like, combine all the different metrics in a way that this is one that I wanna look at, and this is one that I I don't.

Dave: 30:03

That's so cool. Yeah. I mean, I I think this is great. Ben, I can also tell like, you're right on the very first steps of this. Like, there's a there's a Mhmm.

Dave: 30:17

I and, really, it's it's there's so many loose ends to tie up to really make this a really tight process that could work really well. So yeah. I mean, this is really cool.

Michael: 30:29

Well, in the scheduled overnight, so as of recording this, it was only you sent me the message. It was only a couple days ago that they actually added scheduled tasks directly into Yeah. Claude code itself. So I have in in these tasks, I'm just basically telling it every time I won't be at the computer. So weekends, nights, like, go to work.

Michael: 30:51

And I create a Google document, and I'd say this is your just like you would with an employee. These are all the things I want you to accomplish this week. When I'm not using my computer is that these times, go to work. And it it wakes up, and it it does those things and and helps me with my business. And then every morning, I wake up and part of my process now is just reading this.

Michael: 31:16

I have it write me a Google Doc and with some charts and everything and so this is what I've accomplished for you. This and then it so I feel like it's almost a passing of the torch back and forth. And that really to me was a bit of an eye opener. It's not you go and make me a whole bunch of money, I'm gonna go to an island. It is there is half of the day when I want to be a good father and exercise and sleep and and all these things.

Michael: 31:40

You work then. And then there's the other half of the day where I wanna sit at my desk, and I like we talked about, you have to love doing this. I want to do some of this. And Claude is sleeping, and I'm looking through everything it's done and taking the torch and saying, okay. This is the strategy today I really wanna nail in.

Michael: 31:56

I'm gonna start working on the the cruncher with it. I'm gonna start trying to make this a decent strategy. And then it it's good to pass the torch back. And this is something that I think I don't know I don't know how, like, meta I wanna get with this, but there there's a problem when you're a, self employed and b, a trader for a living that it's very hard to shut off and and live a life. Because in a lot of respects, the amount of money that you make is tied to the amount of work that you do.

Michael: 32:28

And this has made it so much easier for me. Because now when I go pick up the kids at school, it's not a 100% idle. I can be idle, but I the the work is still happening in the background, and that that gives me a sense of satisfaction of, okay. I am still pushing forward in my trading, and I'm still pushing forward in my business even though I'm, you know, pushing a kid on a swing instead, and it's been it's been awesome. Well, I think we're at the utopia zone of AI and whether or not this is the top.

Michael: 32:58

Who knows? But it's feeling pretty good right now.

Dave: 33:01

Yeah. I think that's great. Alright. So why don't we push off? I've got several more questions.

Dave: 33:07

I wanna I wanna think more about this, and let's talk about it on the next episode. I think this a really good start. I just have lots of things, and I know it's gonna take too long. So let's, why don't we wrap up there? And I I I think the listeners will have a lot to think about for a week on this.

Dave: 33:26

But, yeah, this is really cool. And, yeah, I've I've got way more questions to to bring to you next week.

Michael: 33:31

Yeah. I knew this was gonna be a this might be a miniseries because this was a there's a few like, I feel the first, like, big light bulb moment for me in systematic trading was just and and this was with trade ideas and the thing that you would build with BrokeragePlus. It's like, man, this thing just this trades for me. I had an idea, and that that idea made me money. That was, like, the first holy shit moment.

Michael: 33:55

This feels like another moment where I'm like, wow. I can now do way more work than I ever thought I could before, and I always feel like I'm I'm pushing forward. So, yeah, we might we might rant about this for a while. I know there's some people who are not really keen on AI, but I I I very much encourage you to to pay attention because we're not we're not using it wrong. We're we're coming up with ideas to use it as seasoned systematic traders, so I think there'll be a a huge amount of value here.

Michael: 34:23

But, yeah, I wanna keep writing. But, yes, we'll we'll do that in the next one. And as always, I'm Michael Nauss.

Dave: 34:30

And I'm Dave Mabe. We'll talk to you next week on Line Your Own Pockets.