Let's start with the problem.

Every quarter, thousands of public companies file documents with the U.S. Securities and Exchange Commission. 10-Ks. 10-Qs. Form 4s. Proxy statements. IPO registrations. These filings are public record — legally mandated disclosures that exist precisely so that markets can function fairly, so that any investor, not just the well-connected ones, can know what's happening inside the companies they own.

In practice, this system has a flaw.

A Goldman Sachs analyst has a Bloomberg Terminal, three monitors, a team of associates, and proprietary NLP pipelines that process EDGAR data before the market even opens. A retail investor has a brokerage app with a "News" tab and whatever shows up on financial Twitter.

The information is the same. The infrastructure to process it is not.

`sec-analyzer-ai` exists to close that gap. It is a Telegram bot that monitors SEC filings for any company you choose, analyzes them using free large language models, compares current risk factors to prior filings to flag what changed, tracks insider sentiment over time, and delivers everything to your phone — automatically, in plain language, at zero cost.

That's the pitch. Here's the honest story behind it.

I want to be transparent about how this project actually got built, because I think it matters.

There was no team. No venture funding. No roadmap document reviewed by a product committee. Just me, a problem I found genuinely interesting, and an enormous amount of time spent in the kinds of places where developers go when they're stuck — deep in forum threads, digging through GitHub issues on libraries I'd never used before, asking AI assistants questions I was slightly embarrassed not to already know the answers to.

Sometimes the solution came from a three-year-old Stack Overflow answer. Sometimes from a model that confidently gave me the wrong approach, which I then had to unbreak. Sometimes from just trying something, watching it fail, and trying something else.

I used nearly every resource available to me. Documentation I barely understood. Community posts from people solving similar problems. Trial and error at a scale that would be inefficient in any sane engineering environment but is, I've found, how hobbyist projects actually get built.

The result — after all of that — is a tool I'm genuinely proud of. Thread-safe architecture. Bilingual UI. Thirteen form types. A risk-factor diff engine. A sentiment trend tracker. A side-by-side company comparison command.

And here is the part I find both funny and a little humbling: so far, the only person who has used it is me.

That's not a complaint. It's an invitation.

That's the trade. Your bug report, my gratitude, and hopefully — a better tool for everyone.

What It Actually Does (The Short Version)

For the first-time reader, here is what sec-analyzer-ai does in plain terms:

Core capability: You give it a list of stock tickers. It watches the SEC's EDGAR database for new filings from those companies. When something arrives — a quarterly report, an insider transaction, a major ownership change — it analyzes the document using a free LLM, extracts the parts that actually matter, and sends you a structured summary on Telegram. No subscriptions. No servers. Runs on an old Android phone in the background via Termux.

Supported forms: 10-K (annual), 10-Q (quarterly), 8-K (material events), Form 4 (insider buy/sell), SC 13G/D (institutional ownership changes), S-1 (IPO filings), DEF 14A (proxy/executive compensation), 20-F and 6-K (foreign companies on US exchanges), and several others — 13 types in total.

Languages: English and Turkish UI, switchable live with /setlang. The LLM output follows the selected language.

I have my own backlog. But the most valuable features in any tool are the ones the builder couldn't see. You might use this differently than I do. You might track different companies, care about different signals, need outputs in different formats. I genuinely don't know — because, as I mentioned, I'm currently the only user.

🔗 Repository: github.com/authorturker/sec-analyzer-ai

If You Find a Bug, That's the Best Thing You Can Do. Seriously. A bug report from someone who isn't me is worth more to this project right now than a five-star review. If something breaks, if the wizard doesn't flow correctly on your setup, if a form type returns garbage output, if a command behaves unexpectedly — open a GitHub issue. Or just message me. The repository link is above.

You won't be complaining. You'll be contributing.

If you become one — even briefly, even just to report a single error — you'll have helped turn a hobby into something that might actually matter.