Transcription
I turned Claude into a fully-fledged financial advisor. It tracks my portfolio in a dashboard. It gives me suggestions every single day based on what happens in the market. It even messages me every morning to make sure I don't miss a single thing.
It's actually trained to behave like a McKinsey or a Goldman-level financial analyst. Because the real difference between this and just chatting to Claude or ChatGPT is that this is actually trained on real data. Today, I'm not just going to show you how to build a dashboard. Maybe you already know how to build a dashboard. I'm going to show you the full architecture so you can use this in your daily life. The full memory architecture so it actually has context about you because the thing with AI is it's basically useless unless it knows everything about you.
So to turn Claude or ChatGPT into a fully-fledged financial analyst, what you need is the correct data. Then you need to plug that data into the LLM, and that's where the magic happens. So without further ado, let me run you through the build. I'm honestly excited for this video because AI has made such a big difference with portfolio management. If you're like me, maybe in the past you struggled with complacency. You struggled with taking profits. You struggled to adapt quick enough to market changes.
This completely changes that because every single morning I get an update on Telegram. I'm going to show you how to connect it to your phone today so I don't miss a single thing. So when Trump comes out and there's a headline, or when there's a silent shift in the crypto market or a silent shift in the AI market, I'm getting alerts so I can make portfolio moves, or at least so I'm aware of what's happening. This is how you go from a passive investor to an active AI investor.
So to begin, everything starts with information. The reason you're probably not getting great financial advice from ChatGPT is because it doesn't have the context on you. So, you need to do an exercise first, just like you would when you sit down with a normal financial advisor. They would ask you everything about yourself. They'd ask for your tax situation. They'd ask for your income situation. They'd ask for your goals, how you think about finance, what you were trying to achieve in your portfolio.
Because the reality is every single person's portfolio is different. For example, I hold more cash because I have businesses which require more cash flow. If I ask ChatGPT how to allocate my portfolio, it'll tell me to do a 60/40 stock bond split. That makes no sense for me. I obviously don't need to hold as many equities because I'm taking on risk in other areas. So, your portfolio should be risk-adjusted to your situation. You might want to take on way more risk for a particular reason that is only relevant to you.
So, the first thing we need to do before getting any sort of financial advice from AI is run a system prompt which I've already created for you. And by the way, the full architecture, all the prompts, everything for this video will be available in the description down below for free. You can sign up for the newsletter where we're sharing AI finance workflows. It's an amazing newsletter to help you guys level up your finances using AI. And then once you've signed up, you'll get access to the Instagram community. In the pinned link, you'll find the corresponding files for this video.
So, if you actually look at the system prompt here, we have set the role for Claude. "You are a McKinsey caliber strategy advisor specializing in personal investment frameworks for high-performing operators, founders, executives, partners, and serious self-directed investors." It goes through operating principles, interview structure where it's basically going to ask you a bunch of questions. Phase one, it's going to ask you about your operating base, income, liquidity, family, health. Phase two, it's going to ask for a capital base, time horizon split, ability and willingness to invest and add capital. Phase three, philosophy and mindset, getting into your mindset about how you think about markets, helping you identify strengths and weaknesses, which the AI can help supplement. Phase four, behavioral nuance, known blind spots, routines, and rules. Phase five, preferences and anti-preferences. And phase six, goals. Phase seven, stress tests.
You might be thinking, "Oh, but this is going to take some time to answer these questions." Of course it will. But if you were sitting down with a McKinsey-level consultant, you'd probably have a two-hour meeting. And then you'd have regular catch-up meetings to outline your financial philosophy. So, you can probably do this in 20 minutes. And what I recommend is actually voice prompting. So, paste the prompt in, which once again is available below. Then, you want to put it into Claude or ChatGPT. By the way, you can use either model for this build. Then you want to go through and answer the questions. And I recommend voice prompting because it's going to be way quicker for you to go through these questions. And it basically breaks it up into phases for you. So when you put in the system prompt, it's not going to ask you 100 questions at once. It's going to go step by step through the prompt asking one question at a time.
Once you've gone through this process, and I've done an example process for today's video. What you'll then be able to do is download a one-pager. This will basically be a document. This is a markdown file, but you could also save it as a PDF. Markdown files are .md files. This is basically how AI reads your data. You will then get a one-page document about yourself at the end of this process. So, it'll have your Northstar, core philosophy, time horizon, mindset rules, personal finance details, anti-portfolio details, meaning if you have any hard rules of things that you don't want to do in markets.
Then what I want you to go ahead and do is create a folder on your desktop. This folder will contain all of the necessary memory for the AI to read. The truth about AI is we're no longer in a prompt engineering environment. Of course, you need good prompts, but what is really more important now in AI is having a good memory system. So, a system where all your data is logged, updates over time, that you can take with you and plug into any LLM. You want a structured investment folder which always has your data because then you own your own data.
The problem with sending messages into Claude is that it's basically a black box. You don't see what memory it's saving about you. So, you have no sovereignty over your data. That is very bad for portfolio management because in finance, things need to be precise. You need to know exactly what the AI knows about you so you can verify its output. That's why the entire build today hinges on a memory system. The simplest version is a folder on your computer. And you can get more advanced over time. For example, putting your memory system into an online application like Superbase, which will then allow you to have your data in the cloud and access it from anywhere. But I recommend just starting with the desktop and starting on Claude Code/Codex or GPT.
I'm showing you today's video on Claude. However, you can do this exact process on ChatGPT and Codex is actually very, very good, arguably better than Claude Code. And the process is the exact same. You still want to create the same folder. You still want to create the same files. You still want to do the same prompt. The only difference is instead of interacting on the GPT app or the Claude app, you'll be interacting on the specific Codex application. And by the way, in AI when we talk about code, sometimes people freak out. It's actually not about coding. The coding interfaces like Claude Code simply enable you to access local and edit local files on your computer. That's all it is. And you're going to need that if you want to build the dashboard that I'm about to show you and build the workflows that I'm about to show you to get live market updates.
And by the way, this tutorial works for your stock portfolio, your crypto portfolio. But how I personally use my AI custom-built memory system is to have an overarching view of my finances. So, it'll have my crypto portfolio, my stocks, my gold, my cash positions versus what's happening in the business, and my profit and loss for the month, which I'll also show you today. And that helps give the AI full context to give an output.
Now, let's talk about how I've actually structured my memory folder. You'll see there are four files. One of the files you just created by doing the prompt. This is going to house all of the data about you, so the AI knows about you. Then, you're going to create an `instructions.md` file. Of course, the architecture for this and the prompt for this will be available for free down below. But basically, what you want to create here is instructions to tell the AI how to behave. So, for example, "You are my personal investment strategist. Operate as a McKinsey caliber level advisor, probing, direct, no fluff, no flattery."
A very important part of the instructions are your rules. Number one specifically solves the entire memory problem with AI. The memory problem you typically have is if you give information to a chat like "I'm no longer interested in investing in equities because of XYZ," you might open a new chat the next day and then it forgets that. And you know, if you ask for financial advice, it'll just start suggesting, you know, buying equities for example, when you've already given it information that you're not interested in buying equities right now. And that's because ChatGPT and Claude, they don't update memory very well. So in order to take control of your own memory, this line in the instructions: "Read investor one-pager and memory before every response. Treat both as canonical. Whenever new information emerges in conversation, update memory.md immediately. This covers life changes, jurisdiction or tax shifts, income changes, new positions, new constraints, behavioral patterns, observed and evolving views. Append, never overwrite. Date every entry."
That line is so important. So you basically have an instructions file and then you have a `memory.md` file. The `memory.md` file is essentially your brain. This is the file that the AI is automatically updating when you're interfacing with it. And this builds over time. So the more you chat with it, the more you will add to the memory log. So the memory file is an ongoing log. The instructions.md file tells the AI how to behave. And then your investor one-pager sets the scene so it has all the information on yourself.
And then the last thing I went and did is make a financials folder. So I have all of my financials for March, April. So you can actually see, you know, your bank statement, your profit and loss. So the AI could actually see your operating income because, you know, investing is also based on cash flow. It's also based on income. And you can actually use this, and I'll probably do a separate video on this, to create your own graph for personal expense tracking, which is one of the really cool things about AI.
So I've got this folder system, but you might be wondering, Miles, how do I actually create this? Well, what you want to do is go into Claude Co-work or Claude Code. You can actually use either for this video. I'll be using both. Co-work is a little bit better for just interfacing with your folder day-to-day, reviewing finances, chatting with Claude. It's just smart. Claude Code is a little bit better at the dashboard build and the more advanced stuff which we're going to get into. But if you want to do this on Codex, once again, it's the exact same process. You simply want to connect your folders to Codex.
So once you've created the folder and you've got the basic information in there to begin with, all you really need is the one-pager, by the way, because you can actually just get Claude Co-work or Claude Code to create these other files based on the prompt that I'm going to give you. So all you need to do is start with one folder. You simply want to click on the folder, click open, you want to click allow, and now it can access all of the information in this folder. So if I say, "Create a new entry for May 2026 in the financials folder," remember Claude can actually edit files on your computer. It'll basically add a new folder here. And by the way, to access Co-work, you just go into the top left-hand corner and you just toggle between the chat and Co-work. The normal chat can't do this because the normal chat on Claude.org or GPT doesn't access local folders. You actually have to drag files in, and that defeats the whole purpose of a memory system that you own. So that's why you need to use the coding interface or the Co-work interface. You can see it's now creating the entry. So this is how you create all the files and manage the files on your computer.
For example, if you drag in your bank statements every single month, it'll be able to automatically consolidate those statements and then create a dashboard to visualize this stuff and calculate it for you. So this is one of the biggest game-changers and probably the highest leverage thing from going from, you know, a non-AI native investor and financial manager to someone that is AI-native. It's just using AI to actually read and consolidate your data. It saves so much time. Honestly, I think for many people it's actually going to save them needing accountants, not only financial advisors.
And I also want to say, I know, you know, the title of this video might be provocative, but I'm not saying don't have a financial advisor if you need one. Of course, specialization is important. Of course, these AI models, they still hallucinate and you need to double-check data, and this isn't actually meant to fully replace them, but I think it will have people reconsidering the need for them. Like for example, you know, I was meeting with a variety of different firms and my portfolio allocation is actually quite passive and basic. So I realized with a good AI system, I don't have, you know, a need for a super high-cost financial advisor. I might get advice here and there on a consulting basis, but I don't need full management because AI can take over a lot of that for me. So it'll really depend on your situation. But of course, this video is not financial advice and you should still take the AI prompts with caution.
So now you understand the memory system behind this all. Once again, we'll have the architecture in the description below. Let's get on to the fun part, which is actually building the dashboard and then building the workflows for market analysis and market updates because that is the real source. But you know, in order to create that, we needed a strong data system so the AI actually knows what to do. But before we could do that, of course, we needed to create the memory system, or else, you know, the rest of this wouldn't have been effective.
Now, what you want to do is switch over to Claude Code or Codex and actually create your dashboard. Of course, I'm going to leave a prompt for this, but if you wanted to do it yourself, you basically give it access to the folder and tell it you want to create a portfolio analysis dashboard. I'm going to save the time in this video of going through the whole process again because it was a, you know, very long process that I went through, but I finally created a dashboard that I'm happy with. Now, I actually have a more advanced one that I'm personally using now. I'm going to save that for a future video because that is a much more complex build. And in today's video, I just wanted to go through the basics of how to turn Claude into a financial advisor, an analyst. And then we can get into the more advanced stuff in a future video, like Claude memory systems, auto-generating code, and auto-generating updates.
And something that's really cool is Claude Design to actually make your dashboards look even better. Like I actually created a personal OS on Claude Design, which ties into this system. I obviously have fake data just for the purpose of this video so I don't leak my real stuff, but you can see I've actually created a dashboard for myself. By the way, this video is probably out by now, so I'll link it below. It's on AI Edge. Very, very cool video. And this finance system actually links to the other one. So, I have a dedicated finance tab here which opens up a dashboard. And then I have my, you know, personal tracking here. And this basically logs all of my tasks, my task management, my priorities, my weekly goals, my daily goals, etc. And I do this in tandem with the portfolio management tool.
So, in order to build this, you want to do it on Claude Code. It doesn't take long. Now, there are a few cool things about this dashboard I want to run you through. Firstly, it has live data connected to Yahoo Finance. What you can actually do on the back end is you can create a Google Sheet. Let's say like you had a lot of crypto. You could actually connect an API from your exchanges or from CoinStats into the Google Sheet. So, it populates your positions in real time and then you can plug that Google Sheet into the back end. You can do that pretty easily on Claude Code. If you literally ask Claude Code how to do it, it'll run you through how to do it. For many brokerages, you can also actually connect an API which can also help you.
The other thing you can do, you know, if you're not trading that frequently, is simply just enter the data manually on Google Sheets. If you're not dealing with a crazy portfolio, you'll probably be able to do that. Or to make it even easier, you can just screenshot your positions, send them to Claude, and get it to create a sheet for you. And then you can just update periodically. So the level of complexity you want to undergo in the beginning will determine how automated this is. For me, I connected APIs to a Google Sheet in the back end to make everything up to date. But if you want something simpler, you can simply, as I said, just screenshot and create it periodically. But of course, the numbers will still update on the dashboard because it's connected to Yahoo Finance. You can also add other crypto connectors or stocks for like lower cap stuff as well. So you capture everything. So you're still going to get live up-to-date data as long as you're maintaining your positions.
And you can actually set up a system so it can actually trade and execute for you. There are websites, and I've actually been experimenting with these even for equities, to automatically execute your trades. And you can also do the same with crypto like Bybit, Bybit, Binance. They actually all have APIs now. You can create an API for them where you can actually use Claude to trade just through Telegram or just through Claude Code. Like it's absolutely insane. That's I'm going to do a future video on that. I'm going to do a $20,000 challenge where I put $20,000 in and I get the AI just to trade on my behalf.
So this is the dashboard part. There are many ways to do it. The way that I've done it will be in the description below, but it's basically using a Google Sheets integration. If not, you can screenshot. That's not the real source. I mean, there are other ways to visualize your data. I mean, Plexity has it now. There are other apps like Sylvia that allow you to plug in brokerages. You know, why do I like doing it myself? Well, I have some more obscure positions. So, the benefit of doing it yourself versus using one of these pre-made platforms is, you know, I can enter properties that I might have that might not be on the books with live Yahoo Finance data. I can enter illiquid positions like my VC portfolio, which I could factor into this. Like anything illiquid, any, you know, super low cap crypto as well, I can add to this as well myself. That way, I'm not boxed in like I would be with some of these other applications.
But the real source is actually what's below. It's not even just the data. It's this. So, when you click, because it's plugged into your real data, so this will be much better if it's actually got your information. This won't even do it justice on screen, but when you click "Generate Morning Brief," it basically audits the portfolio. So, it automatically goes through all of your positions. It audits the portfolio and it gives you advice. Once again, remember the system that we built is based on McKinsey-level, Goldman-level analysis. And you can of course improve this over time by even feeding it the real way that financial analysts behave.
If you watch podcasts, videos, like let's say, this is a really cool thing. Let's say you watch a video from a creator that you like or a thought leader that you like and you're like, "Wow, that investment philosophy is amazing." What you can actually do is you can just take the transcript on Glasp or on NotebookLM, put it into your LLM chat that's connected to the folder and say, "Hey, review my portfolio and weigh it up against the philosophy of this person." And then it will actually tell you where your portfolio aligns with this person or misaligns with this person. And I found that really interesting. Like when I've had big guests on my channel like Jordi Alexander have actually done this. I've said, "How does my portfolio fit into his thesis?" Or Jeff Park when I had him on the channel, you know, he had this thesis around sovereign assets. I'm like, "How sovereign is my portfolio allocation?" So once again, if you have the memory system, it'll actually, you know, enable you to do these cross-checks and cross-references, which is really important.
So as you can see here, it's actually created a brief based on my personal portfolio and what's happening in the market. So April payrolls, confidence high, the inflation input, the Fed statement, earnings, the decision, etc., etc. It also gives you a risk profile. So it says, "Acceptable risk, you're still 11% Nvidia after the trim." So this is an analysis that you generate.
What you can also do, and I do this a lot, is you can just go into Claude Code because remember the dashboard is just a representation of the underlying data that you've set up on your folder system interfacing through Claude Code. So you can actually just ask Claude Code for questions about your portfolio. For example, I just asked it, "Where am I underweight given the current macro environment? Gun to your head, if you had to make one portfolio change right now, what would it be?" And it basically goes through the changes I should make. "Trim Apple because I'm overweight. Look to index more Bitcoin." And obviously, you don't just copy the AI because the AI said it, but it can get your mind wandering. And you know, I'll go through big prompting sessions where I ask like, "Where does the value accrue on AI like a hyperscaler is potentially overvalued?" And you know, if that is the case or if this person is correct, what would that mean for my portfolio? And how can I look to diversify or hedge? Or even if you're stuck in like, you know, stocks are running, I'm underexposed to equities, how can I look to enter equities based on my cash position? It actually gives you quite good advice. It can actually get the wheels turning on things that you know nothing about. And because it's plugged into your portfolio, all the information is relevant.
And the other thing that I do is I cross-reference with other LLMs. So I'll put my memory system into Codex and then I'll ask GPT the same questions that I'm asking Claude in order to find out if it has a different opinion because I want to get opinions from multiple models. So that stuff is really helpful. But ultimately, I make the final decisions. I just use AI to go through my thought process and help give me advice.
So now I want to show you the coolest part of this video, I think, and it's the automated messaging every single morning. So I wake up now to a brief on Telegram every morning which goes through what's happened overnight. So I don't need to scroll through X and see what might be relevant to my portfolio because I have all the information that's relevant to my portfolio that gets sent to me overnight and then I can decide in the morning whether I want to consider making any decisions later in the day. Just this awareness overall, understanding my position every morning, understanding my cash position every morning, understanding the risk that I'm taking on every day. This is really helping me now, especially in an environment that is this volatile compared to what was happening before where I would, you know, essentially bury my head in the sand or, you know, I'd just be busy and I wouldn't check my portfolio. And that led to a lot of complacency like not profit-taking, not managing risk, not buying when there were big corrections because I simply wasn't aware of things.
So really the best part about this AI system is not that it's going to take over investing for you or automate absolutely everything. It's just giving you that extra awareness that is going to enable you to make better decisions. Think of it as a force multiplier.
So to actually connect it into Telegram in Claude Code, and this is part of the build, you want to set up BotFather. So you basically find BotFather on Telegram. So you open Telegram, search BotFather, then you create the bot/newbot, then you copy the token into Claude Code, and then you message your bot. It's actually really simple. Claude Code will run you through the full setup. If you have any questions, of course, you can ask. Then you just send a message to Claude Code like, "Send me a market update based on the main headlines affecting my portfolio every morning at 5:00 a.m. UTC." So, it's ready when I wake up and it's going to automatically connect to BotFather.
So I'll give you an example now. I'm going to click "Send Morning Update." And what this is going to do is it's going to send an update to my phone on Telegram. So you can see it's composing the update and this will happen automatically. So obviously it defeats the purpose clicking it on the site. But I want to give you an example. And then as you can see, it's just come in on my phone. So you get this every single morning. It'll give you a market update. So how the market has gone in the last 24 hours, the impact, so what is currently happening, suggested action on your portfolio, and any major headlines that have happened overnight. And of course, you can update this based on your own preferences. And I'm just showing you a really basic example, but imagine how over time you could refine this to your own needs. For example, if you were a crypto trader, you could get an update on all your leverage positions connected to exchange APIs overnight. So you have the exact data on your positions and things that may affect those positions all consolidated into one place. Or for example, if you were a stock market low-cap hunter, you could have an AI scanning X and automatically sending alpha from the accounts that you follow without requiring you to be on X and giving you an update on Telegram every morning. That's also a video I'm going to be doing in the near future. So the possibilities are endless.
But what I wanted to do today was show you the architecture and show you a system for taking your finances and consolidating them into one place and how to maximize the intelligence of AI to get real financial advice. And the majority of that is just based on having a good memory system.
Just a few tips and tricks and you know, in case you're confused on anything, I'll run through maybe some of the questions you have. Now once again, you can use Codex or Claude Code for this. You can interface with Claude Code like a chat. Don't let the name "code" scare you. You can literally ask it for financial advice. It's still connected to Opus 4.7. The other thing I would do is put it on extra high or max. So, it's actually putting an effort and it's not giving you lazy responses or doing lazy coding. If you do want to interface with the chat, you can switch over to Claude Co-work. You won't have access to the dashboard through Co-work, but it will still be able to access the same folder. So, if you just want to talk about markets based on your memory system, you can still do so through Claude Co-work.
And the exact architecture for the dashboard build and also all of the prompts for the memory and the consulting prompt. It's all going to be available in the description below. Sign up for the newsletter where we're putting out alpha on how to use AI to level up your finances and your trading and then you'll get access to the Instagram community where you'll be able to find the link to the prompts in the pinned drive.
So hopefully you enjoyed this video. We're going to be doing a lot more AI content like this in the future as obviously now it's a big part of my investment stack. And I'll see you in the next one. Thanks for your time today. Have a nice rest of your day. Peace out.