Hi guys,
I have been workign with AI-coding tools such as codex cursor claude etc.. for about a year now for work, but have been trying it out for automating DD's / building my own screeners using my preferred method of (value) investing, which is mainly in the smaller cap space with asset backed floors.
I was wondering if anyone wanted to share their success stories if people have been doing similar stuff? It's worked out pretty well for me so far. The real edge for me is being able to distill the methods I used before and trying to automate those + building a system that improves itself. Every time I run into something in a stock DD I use the AI tools to add the findings into the system in order to catch it in future DD. That's where building your own custom solution using AI assisted coding tools really does wonders.
Pulling data from yfinance and EDGAR is free, there are many free API's to build your own custom screeners, i'd recommend everyone to try it out for themselves!
Some tips i have if you want to try this out:
-
Seperate between the deterministic layer and the AI-analysis layer. This simply means you build a set of metrics with a script, so you don't have to use an AI to calculate it for you, this saves usage.
-
Find methods to point LLM's in the right direction and decide which perspectives / directions they should be pointed to, that's where the deterministic part can become a real differentiator. For example: the industry determines whether or not which metrics are important. An LLM will probably figure that out by themselves, but a more fine-tuned screener can really help pointing it into certain directions that addsomething to the DD / valuation thesis.