My goal with this newsletter is to make you an investing superhuman with AI.
I want to level set this week with a clear action plan on how you can really master these AI tools and techniques.
We've seen explosive growth in our little community and I genuinely believe we are building together the future investing.
Welcome to our new members who've joined!
Here's the 5 step plan I'll be covering:
Level 0: Mastery of Investing Strategies
Level 1: Mastery of Prompt Engineering
Level 2: Use Cursor and Think Like an Engineer
Level 3: Simple Automations
Level 4: Enterprise Grade AI Buildouts
Level 0: Mastery of Investing Strategies
We need to be clear what AI can and cannot do for investing.
What it can do: Improve your efficiency by 10-100x, guide you through your gaps of knowledge, find due diligence facts you wouldn't have gotten otherwise, automate majority of tasks, and equip you with new technical skills like coding / data science.
What it can't do: Tell you the exact market inefficiency to exploit
In other words, you need to be directionally correct to "lead" the AI and have a clear understanding of your investing style.
For the hedge fund, private equity, and institutional managers who read my newsletter - understanding your style will be increasingly important as AI systems enable you to accelerate exploitation of those opportunities.
For individual investors who read my newsletter - I suggest studying strategies and tailoring it to your unique strengths (eg, if you are from Germany or Brazil, you may have an edge picking local winners).
I highly recommend the book More Money Than God - this was required reading when I did my summer internship at Blackstone's Hedge Fund Group back in the day. It shaped my views on mixing + matching different styles into my own.
Level 1: Mastery of Prompt Engineering
Your outputs to language models are only as good as the inputs (garbage in garbage out).
You need to be as specific as possible in your prompt engineering!
Why is this?
The elementary explanation for how modern LLMs work is it reads the text that’s already there and chooses the next bit of text it believes is most likely to follow.
So the more specific context you give the better results you will get.
Before I work with my AI, I like to sketch out what I am trying to achieve
Before you submit the actual prompt, I like to work with the LLM to craft a prompt together. This will turn into a multi paragraph prompt tailored exactly to what you need.
Understand the pros / cons of each major model provider (some outperform others in certain tasks). Use the right model for the right task. Here is an old comparison of platforms I made previously: https://www.davewang.ai/resources/ai-platform-comparison-investing
This is the level where you will achieve massive unlocks to investing.
Cursor is a AI developer platform where you can code using plain English. We've built some really cool stuff in the last few weeks with Cursor and you've seen how useful it can be for investing.
Cursor feels like a scary jump but if you can get past the initial learning curve you will effectively have a 'quant' as part of your investing team.
The possibilities are endless....example things you can do range from simple data analytics with Python to shipping full stack investing mobile apps or websites with proper data pipelines.
I majored in statistics in college and data science views that would've taken me days to make now can be done in <1 hour while scrolling TikTok (I literally do this lol).
My tips to get good at Cursor:
Think like an engineer or product manager. Scope out the project plan (together with the LLM!) and design plans of action to tackle bit by bit, feature by feature.
Learn how to sign up and put API codes in your .env files. This will unlock so many use cases unavailable from out of the box LLMs.
Be patient. You will get frustrated at some points but being calm and debugging together with the AI will get you your results faster (+ serve as a learning tool for future projects).
Just vibe code more. The best way to learn is by doing.
I don't have any guides yet on using Cursor for Finance ... Feel free to reply to this email if you'd find it helpful (I read and reply to every email).
Level 3: Simple Automations
Do you find yourself repeating the same tasks to your LLMs?
This is where simple automations come into play to free up your time with AI.
There are several no code automation platforms out there but my preferred tool is n8n (visual editor for hooking together workflows with AI layers).
With this, you're able to customize AI tasks with your preferred platforms (Gmail? Outlook? Slack? Telegram?) and even fully automate repetitive tasks.
Here's some examples of n8n tasks I automated:
I have a daily news scraper that sends me pre-market the top 5 pieces of news from WSJ / Bloomberg / CNBC. AI summarizes this, tells me the stocks or asset classes impacted by the news, and suggests what I should focus on for the day.
Simple automations are great for anything with deterministic workflows. Think like first year analyst to VP level work. You can get quite complex stuff (because of Python integration) like automated AI comps analysis and beyond.
Level 4: Enterprise Grade AI Buildouts
If you understand your investing strategy, this is how you can win back your team's time and shift work from "monkey work" dominant to "decision making" dominant. Which is really where the alpha is made - the deep (human) thinking about the levers that really matter for an investment.
How does an enterprise grade AI system look like?
Here's a high level example of the system I've designed for my own investing (I am primarily a crypto and equities investor).
Each pod has agents and subagents that flow from idea generation to execution of a trade. All trained on my preferred investing framework, sourcing methods, risk tolerance, etc and human approval at critical junctures like manual review of due diligence materials. The quant pod oversees the portfolio as a whole with both traditional methods (eg factor exposure) and AI methods (eg qualitative news risk indexing)
The beauty of designing your own enterprise grade AI system is you can feed it your own history and prop data to iterate on. For example, you can append an AI agent that learns from track record and can reference your trade journal (Lone Pine, one of the top HF's, pays millions in salary for this support).
The tools you need for this:
LangChain / LangGraph for building AI agents
Local hosted or cloud based AI (ChatGPT, LLama, etc)
Database storage (eg Supabase)
CrewAI for agent coordination and task delegation
API access especially to raw data (this is where it gets expensive)
If you're at a fund needing assistance implementing this, I'm happy to chat. I have a team of killers (Stanford PhDs, engineers with years of back end experience, etc) that specialize in building these systems.
That's all for this week. Let me know if you have any questions on the above action plan - I'm rooting for you.
There's a billion dollars waiting for you inside your laptop. You just need to figure out how to click the right buttons to get it out!
Personal News:
I started a YouTube - subscribe if you learn better from video format.
I am posting about building AI investing systems and outputs I see from my AI analysis. They're short form videos for now and I'll figure out long form later. I'm also taking requests from community for video topics :)
Pardon the awkwardness on the camera, I'm still figuring out how to act on camera in an engaging way (it's harder than it looks lol).
If you made it this far, forward this to a friend who might find this helpful