Show HN: I made a free tool that analyzes SEC filings and posts detailed reports (signalbloom.ai)

Done. Thank you!

Really nice of you to encourage, thank you! I've thought about some of those.

The first one is an interesting idea, but offering any real financial advice is regulated, which is a tread carefully territory

Second one is achievable and a good one at that. In fact, there is a lot of canonical data out there that when used carefully can provide a very nuanced picture of the current state of economy.

Third one, currently I have just a twitter account where the earnings reports are auto posted the second they are ready (that is, until I run out of free api limits). Their API cost for at the higher end makes it only viable if I am doing it full time and there was a clear edge by analyzing the data. My gut feeling is, most of the work will go into filtering out low signal posts idk

Was figuring out how to do the same thing, but for an Indian context, and specific to one's portfolio, by perhaps reading emails. Was still ideating it, cause I am working on another thing at the moment.

I'm interested in this topic too, but am approaching it from an auditing rather than from an investing angle.

Depending on how much time you have, the first thing I would recommend is to watch Karpathy's 'deep dive into LLMs' video (https://www.youtube.com/watch?v=7xTGNNLPyMI). It is 3.5 hours but it will give you so much instinctive understanding of how these things work and will likely save you a lot of time.

Purely from a technical standpoint, things aren't that difficult really. I'd recommend get yourself an API account with all major providers (because each frontier LLM evidently has their own strengths and weaknesses) and start experimenting with your data. The feeling you'll get when you first get an API response will keep you going! :)

Having said that, a caveat is, LLMs do hallucinate from time to time. If your data needs accuracy, definitely don't do the "throw everything at LLMs and let them do the work" but I'm sure you'll figure it out from trial and error. Good luck!

Thank you! I love this about HN (well, I've been a lurker mostly for 18 years) that people point to things you haven't even thought of.

Thanks! No we (had some help) built the full pipeline from the ground up using EDGAR archives. Cleaning up SEC data was a nightmare but we eventually got there

Much appreciated. The only donation I can think of is if you help spread the word. I learned a tough lesson this week that it is extraordinarily hard to market a product lol

P.S. If it wasn't for @dang putting it in the second-chance-pool, this post would have never been seen by more than 2 people.

I swing trade on the side, and tbh news don't matter - manually all I do is have an opinion on an absurd price range, I do a buy order at the low end, it eventually fills, I set the sell order at the high range, and one day it'll fill. If I want to earn while waiting, I do options.

Automate this maybe ? Don't follow the news because you'll never fully understand the implications: firing half the employees could be good news because it relieves cash stress or bad news because it means no more sells of the company's product. Better to at least just look at the post-news trend and position with it?

I think most people end up with a timing issue: they want to predict time of the change of movement but why don't they just predict the price ? Remove time, and you'll never "lose money" - but your time-value will maybe be less efficient than positioning on an ETF and waiting.

Basically, if one could do a trading robot making profit overall, based on public news, it would already be done, and therefore you would be unable to compete. Flash news, it's already been done. You're like a little cell appearing in an ocean of sharks, the evolution of your competition already outpaced you.

I disagree with this. Zero is an arbitrary and often useless intersection. A stock worth $300 is going to show meaningless movements at that scale.

One child comments makes a better suggestion. Display only a change in % since period start. Who cares about the absolute price. You only care about percentage change.

And yes, all charts would automatically start from 0 as a side effect.

Charts like this are standard practice in finance. You're interested in the relative movement, not absolute numbers, so a chart that starts from zero is practically useless since the change you're trying to see is too small.

> Always zero out your graphs

Great way to end up with useless graphs where you have a tiny line at the top that's been compressed to the point where you can't see any changes...

The "rule" about ensuring axes are zeroed is for bar charts not line graphs.

The earnings filings are available on SEC. Where the product makes a difference is by analyzing interesting patterns and contextualizing it with different means

e.g. this NVDA earnings report: https://www.signalbloom.ai/news/NVDA/nvidia-revenue-soars-ma...

beyond the headline revenue/EPS miss/beat which is trivial, it correctly identifies a trend of margin compression and charts it. It also identifies the gaming revenue divergence from the datacenter revenues and so on. You can check out pretty much any report and odds are, you'll find things that weren't called out in the earnings release

It really seems like it is, I made this YouTube channel to do it as well

https://youtube.com/@corporatedecoder

Curious what you've had in mind here regarding auto generated content and using SEO? How is this supposed to make money?

This is really hard to quantitatively track, but there is a similar concept called the "hairy back chart" where you compare management guidance to actuals (or street guidance, which is usually developed in conjunction with management, to actuals). E.g., https://www.mckinsey.com/capabilities/strategy-and-corporate...

Thanks, both are really good suggestions!

That's an interesting idea, thinking about it.

The plan is to address all SEC forms anyways so this will be a natural extension

I think this goes back a bit to what personas you are trying to solve for. Many companies have a generic investor presentation on the website.[1] Every professional investor will have access to sell-side coverage and initiations of coverage is basically this (an overview of the company, key drivers, etc.). Retail investors will often have access through wealth managers. And if it went public relatively recently the IPO deck will function as this.

[1] E.g., https://s27.q4cdn.com/632832908/files/doc_financials/2024/q4...

would be interested to follow along! Where would you source the filings for european stocks?

I lost track of the hours lol. Technically it is a side project but I have been putting all my spare time into it. I had some help too.

For European stocks, is there a common EU regulatory body where all filings are submitted? That would be a good start

And way before the current wave of LLMs, I did some consulting for a company that provided these kind of summaries over a decade ago, and I’m sure they weren’t that early themselves

It's pretty trivial to find insider trading looking activity as a human, so I doubt it would be particularly difficult to write up an algorithm to do it. Of course you wouldn't come close to catching everything but you'd catch blatant stuff.

The problem is as a private citizen/company (1) you have no idea who placed the trades and (2) even if you did, you can't enforce the law.

Unusual Whales tracks as much as they can... https://unusualwhales.com/insiders/trades?transaction_codes[...

Finding insider trading is not the difficult part, it's being able to prosecute it that is going to be a challenge.

I have a prototype doing just that! That is on the roadmap :)

Although, it is not as exciting as some make it sound. Most insider trading reports are utterly boring. There are a small percentage that are valuable though

By "insider trading" I assume you mean illegal insider trading. There are a few levels:

(1) explicitly illegal insider trading where you trade on MNPI and hope no one notices (e.g., CEO sells shares before bad-news results). I don't think you could track this because usually the person trading is either not a prolific insider (a janitor who sees a draft report in the trash can) and therefore how would you track them OR a prolific insider using some sort of shell (e.g., wife/cousin) [or some combo]

(2) totally legal and ethical "insider trading." The SEC allows insiders to trade, you just have to do it with proper disclosure and in open periods. As noted Unusual Whales and others track this because in theory this is a signal—if mgmt is selling does that mean they no longer believe long-term in this company (irrespective of near-term MNPI)? The flip side obviously is that management cannot eat shares and sometimes wants liquidity for lifestyle reasons.

(3) everything in between—shadow trading, auto selldowns that are turned off if you realize there is good news, etc.

Tracking (2) is as mentioned kind of useful for investors but I would argue not a great signal, tracking (1) is mostly useless because it's a post-hoc analysis (most analyses here are—there was a massive spike in trades before/after a big announcement, or more obviously, a big block trade) and tracking (3) is similar to tracking (2). Btw there is also a lot of academic evidence that insider traders of type (1) overestimate the importance and misestimate the impact of their MNPI—knowing J&J had a drug that passed FDA may not matter if (a) there are twenty drugs that failed and (b) the street assumed it would pass. (none of this is legal or trading advice)

This is a really good suggestion (changing "news" to "latest filings"), done, thanks!