I Built a Stock Research Site So I Could Stop Pretending to Understand SEC Filings

Published March 11, 2026

Side Project SEC AI Investing

TL;DR: I live in the Bay Area, surrounded by IPO hype, but couldn’t actually read an SEC filing — 340 pages of legalese, made worse by English not being my first language. So I built Stockadora: a free site that uses AI to turn dense SEC filings into plain-English summaries, updated daily, no login required. This post is the origin story.

Series: Building a Side Project That Runs Itself

I’ll be honest with you: I am not a serious investor. I don’t have a Bloomberg Terminal. I don’t have a finance degree. I have a Robinhood account, a loose interest in tech companies, and a habit of getting curious about things at the worst possible time — usually 11pm on a weeknight when I should be playing games.

I live in the Bay Area. That means I’m surrounded by people who work at startups, and every so often one of those startups announces an IPO. People in my circle start talking about it — whether to buy in, what the company actually does, whether the valuation makes any sense. I was curious but had no idea how to evaluate any of it. It felt like everyone around me had some intuition I hadn’t developed yet.

Then Robinhood added IPO tracking to the app. Suddenly there was a feed of upcoming IPOs, right next to my portfolio. I found myself scrolling through it, reading company names I half-recognized, wondering what they actually did and whether they were worth anything.

But I didn’t want to blindly jump in. Buying a stock just because it showed up in my feed felt like gambling. I wanted to understand what I was looking at — at a minimum, what the company does, who runs it, and what they themselves say the risks are. That’s all public information. Every company going public has to file an S-1 with the SEC. It’s all right there.

So I opened one.

Three hundred and forty pages. Dense legal language. Risk factors that appeared to have been written by a team of lawyers specifically paid to make investors feel nothing. And here’s the thing: English isn’t my first language. I can read it fine, but when a document is written to be deliberately opaque, with financial jargon stacked on top of legal boilerplate, I hit a wall fast. I closed the tab.

That was the itch.

What Is SEC EDGAR, Anyway?

If you’ve never heard of it: EDGAR (Electronic Data Gathering, Analysis, and Retrieval) is the SEC’s public database of company filings. Every public company in the US has to file there. It’s free. It’s public. And it’s… not exactly reader-friendly.

The main filing types you’d care about as a casual investor are:

S-1: The IPO registration statement. Usually massive, contains the full business model, financials, and risk factors.
10-K: Annual report. Think of it as the company’s end-of-year recap, but written in legalese.
8-K: Current report. Used for material events — earnings surprises, leadership changes, acquisitions. Usually shorter, but they come in fast.
Form 4: Insider trading disclosures. When a CFO buys or sells shares, this is where it gets recorded. These are small documents but surprisingly interesting.

All of these are publicly accessible. All of them are technically readable. Most of them are genuinely painful to parse if you’re not trained for it.

The “Itch” Moment

After giving up on that S-1, I went looking for something better. There are services out there that summarize SEC filings — but the good ones are behind paywalls, and most of the free ones are either shallow (“Company X filed a 10-K today”) or clearly designed for professional analysts, not people like me.

I didn’t want a financial analysis service. I just wanted someone to explain what the filing said in plain English — minimal jargon, no assumed background knowledge, just the substance. What does this company actually do? What are the real risks they called out (not all 47 of them)? Did any insiders sell a suspicious amount of stock before that earnings miss?

That’s a narrow but real use case. And it turns out, modern LLMs are pretty decent at it.

So I started sketching something out. The goal was simple:

Pull filings from EDGAR automatically
Run them through an AI model to generate plain-language summaries
Publish the results as a free website — no signup, no subscription, no support queue

That last point mattered to me. I’ve seen too many side projects turn into customer service jobs. I wasn’t interested in that. I wanted to build something useful, put it online, and then go back to my normal life — which includes, I’ll be honest, a not-insignificant amount of PC gaming.

The Audience I Was Building For

There’s a large gap between “I have a Robinhood account and occasionally wonder about the companies I’ve invested in” and “I am a professional analyst with access to Bloomberg.” The casual investor crowd is enormous and almost entirely underserved by financial data tools.

Bloomberg Terminal costs north of $20,000 a year. The professional-grade services like FactSet or Refinitiv aren’t much better. Even mid-tier services aimed at retail investors tend to require a subscription, assume you know what EBITDA is, or present data in a format that’s dense enough to be intimidating.

And none of them are designed for someone who isn’t a native English speaker and doesn’t have a finance background. SEC filings are hard to read even for people who grew up with English as their first language. For people like me, the combination of legal language, financial jargon, and sheer document length makes them effectively inaccessible without help.

What I wanted was closer to “a smart friend who happened to read the filing for you and will give you the summary in five minutes.” Plain English, no assumed background knowledge, accessible to anyone who’s curious about a company.

The not-financial-advice angle was also important to me from the start. Stockadora is about making public data readable, not about telling anyone what to buy or sell. There’s a real difference, and I wanted that to be front and center.

The Recurring Challenge

Throughout all of this, I had one recurring constraint: I have a full-time job. I’m not building a startup. I’m not looking for VC money. I’m a developer with weekends and evenings and a list of games I want to get through before I die.

That constraint shaped every single technical decision I made. Not “what’s the best tool for this?” but “what’s the tool I can maintain at 11pm when something breaks?” Not “how do I scale this to a million users?” but “how do I make it so this thing runs itself while I’m not looking at it?”

That became the unofficial motto of the project: I still want to play games on weekends — this thing needs to run itself.

And here’s the surprising part: that constraint actually led to better architecture than I would have built otherwise. When you can’t accept the overhead of a complex system, you’re forced to keep things simple. Simple systems fail less. Simple systems are easier to fix when they do fail.

What It Became

The result is Stockadora — a site that publishes daily AI-generated summaries of SEC filings, Form 4 insider activity, and news digests for public companies. It’s free. There’s no login. It updates automatically every day while I’m doing other things.

Over the next several posts, I’m going to walk through how I built it — the architecture decisions, the tech stack choices, what went well, what broke, and what it actually costs to run. These posts are not a tutorial and not a pitch. They’re just an honest account of what it looks like to build a real project with real constraints.

Next up: the most counterintuitive decision I made — using a static site generator for a financial data platform.