YT YouTube2Text

YouTube2Text → Transcripts → The Simple 4-Step Process To Build Your Own AI Trading Assistant With Claude (for Beginners)

The Simple 4-Step Process To Build Your Own AI Trading Assistant With Claude (for Beginners) — Transcript

by SMB Capital · 7,718 words · 1,229 segments · language en · Watch on YouTube

Open in transcriber Summarize with AI Download SRT/VTT

Full transcript

0:00Two months ago, I couldn't write a line
0:02of HTML. Today, I have a personal AI
0:05trading assistant that tracks every
0:07trade I've made this year. It flags the
0:09moment I start repeating mistakes I've
0:12already paid for, auto-populates my
0:14dashboard while I just talk to it about
0:16my day, and runs a full ritual at 4:15
0:20every afternoon without me lifting a
0:22finger.
0:23I built the entire thing in a single
0:25afternoon. And by the end of this video,
0:27so will you.
0:29Claude Code unlocked this for anyone,
0:31and the traders already using it are
0:33pulling ahead fast. I'm not going to
0:35talk in theory.
0:36I'm not going to tell you AI is
0:38powerful. You know that.
0:40I'm going to show you the exact
0:42four-step workflow I use, and take you
0:46from a blank page to a working trading
0:48assistant today. Every prompt I typed,
0:51the exact order, the shortcuts, the
0:53guardrails that kept me out of trouble.
0:56You'll have them all.
0:57And you can pause this video, copy them,
1:00and paste them into your own session as
1:02I go along. That's the point. You're not
1:04copying my dashboard. You're learning
1:07the workflow that lets you build yours.
1:10You have everything you need to close it
1:12right now, today. All right, let's get
1:15into it. Garrett, how's it going? It's
1:16good, man. So, I'm excited about this
1:19one because as we talked about it for
1:22a few minutes, briefly, and you kind of
1:24filled me in on what you wanted to talk
1:25about today,
1:27I love it because this is like this is
1:29just let's lay out
1:31the actionable stuff, right? And I I I
1:34thought about myself in the situation,
1:36but I'm somebody who knows how to code,
1:38right? I come like we've we figured out
1:40how to code and and do a lot of this
1:41stuff on our own before AI came in.
1:45But even then, like once
1:47Claude Code came about and co-work and
1:49all you know the agentic AI stuff, like
1:52like I was even even kind of coming from
1:54a coding background, I was kind of like,
1:55where do I start, right? Like I was
1:57asking you all kinds of questions like,
1:58what are you even doing? How do you how
2:00do you kick off this project? So, I'm
2:02thinking about you know, people who
2:04people I know, traders I know, who have
2:07zero coding background,
2:09not much AI experience, right? And I've
2:11mentioned my dad before cuz he's like he
2:14he can trade. He knows he's doing. Um
2:16but he doesn't necessarily
2:19know what he's doing with the AI stuff,
2:20and it's like for him, it's like, okay,
2:23can I just give you like a recipe for
2:26how to build your own AI assistant that
2:31you could follow and do the same kind of
2:33stuff that Tim is doing? And that's what
2:35we're going to do today, so that's why
2:37I'm excited. So, why don't we just
2:38start? What's Where do Where do we begin
2:41if you're just getting into this and you
2:43want to create your own AI trading
2:45assistant?
2:47And the big like takeaway I want to have
2:51today is that it's not about the
2:53dashboard that we're building. The idea
2:55is you're going to have the steps, four
2:58steps, that you can build anything. And
3:00what I've found, like what makes Claude
3:03awesome,
3:04is it's not like you're doing crazy
3:06complicated prompts. It's not like you
3:09need to really know specifically what to
3:11ask. You have to just be curious, and
3:13like the possibilities are endless. And
3:15if you follow like four steps, which we
3:18will go over, like you can build
3:21anything.
3:22So, let's just jump into it, right?
3:24>> what's the take If the takeaway is not
3:27to not the cuz you're going to show the
3:29tool that you built, right? The the
3:31dashboard.
3:33But you're saying the takeaway is not
3:34the dashboard. That's going to be like a
3:36a fun accident that people can people
3:39can certainly copy that and take away
3:41that from the video, but what is the
3:42takeaway to you?
3:44The takeaway is going to just be the
3:46workflow. Yeah. Because with this
3:48workflow, you're going to be able to
3:49build any tool you possibly want. Like
3:52really any project within your trading
3:55process
3:56that would help you
3:58make you you make decisions,
4:00discretionary trade, quantitative trade.
4:02You can really build anything, and these
4:05four steps, all you would replace is
4:08pretty much keywords in that first
4:10prompt, which we will show you uh today.
4:12All right, so you're going to give us
4:13the blueprint, and then we can cater it
4:15to anything we can dream up that is
4:18specific to our own trading, which is
4:20great cuz we're all different. We all
4:21have different processes and strategies
4:23and stuff like that, too. So, what works
4:25for you might be different for somebody
4:27else. All right, let's go. And I'll uh
4:29at some point take over the screen and
4:31just use my own Claude.
4:33So, to start, Garrett, do you want to
4:35pull up image number one? Would you like
4:37to gain the biggest edge a retail trader
4:40can get? All of our daily and weekly
4:43in-house trader meetings are now
4:45available to you. Just head over to
4:48smbtradingfloor.com
4:50to find out more. I just kind of want to
4:52flash this to you cuz this is the
4:54takeaway. This is the main point. So,
4:56these are my four steps. You could have
4:58more steps. All you need is these four.
5:01And this is exactly how I built the
5:04trading dashboard
5:06on just a Sunday afternoon. And it
5:08starts with plan mode. So, I go into
5:11Claude Code, and I'll show on my screen,
5:15but you go into plan mode, where Claude
5:17doesn't accept any edits. It's not
5:19making any changes to scripts or files.
5:21You're just brainstorming with the goal
5:24of coming up with a concrete implement
5:26implementation plan.
5:29Step two is you have that plan.
5:32It's going to be just an
5:33.md file, and then you're going to go
5:36into build mode, where you're going to
5:38give Claude Code that plan, accept
5:41edits, and it's going to build the
5:42skeleton of the dashboard.
5:45Then the third one, the third step, is
5:48where it gets fun. This is where you
5:50personalize it. You can say, all right,
5:52I actually want this stat in this tab. I
5:55actually missed a tab, so can you add a
5:57tendencies tab?
5:59And you can really add any element that
6:02you want. And one cool thing um that I
6:05do want to show you is you can add this
6:07coach layer, a performance coach, where
6:10it looks at For me, it looks at
6:12different trade write-ups, and looks for
6:14tendencies and mistakes, and is able to
6:17do analysis that you might not be able
6:19to see yourself.
6:21And then fourth, you have this cool
6:22thing, how do you implement it in your
6:24daily workflow, into your routine? And
6:27Claude Code got this cool thing called
6:29routines, where you just schedule it,
6:32and you can have it ask you certain
6:33questions to help fill out your
6:35dashboard every day. So, it asked me
6:37like, what trade did you take today?
6:39Do you want to add anything to the
6:41journal, best ops, etc.?
6:43And those four steps,
6:45emphasis on
6:47plan first before you build, then you
6:50build, then you personalize, then you
6:52make it repeatable.
6:54All right, and of course we're going to
6:55go into each of these and how exactly
6:57you do each of these steps. Um but when
6:59just to clear this up, so when you say
7:02dashboard, you're like, okay, we're
7:04going to build, you know, in step two,
7:06build mode, and that's where Claude's
7:08going to go in and actually build the
7:09dashboard. Like, what are you talking
7:10about? Like, pretend that I know nothing
7:13about what Claude Code even is. Like,
7:16what are you doing? Dashboard where?
7:18Where is this dashboard?
7:20100%. Yeah, so
7:23we've used TraderView in the past. Um I
7:26know that is common just in general.
7:28There's a bunch of other ones.
7:29The idea was I'm going to build my own
7:31TraderView
7:33that's specific for what matters in my
7:35trading,
7:36and include stats that might not be
7:39included in TraderView. So, TraderView
7:41being like a performance tracker that
7:43gives you statistics on your different
7:45setups, and you lets you tag trades, and
7:48gives you your P&L breakdowns, and
7:51win-loss ratio, and all that stuff,
7:53right? Exactly. And when you say
7:56a dashboard, like when we go to
7:58TraderView, it's like a website, and we
8:00like log our trades, and we can look at
8:02it. So, when you say dashboard, are you
8:04building
8:05like a Tim website?
8:07Yeah, it's an HTML. Okay, so it's So,
8:10it's an HTML that's now going to show
8:12you that's personal to you, no one else
8:15can get in there, right? Yep. And and
8:18it's personal to you,
8:20and that it's going to display
8:23everything that you're creating here
8:25that is like going to be constantly
8:26updated. So, when you say dashboard,
8:28that's where we go. It's a it's an
8:29internet webpage, okay.
8:31And basically then you will input
8:34trades, input any information you want
8:37to update the dashboard just in Claude
8:41Code, and then it will update it for me,
8:43and then resave that HTML website. Yeah.
8:47Okay.
8:49Great.
8:50All right. I think, Garrett, let's just
8:52let's just go into the first prompt.
8:55So, So, we're go we're going to go in
8:57into plan mode, step one, right?
8:59>> Yep. We're in Claude Code. Um just to go
9:03over it like some basics, this is the
9:05desktop app of Claude.
9:07Um you have chat here. You can then go
9:10to Claude Code. I'm starting a new
9:12session.
9:13And this is where you can go down to
9:16accept edits. So, that's build mode
9:19right now. You're allowing Claude to
9:21edit anything, change anything.
9:23And if you click this plan mode, now
9:25we're in plan mode.
9:26Okay, and let's back up a second, and I
9:28don't want to belabor this stuff, but I
9:30know that there like I mean,
9:32there are a lot of people that don't get
9:34this because I didn't at one point. Chat
9:37mode is one thing. Claude Code is a
9:39different thing. So, chat mode is like
9:41what we all think of like chat GPT,
9:43where we can just
9:44talk to it, and it can give us answers.
9:47Claude Code, the difference, the agentic
9:49part, is that it is an agent for you, so
9:51it's building things. It's coding for
9:53you, and building these
9:55these websites and consoles, and it can
9:58even manipulate your computer through
10:00like code work and stuff like that. So
10:01you're
10:03There is a Claude chat, but you're going
10:05into Claude code. So now you're in the
10:07agentic part, but you're choosing a
10:09toggle
10:11that allows you to plan with Claude
10:14code. So it's not actually going to do
10:16anything yet.
10:18Correct?
10:18>> Exactly. Exactly.
10:21And this is
10:22the reason this is the first step is
10:25before you jump into any project, you
10:29really want to come up with a plan. And
10:31you'll have Claude write the plan, and
10:33you can revise the plan.
10:35And
10:37it will just save so much time down the
10:39road of debugging, trying to fix any
10:41problems, anything you don't like with
10:43the plan before it starts changing
10:46things.
10:47This is the first prompt
10:50that we want to use. And I'll put it in
10:53there and I'll also read it.
10:55The first prompt I wrote, and this is
10:58the exact prompt that I used that
11:00Sunday, is I want to create a trading
11:02dashboard that I update every day with
11:05the trades I took, the best ops of the
11:07day, and I want to model it slightly off
11:11of a template dashboard doc with three
11:13images of a trading dashboard that I
11:15attached. But I want to incorporate a
11:17lot more into it. So for right now, I
11:20wanted use the ask me questions mode for
11:23every tab that I want on the dashboard.
11:26And since we're in plan mode, let's
11:27really take our time. The idea is I want
11:30to create my own trader view super
11:31unique to the stuff that matters to me,
11:34as well as a bunch of potential problem
11:37patterns that I want to include and have
11:39you aware of this is going to take a
11:41good amount of time.
11:44I don't want you to decide on anything
11:46yet. So let's just take it slow. And
11:48then I reiterate use ask question mode,
11:50ask user question mode for every major
11:52decision. Batch three to four questions
11:55per round. After each batch, update the
11:58plan document.
12:00It's great. Okay, so that that first
12:02part was like one sentence. So bravo for
12:04that. I would have been
12:06kicked out of
12:08my English classes in college for that
12:10one. But that's great. Um
12:13So you're you're basically just giving
12:15it everything up front so that you're
12:16kind of preparing it to to get on the
12:19same page as you.
12:21Do you want to sit in on all of our
12:23daily and weekly in-house trader
12:25meetings? There's no bigger edge retail
12:27traders can get. Visit
12:29smbtradingfloor.com
12:31to learn more. And anyone in the comment
12:33section, like the next area of Claude
12:36that I really am trying to focus on is
12:39how to use less tokens, how to be more
12:41efficient. And even rereading that out
12:43loud, you know, you can prompt more
12:46efficiently than that, more concise.
12:49What I do What I really wanted to give
12:51was the exact thing I did to build this
12:54because it's that simple and everyone
12:56could do it. But if anyone has any tips
12:59on
12:59token usage, efficiency, It's It's as
13:02cluttered as my apartment is right now,
13:04as you can see. Maybe you can see it.
13:06It's kind of blurred, but we're
13:07renovating right now. So there's stuff
13:09There's stuff everywhere. And I
13:12I can't think straight when when stuff
13:13is all over the place, and that's what
13:15that sentence reminds me of.
13:17Um All right. So I'm going to run it,
13:19and this might take a sec, and we will
13:22see.
13:26All right, here we go.
13:28So
13:29first a couple quick points.
13:33Once it ran, it saw the prompt, and what
13:37ask user question is
13:39is one way of trying to be more
13:41efficient with Claude code, where when
13:43you're in plan mode, rather than using
13:46it like a LLM chat GPT interface where
13:49you're asking a question, it responds,
13:51you type another response,
13:53if you say use ask user questions, it's
13:56going to be building the specs
14:00of our skeleton dashboard, and it's
14:02going to do it in a way of just
14:03drop-down questions, where you have a
14:06question and then you choose
14:08your response. And basically clicking
14:12>> choice, basically.
14:13>> It's Exactly. It's like a multiple
14:15choice, and just clicking one of those
14:16responses takes up way less tokens. It
14:19might even be zero tokens, but way less
14:22than typing responses out. Okay.
14:28And the other one thing to mention is
14:32in the prompt, you probably heard that I
14:34said I had dashboard examples,
14:37templates.
14:38And I basically just went into my own
14:41trader view, took a screenshot of a
14:44couple of the pages,
14:46and I'm giving it to Claude. Basically
14:48trying to get an idea of like this is
14:50the end goal.
14:52Similar to trader view. And it just for
14:55me, I like using templates. I do it when
14:57we do our scripting. I'll give it a
14:59template script. So it doesn't just
15:01start from a blank slate.
15:03That's just a personal preference.
15:05Um
15:06>> giving it three examples of what you
15:09might want it to look like using
15:12trading view
15:13or trader view.
15:15Um service or
15:16>> one of them One of them I just found on
15:19Twitter. And I was like, "Ooh, that
15:20looks pretty good."
15:22Okay.
15:24Um And this was I forgot what I was
15:25going to say.
15:26>> even you. You're You're not Nick.
15:27>> No, I'm not Nick.
15:29>> [laughter]
15:29>> Okay.
15:31But I was like, "Oh, I like how that's
15:32set up. I like the tabs on the left."
15:35Like it has a lot of good information,
15:37good stats.
15:38Okay. And it just has like a better idea
15:42of, you know, what we're trying to
15:44accomplish. Yeah.
15:46And so this is where it gets fun. So
15:48confirming the platform, local HTML
15:51single page app that you open in
15:53browser.
15:54So it's asking how I want this to exist.
15:57So I want just an HTML file that I'm
16:00going to get in my file explorer, and
16:02I'll be able to
16:05easily get it. So I just click that, and
16:07now we move on to the next question.
16:09Where should the dashboard data live?
16:11Daily entries, notes, tags.
16:17And usually, yeah, human readable, get
16:19friendly, I'll go with the first one.
16:21And they tell you sometimes like I
16:22recommend this.
16:24Okay.
16:26How does data get into the dashboard
16:28each day?
16:31So you can run a script that
16:33automatically pulls
16:35your trades. You can do a manual entry
16:38where you just type in every trade.
16:40You can even say not sure yet.
16:42And we'll defer until the end.
16:45All right.
16:46So what are you going to What did you
16:47choose?
16:48So on this one, I ended up doing my own
16:52where I was
16:53saying, "I want
16:56to create a routine
16:58on the last step
17:01that asks me certain questions."
17:15So I don't want it auto pulling
17:17anything.
17:18What I found for my own trading is I
17:19like I'll give it the executions. I will
17:24then also tell the P&L what pattern it
17:27is for my playbook, and then I'll have a
17:28short write-up.
17:30So you can then just type in your own
17:32and other, hit next.
17:35What is this dashboard for? Pick
17:37everything that applies.
17:40And just to keep it pretty simple,
17:43you can do as many as you want.
17:47I'll do
17:50review what I did, so the historical
17:53review, and the coach accountability
17:56view. So you have that coach layer.
18:00When I did it, I did all of them, but
18:02just for time.
18:03Yeah, great.
18:06And you ask those four questions, and
18:09then it will update the document, our
18:11plan.
18:19All right, now we got our second batch
18:21of questions. And what's fun is it's
18:24feels like brainstorming, you know?
18:27So what does best ops mean?
18:34I want my best ops to be setups that I
18:36saw and didn't take, setups that I took,
18:39market wide.
18:45I want my patterns to be trade tags,
18:48very similar to trader view,
18:50plus
18:51detected patterns so I could group all
18:53my trades together.
18:55The end-of-day routine, what should it
18:57ask?
18:58Trade write-ups, best ops.
19:01We'll do
19:04We'll do all those.
19:13Again, I'll do all of them. Just getting
19:15all the stats.
19:20So this is still just step one, and
19:23you're going to keep getting asked
19:24questions in these batches.
19:26You'll go through all the questions, and
19:29then it'll say, "All right, I have all
19:31the information we need, and your .md
19:34implementation plan
19:36document is finished.
19:38And all you'll do is save that .md .md
19:42file,
19:44then go to new session,
19:47make sure you're out of plan mode, hit
19:49accept edits,
19:50and just paste in
19:52your plan and say, "All right, time to
19:55build." And it'll build the skeleton.
19:57Okay, so how does So, it decides when it
20:00is ready
20:01when it's complete, when the plan is
20:03like if the if it if the plan's
20:05complete, it tells you or do you have to
20:07tell it
20:08that we're we're good here?
20:10It will ask a question basically saying
20:13is there anything else
20:15you want to go over before, but it'll
20:18say, "I have all the information I think
20:20I need to build the skeleton." Okay. So,
20:22at that point I say sweet because you
20:24can always personalize it after.
20:26Right, got you. So, that gives you a
20:28great jumping off point because now it
20:30it kind of understands this architecture
20:32ahead of time before it starts
20:34coding and and creating this thing. Um
20:38Okay.
20:40Exactly.
20:42And just lastly, before I stopped the
20:45screen share,
20:47what ends up happening is
20:50every day on the close, this is my daily
20:52close ritual,
20:54this was 34 minutes ago. I get a pop-up
20:56on my desktop.
20:58And
21:00I even moved it back a little, but then
21:02we had the podcast today. But, it'll
21:04say, "All right, I'm here for the daily
21:06close. What trades?" It'll start with
21:08the first question, "What trades did you
21:09take today?"
21:10And then I told it, you know, "I'll come
21:12back later just this one time."
21:15So, all right, so you're you're finished
21:17with the plan mode. You have it gives
21:19you a that file that the plan is on.
21:22And you're then going into the execution
21:26mode
21:27and giving it that plan.
21:29And what you're showing us here is the
21:30end result. Like, this is what This is
21:32what it what it asked you today.
21:35Do you have to do anything now? Like, in
21:37step two? Like, as we enter into step
21:40two, which is the What did you call it?
21:43Um
21:44execution mode?
21:46Build mode. Build mode. Right. So, what
21:49what are you doing in build mode
21:52other than giving it the plan that you
21:53just created in step one?
21:57That's all you're doing. That's
21:58Basically, it's literally you just you
22:01have this now, so you can
22:03start a new session
22:05and you'll then copy and paste the plan
22:08in. So,
22:10here's just a different plan, but you
22:12can copy and place one in
22:16where you just go
22:18there, add a file,
22:21and basically, what you saw on the
22:23routine was it looked for this daily
22:25ritual file.
22:28So,
22:36I I don't even actually want to do it,
22:38but you would post paste it in and say,
22:40"Here's my implementation plan for a
22:42trading dashboard that I'm trying to
22:44build. I want you to build the skeleton
22:47of this." And once that is built, the
22:50foundation's there, I'll most likely
22:53personalize and change up the layout,
22:55the visuals, and the stats included.
22:58Got you. So, that's step two. So, step
23:01two, build mode, it's going to see that
23:03plan, it's going to build the skeleton,
23:05right, of that architecture. Now, you're
23:06going to be able to see it, right? It'll
23:07give you like a link or something to go
23:10to that
23:11that website, right? And you'll be able
23:13to see it, and then then you can talk to
23:15it in build mode to tweak
23:18what it has built already. Correct?
23:21Exactly. Okay.
23:23And
23:24just to show, we don't need to go
23:26through everything.
23:28Here's when I did it myself.
23:31So,
23:34this step three now, we built it.
23:38Step three now looked like me
23:40personalizing it. So, it was very like
23:44one-off, one-by-one, I want to make this
23:46change. Mhm. For example,
23:49it had this chart where it was grade
23:52accuracy. So, it's each of my grades,
23:55A+, A, A-, B, and did I grade the trade
24:00accurately or not? Mhm. So, I'm like,
24:02"That's great.
24:04Can we actually have the exact accuracy
24:06percent as well and the cumulative P&L
24:08on the X axis?" So, it's small changes.
24:11Mhm.
24:13And then it's like me adding in
24:15our playbook.
24:17Okay, yeah, great.
24:21So, that is all of the personalization
24:26stuff.
24:27So, that's enough of probably showing my
24:29exact poorly worded prompts.
24:32>> [laughter]
24:34>> But, step three is honestly where it
24:36gets fun, where I'm adding adding
24:38journal entries, I'm adding everything.
24:40And then step four is you have this HTML
24:44built. It has all the tabs you want.
24:48You personalized it, you add your
24:49trades, you can backfill your old
24:51trades. So, now it has data, it could
24:54show your graphs.
24:55And then you're like, "All right, how
24:57could I make this so easy to input?"
24:59It's basically your DRC. So, how could I
25:01make this so repeatable that I can be
25:04consistent and I do my DRC every single
25:06day. And the easiest way is then telling
25:09it I want to set up a routine where this
25:12is scheduled at 4:15 or 4:30 every day.
25:16And then you say, "What do I want to
25:17input?" And it'll ask you again.
25:20And that's it.
25:21That's great. Four steps. So, let's
25:23review the steps.
25:26Number one is plan mode
25:29where you're coming up with the the
25:30thing and it doesn't do anything. Then
25:32it's build mode where you give it the
25:34plan.
25:35And then it actually just builds the
25:37skeleton right there. That's the easiest
25:39step.
25:41Yes, that is.
25:42>> Step three is personalization. So, now
25:44you're seeing what it built. You go to
25:45the It gives you a link. You go to that
25:47webpage. You're saying, "Oh, cool. All
25:48right, well, this is my
25:50this is my um
25:51What did you call it? A console or a
25:55dashboard? This is my This is my
25:57dashboard. Now, I want to edit and tweak
26:00things to my liking. I might want to add
26:02a few things. I might want to change a
26:03few things. So, now I'm going to talk to
26:06Cloud Code in build mode.
26:10Get it all tweaked the way I want. And
26:13of course, that could be a ongoing
26:14process, right? As you find new ideas
26:17and stuff as as time goes on.
26:19Step four
26:21is
26:23set up a routine. So, you're basically
26:25telling it, "Okay, like periodically,
26:27whether that's every hour, every day,
26:29maybe after the close, maybe before the
26:31open,
26:32maybe there's something that happens
26:34over the weekend, you're saying like I
26:36want you to either update at these
26:38points or I want you to hit me up at
26:39these points and I can I can give you
26:42inputs." Here's a mind-bending stat for
26:45you. SMB traders have 20x to 40x higher
26:49odds of success than independent
26:51traders.
26:52Find out why and how to greatly increase
26:55your odds by visiting
26:56smbtradingfloor.com.
26:58The reason like we skipped over some of
27:01the personalization and some of the
27:03questions is again like that's not the
27:07point. Like, if you have the first two
27:09steps down, like you can build anything.
27:12Yeah.
27:12>> Like, it's all in that first prompt.
27:14You're just saying in plain English what
27:16you want it to do. What do you want to
27:18build? What tool? And then you go
27:20through all of the questions and you
27:22answer what you want included, how you
27:24want it to look.
27:25And step three, build mode,
27:28it's not that fun to look at. So, if
27:30anyone's thinking, you know, we're
27:31hiding stuff, it's just it's building
27:33and you're just sitting back doing
27:35nothing, hands off the keys.
27:37Well, I think the order of operation
27:39helps um the learning curve a little bit
27:41and like demystify this whole thing.
27:44Realizing that all you're doing is
27:45talking to it, but if you can do it in
27:47an organized way and prompt it so that
27:50it doesn't get confused or overloaded
27:52and that you know,
27:53"Okay, A B C D. Like, here's my Here's
27:56my order of operations." I think that
27:57helps a lot. That definitely helps
27:59organize my mind versus just this
28:02free-for-all like, "Let me open up Cloud
28:03Code and just
28:05go nuts, right?" This is This is a very
28:08thought-out way to do it, which I
28:10enjoy. What Tim, what's the What have
28:13you been getting out of this particular
28:14tool that you built? Like, what is your
28:16favorite part about it? Like, something
28:18that you didn't have before that you
28:19couldn't do before that you can do now?
28:22Honestly, should I just show it?
28:24Yeah, let's Yeah, I'd love to. Yeah, let
28:26me let me pull it up.
28:28So, here it is where it starts with
28:31today's entry.
28:33And, you know, it has my trades.
28:37Exactly, you know, all the little stats
28:40I put. The fun stuff for me, it it has a
28:43ton of tabs, a ton of tabs. What I like,
28:46you know, your classic TraderView
28:47calendar.
28:49You have all your trades based on grade.
28:52Where this is where it's really a
28:54dashboard for me. Like, you know, you're
28:57the same way. Like, grading is our main
29:00focus. Mhm. So, having your trades split
29:02by grade helps a ton and you could go
29:05back historically.
29:07But, what I really like, and this is my
29:09probably favorite part, is I have this
29:12page called tendencies.
29:15And these are
29:17things These are different patterns in
29:21my trading that has shown up in multiple
29:25trade write-ups.
29:27Where a this dashboard and any dashboard
29:29is really only as good
29:32or as detailed or as personalized as you
29:35make it. So, the way I do it is I'm like
29:38giving it a lot of information in my
29:39trade write-ups about how I traded it,
29:42mistakes, what was really good about how
29:44I traded it, how was the allocation
29:47grade, accuracy, everything?
29:50And then it was like, all right, I have
29:52all of your trades for 2026.
29:55I've noticed you mentioned no man's land
29:57sizing five times.
30:00So that's a main trading tendency that
30:02you want to always be aware of.
30:05So did it recognize that as a pattern in
30:08your write-ups or did you tell it, "Hey,
30:12I I want you to give me a heads-up on no
30:14man's land trading if I start to mention
30:16it."
30:17Some of these
30:19it noticed itself. It picked up itself.
30:22Some of them I did tell it. This was one
30:24of them
30:25cuz it was a big focus.
30:28Um
30:29some of the ones that I didn't tell was
30:32this is end of month B cap where
30:34basically I try to limit the amount of B
30:37trades I take and in the past there has
30:40been this over-grading when we get
30:43closer to the end of the month and I
30:45don't want to take as many B trades.
30:48Okay.
30:50Wrong execution style, that's an example
30:52of one it found on its own.
30:54And what's cool is you have this page.
30:57So not only is this a cheat sheet where
30:59you're like,
31:00"These are the things that I want to
31:02focus on that I messed up in the past."
31:05On any day
31:07whenever I input my trades for the day,
31:10my entry with the trade write-ups, my
31:13journal entry,
31:14I have it do a coach's note, basically.
31:18Where it's a performance coach looking
31:20at all of the stats, looking at the
31:22write-ups,
31:24and it knows my mistakes, it knows my
31:26tendency, it knows my journal entries,
31:28so it knows where I'm at and it's going
31:30to tell me, "All right, those were two
31:33no man's land sizing today. Like that's
31:35something we really want to keep an eye
31:36on."
31:38Yeah.
31:38>> Which is kind of fun. That's great.
31:41That's a great way to use it. You know,
31:43you get to focus on your
31:45um your traps, right? You're
31:48tightening those things up that you need
31:50to to not forget about.
31:52Do you have a spot in here for strength?
31:55For like things that like like good
31:57tendencies that you that you want to
31:59leverage?
32:03Goals, but no, I should do more
32:06strengths. I haven't added that yet. I'm
32:08just thinking Dr. Ken is in my head
32:10right now. I'm just thinking like cuz
32:12I'm reading
32:12>> With his new book?
32:13>> I'm reading his new book and like that I
32:15feel like that's something that that he
32:16would say based on what I'm reading,
32:18which is great by the way, so far.
32:20Yeah, I'm enjoying it as well. No, I
32:22should add a strengths one. That's a
32:24that's pretty good.
32:25Here's another section that I really
32:27like where it is all of the playbook
32:29trades
32:30and it tallies them cuz we're marking
32:32them, gives you your stats, your win
32:34rate. But then here's my checks in
32:36favor. Here's my rules. Here's how I
32:38execute these trades and it's all within
32:41this. So it's a one place for everything
32:43for me in my trading. So how did it get
32:46those checks in favor? Did you give it
32:48your playbook? Exactly. Got it. Okay,
32:51great. So you're just
32:52creating a um
32:54you're just giving it a lot of the the
32:56information and and putting it funneling
32:58it all into here so that it sort of
33:00understands your business in a way that
33:03allows it to help you. Very catered to
33:06to the individual, which I think was the
33:08main point that you made
33:10right off the bat.
33:12Exactly. Where again, you want whatever
33:16you're focusing on to be the main point,
33:18TraderView didn't have a good way of me
33:21grading the accuracy of my trades.
33:23It didn't have a great way it has tags,
33:27but they could get a little convoluted
33:29as we've seen with TraderView and all
33:31your different accounts. But just simple
33:33win rate P&L by pattern. Mhm.
33:37I mean, you can throw in anything. I
33:38even have Claude tips in here.
33:43All my slash commands.
33:46This is great.
33:48This is great. Tim, thank you so much
33:50for opening up this for everybody
33:53because I it's been
33:55it's been helpful for me, somebody who I
33:57think
33:58you know,
34:00I think was sort of late diving
34:02headfirst into this stuff because I've
34:04built so many things from actual coding
34:08that I wasn't as like
34:10enthralled by all this, I think, but the
34:12way a lot of people were initially when
34:14I heard about it because I was like,
34:16"Yeah, I've like created algos and stuff
34:18and like I don't know what more I like
34:20can do." You know, like I didn't feel
34:22like I had a need for it. But then like
34:25as you've kind of played around with it
34:26and showed me some of the applications
34:28and how powerful this stuff is and the
34:30fact that like anybody could really do
34:32anything and
34:34um like we're not building web pages. Um
34:36this is
34:37this is great. It's a it's inspirational
34:40for me to get to kind of start to go go
34:42a little nuts here.
34:43Hey, I love it. And I it's fun. I like
34:46showing it. I got a question for you. So
34:48now seeing the four steps, seeing that
34:51you can build a dashboard, but you can
34:53build any tool, anything come to mind on
34:56something you might want to attack?
34:58Yeah, so I mean I think you've nailed
35:00the obvious one because like the first
35:02thing that came to my mind was some sort
35:04of
35:05trading assistant, I guess was what we
35:08what we've been calling it. Um but I
35:10think that my
35:12emphasis would be
35:16different. So like rather than
35:20performance-based,
35:22like yours is very geared on like DRC,
35:25like daily review performance-based,
35:26which I think is great and and I would
35:28certainly adopt that.
35:31But
35:32like you know all the stuff we're doing
35:33like in our notion and like I would try
35:36to
35:37like give it that, right? Where it's
35:39keeping track of
35:41catalysts, it's keeping track of
35:44um setups that are working, setups that
35:47are not working.
35:48Um I want it to understand how I grade
35:51setups so that I can
35:54give it information and maybe get some
35:57feedback on
35:59like what kind of grade the setup should
36:03be. I don't even know how realistic that
36:04is. Like I I kind of have my kind of
36:06have my doubts as to whether that would
36:08be
36:09better than than than what I could do
36:12alone, but it's interesting to me. So I
36:15think that like something some sort of
36:17like dynamic um resource page, I guess
36:21is the way that I put it. Like the like
36:23a resource page for
36:25the stuff that I care about. So there's
36:27all kinds of noise and information in
36:30the market that we we're all aware of
36:31that.
36:32Um but on the other side of that, like
36:35we all have our very clearly defined
36:37businesses and within those clearly
36:39defined businesses are very specific
36:41plays. And those plays have very
36:43specific
36:44checks in favor and variables and
36:47characteristics. And so
36:49my idea is like if
36:51the AI can understand what my businesses
36:54are and what those very specific checks
36:57in favor and characteristics are
36:59and then on the other side I could feed
37:01it information um somehow.
37:05Maybe uh maybe it would become a really
37:07good resource page for me for either
37:10finding setups or even just, you know,
37:12grading things that are there.
37:15That's a pretty cool idea. I like it.
37:18Especially on the grading setups front.
37:21Yeah, I mean catalysts like
37:22>> the I think the reason I thought about
37:24that is just because like that's been my
37:26focus. So it's not that I think it would
37:28be particularly good at it. Like I'm
37:29really I have my doubts.
37:32Um but I
37:34I go there because like that's what my
37:36focus is. So I'm like if I could have it
37:38help me do anything, it might as well be
37:40kind of like what I'm focusing on right
37:41now. So it would be like how can I
37:43figure out how to get it to help me in
37:44that area.
37:46So you know, I'm like and I'm kind of on
37:48the fence with this stuff. Like I
37:50acknowledge it's
37:52unbelievably powerful and it clearly
37:54based on your demonstration, you can do
37:57anything with this thing, which I think
37:58is
37:59awesome because for even stuff like
38:02TraderView, like you can do a better
38:04version just by yourself. I mean, this
38:06is this is why the software names are
38:08quick have been in the in the the
38:10doghouse, right?
38:12Um
38:12awesome. And I can see so many
38:14applications, but we've been doing this
38:16for so long that we're hooked up with so
38:18many things already that are sort of
38:19like integrated and helping us.
38:22Whereas like I think again, like someone
38:24like my dad who trades but doesn't have
38:25the infrastructure that we have. I'm
38:27like, "Oh my god, I have a million ideas
38:28for for you."
38:30Right? Someone who doesn't even have
38:31TraderView, right? Yeah.
38:34But
38:36I also feel like
38:38trading is simple and it and it should
38:41be simple. And of course like we want to
38:43do a really good job of being alerted to
38:46information quickly that's going to help
38:48us. We want to do a really good job of
38:50finding our best setups. But when it
38:53comes down to it, like at the end of the
38:55day,
38:56it's like whether you put the risk on or
38:57not on like a really good setup and it's
38:59like how you do it. And it's like, you
39:01know, it could be the simplest thing
39:04that you that you miss or that you don't
39:06do or that you do based on really really
39:09simple things like
39:12how you were feeling that day or whether
39:14like how clearly you were thinking,
39:16right? So
39:17given how simple trading is and how
39:20simple it should be, like that's my big
39:23sort of reservation or like question
39:25mark with this stuff because the last
39:27thing that I want to do is like get in
39:29the weeds, like deeper and deeper and
39:31deeper into this like never-ending sort
39:33of like complex
39:35um building out this like system that
39:37that seems really fancy, that makes me
39:39feel feel I'm doing like all kinds of
39:41work, but in the end, it's like what's
39:43the ROI? Like how's it actually going to
39:45help me
39:46like crush that really simple moment
39:49more
39:50than I am now. So, I don't want to lose
39:53sight of that. I What I want to do is
39:55like get into this. I want to experiment
39:57with it. I want to use it, but I want to
39:59do it kind of parallel to that thought
40:01without losing sight
40:03of that thought.
40:04Have you heard this study of tens of
40:06millions of traders which shows that
40:08only 1% of retail traders actually make
40:11it?
40:12Don't be a statistic. Visit
40:14smbtradingfloor.com
40:17to greatly increase your odds.
40:19And you can spin your wheels if you're
40:21not careful
40:22like so many directions you can take
40:25things and do way too much work.
40:27But I think like you just said, if you
40:29have
40:30the main goal of my entire trading
40:33dashboard is these are the things that I
40:35want to be
40:37aware of that I want to be able to
40:39access quickly
40:42in or in like to review or like
40:44pre-market.
40:46And this dashboard like if I focus on
40:50these things, that will get me to trade
40:52better in those simple moments
40:54discretionarily. So, never losing that
40:56connection where the whole idea is
40:58you just are doing stuff to make you the
41:01best trader you can be. You're not
41:03looking for some flashy dashboard that
41:06has these cool moving charts and like,
41:08you know, heat maps and crazy stuff.
41:11I I really like what you've done with it
41:13in terms of personalizing it to you and
41:15having it really focus on those points
41:17that can like move the needle. I mean, I
41:18think you've You've done a good job of
41:20that, which is great.
41:22>> people might have less of a psychology
41:24element as well.
41:25Yeah, sure. I mean, like everyone's
41:30everyone's going to be different. Um
41:33and I wonder I guess like my thing is
41:35like
41:36if I can just keep it super simple. Like
41:38again, like don't get carried away
41:39again. That's the last thing I want to
41:41do.
41:42But
41:43if I could just have something to check
41:45in with like during the day
41:47that knows my process and my tendencies
41:50just like you.
41:52Like I might use it more intraday than
41:55say after the close.
41:57Like I You know what I mean? Like I
41:59might be like here's the setup I'm
42:00looking at.
42:02Here are the checks in favor that I'm
42:04seeing. Here are the things kind of
42:05going against it. And
42:08um for some reason like I intuitively
42:12feel really cautious with this setup,
42:14but I
42:16I'm second-guessing myself. And maybe
42:18that's just a terrible path to take.
42:21Maybe talking to an AI [clears throat]
42:23about a goddamn trade it could be
42:26literally like How much risk can I take
42:28here?
42:28>> of the world. Like yeah, like exactly
42:30like like
42:31I'm just I'm just like we're
42:33experimenting here, right? We're just
42:34This is a thought experiment and I
42:36totally understand the potential
42:38pitfalls and sort of like how crazy this
42:41sounds. But I'm interested, right?
42:44Because if you can make it simple and
42:46mathematical
42:48um like maybe it just knows your system
42:50and you can use it as a computer, right?
42:51Like it's like, okay, well, if it has
42:53these checks in favor like so did this
42:55this and this example.
42:57And that these these great examples
42:59didn't have this one either and it was
43:01in the similar kind of market
43:02environment, right? It can go through
43:04your like
43:05backlog and kind of give you that like
43:07right away. Like that that that stuff
43:08might be helpful.
43:10My mind's going to you have a breakout
43:13day. You're getting involved in the
43:15morning. It doesn't close above the
43:17breakout level or close is right at the
43:19breakout level, which tends to be a
43:21tough decision on close.
43:23You have Claude has all your trades and
43:25it's like, all right, like how many
43:27times did a breakout trend trade that
43:29closed below the breakout level work for
43:31me?
43:32Right. Right.
43:34>> you maybe just get stats and it's like
43:36you've taken 20 breakout trend trades
43:38the past year. All of let's say 70%
43:43closed above the level that worked.
43:45Yeah, so
43:45>> And none of them closed below and
43:48worked. And you're like, oh, wait a
43:49second. All right, lower the grade,
43:51maybe get flat.
43:53That's that's a that's a very specific
43:55example of what I'm talking about. That
43:57that That's great. Um That's just where
43:59my mind went. No, that that's that is
44:01that's an actual example that that is
44:04that is right along the lines of what
44:05I'm talking about. And I think
44:07that that just speaks to how
44:10it's all about the inputs.
44:12So, like the better it knows
44:15whether all those breakouts you took
44:17closed above or below the breakout
44:18level. Like it has to know that or it
44:20has to be able to backtest. So, the data
44:23that you have that you're working with,
44:25if you're able to use an API and input
44:28data, if you're able to
44:30do really good reviews, really good
44:32playbooks that have all this information
44:34in it that you can give like it's just
44:36it's all about the inputs, the quality
44:38of the data
44:39in my opinion that you're giving it
44:41that's going to then give it the power,
44:43right? Because just cuz it can process
44:45doesn't mean it's going to be smart.
44:48Right? It has to
44:48>> Exactly. And you don't want to make it
44:50too many decisions on its own without
44:51your
44:52>> I don't I don't I don't want to
44:53>> I don't want to make I don't want it
44:54making any decisions. Like that's my
44:56That's my thing. But I sometimes might
44:59want to know um
45:02about past setups with like, you know,
45:04sometimes we'll run a backtest. It's not
45:06that different, right? But like to do
45:08that, you've got to go write a script
45:10and then do a backtest and even that has
45:13its own
45:14nuances and complexities that kind of
45:17are are you going to really get what you
45:18want to see? And we might come back and
45:20say, hey, you know, every time an IPO
45:22did this, like this was the next day,
45:24right?
45:25Um so, it's not that different than
45:27that. All right, just to summarize, we
45:29went over a four-step workflow that you
45:32can use on any Claude project. I happen
45:35to use it on a trading dashboard. And
45:38this is the Trading Floor podcast. Uh
45:40we're still on Spotify if you choose to
45:42listen there. And in the comments, let
45:45us know what you're building.

About this transcript

This page contains the full transcript of The Simple 4-Step Process To Build Your Own AI Trading Assistant With Claude (for Beginners) by SMB Capital, generated from the public captions YouTube serves with the video. The transcript has 7,718 words across 1,229 segments, with the original timestamps preserved so you can click any line to jump to that moment in the embedded player.

What you can do with it

Use the transcript to take notes, quote the speaker, build a study guide, generate a summary with ChatGPT or Claude via the YouTube Summary tool, or export it as a timed subtitle file with YouTube to SRT. You can also re-open it in the transcriber to translate the transcript into 100+ languages.

Free YouTube transcript tool

YouTube2Text is a free YouTube transcript generator — no signup, no daily limit. Paste any YouTube link and get the full transcript instantly, with timestamps, click-to-jump, translation to 100+ languages, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.