I built an AI that reads boring SEC filings so I don't have to

New Hey IndieHackers!

I've been building a financial research tool called MoneySense, and I just shipped the feature I’ve wanted for myself for years: Automated SEC Filing Analysis.

The Problem I invest in stocks, but I’m lazy. I know I’m supposed to read the Annual Report (10-K) before buying, but they are 100+ pages of dense legal jargon. I usually just read the press release (which is just marketing) and ignore the filings.

This burned me recently when a stock I owned dropped 20% because of a "hidden" risk factor I would have seen if I had just read page 60 of the report.

The Solution I built a pipeline that pulls live filings from the SEC EDGAR database, cleans the text, and feeds specific sections (Risk Factors, MD&A) into an LLM with a strict system prompt to "act like a forensic accountant."

How it works

1. You type a ticker (e.g., NVDA).

2. It fetches the latest 10-K.

3. It generates an "Executive Summary" focused purely on Risks and Red Flags (ignoring the marketing fluff).

4. It extracts a sentiment score based on the management's tone.

WIP / Challenges The hardest part was token limits and "hallucinations." I had to build a chunking system to feed the document in parts so the AI wouldn't lose context or make up numbers. It’s about 95% accurate now compared to my manual reading.

Try it out It’s live here: https://moneysense.ai/sec-analysis

I’m looking for feedback specifically on the "Red Flag" detection. Does it feel too aggressive? Too passive?

Thanks for checking it out! 🚀