[HN Gopher] Show HN: I built an AI tool to analyze SEC filings t...
___________________________________________________________________
Show HN: I built an AI tool to analyze SEC filings the minute
they're released
Hello everyone, I built this tool on Next/Node to automatically
analyze new filings from the SEC and probe the Edgar API for new
filings 24/7. We use AI to analyze the filings the second they are
released. Free accounts to look at real filings (automatically
updated) are available to any who sign up. If you have any
questions feel free to ask.
Author : docdeltaneer
Score : 51 points
Date : 2024-12-03 19:25 UTC (3 hours ago)
(HTM) web link (docdelta.ca)
(TXT) w3m dump (docdelta.ca)
| toomuchtodo wrote:
| How confident are you in wiring this up to trade immediately off
| of the output?
| docdeltaneer wrote:
| Having tested this extensively against historical market
| movements, I'm quite confident in its analysis accuracy. After
| running it against countless filings and tracking post-filing
| price movements, the system's risk assessments and change
| detection have proven remarkably reliable - especially for
| catching subtle shifts in risk factors and management tone that
| often precede larger market moves.
|
| I personally rely on it for my own decisions, but like any
| tool, it's most effective when used as part of a broader
| analysis process. The system is great at rapidly surfacing the
| most significant changes in these massive documents and
| contextualizing them against historical patterns. That's
| incredibly valuable for making timely decisions, especially
| since the market often takes days or weeks to fully price in
| the implications of complex filing changes.
|
| That said, the analysis code has built in multiple validation
| layers and error handling because filings are just one piece of
| the market puzzle. The code is designed to be conservative -
| it'll fail closed rather than make questionable assertions. But
| for what it's designed to do - deep analysis of filing changes
| and their implications - it's proven to be a reliable tool for
| actual decision-making.
| mistrial9 wrote:
| > The system is great at rapidly surfacing the most
| significant changes in these massive documents and
| contextualizing them against historical patterns.
|
| probably worth capturing a few successful situations and
| documenting that in some marketing way.. also probably worth
| capturing a few Very Big Fails and making a ticket out of
| that internally, to improve the system. Seize the day!
| ezero wrote:
| > Having tested this extensively against historical market
| movements, I'm quite confident in its analysis accuracy.
|
| This doesn't say anything. Exactly how much have you tested
| and what were the results?
| cortesoft wrote:
| Most releases happen after trading hours
| toomuchtodo wrote:
| Agreed, my focus is on data released during trading hours,
| intentionally or inadvertently.
|
| https://www.youtube.com/watch?v=ag14Ao_xO4c
|
| EDIT: To those suggesting extended hours trading, check out
| what the order book looks like and who the market maker is.
|
| https://www.finra.org/investors/insights/extended-hours-
| trad...
| ryeights wrote:
| Yes, but you can still trade pre-market and after-hours.
| ssijak wrote:
| you can trade post hours
| jebarker wrote:
| Naive question here: wouldn't the upside of keeping this secret
| and trading on it's output yourself far exceed what you can make
| in subscriptions?
| baal80spam wrote:
| My first question when I saw the page.
| cj wrote:
| I think this tool surfaces information not previously
| immediately digestible by retail investors.
|
| Institutional investors have had systems like this in place
| for decades. LLMs might improve parsing the data in some
| ways, but this is (and was) completely doable before the LLM
| era, meaning there's probably not a huge amount of secret
| sauce here worth protecting.
|
| If anything, a project like this simply improves the
| information asymmetry between retail and institutional
| investors.
| kelseyfrog wrote:
| If it works, yes. If it doesn't work, no it's better to let
| others convince themselves.
|
| It's the same way creating your own get rich quick scheme is
| the actual way to get rich quick, not following someone else's
| scheme.
| colordrops wrote:
| Sell shovels during a gold rush.
| andrewmcwatters wrote:
| Everyone has access to this data.
| sprobertson wrote:
| The information is freely available, this just makes it more
| readable
|
| https://www.sec.gov/cgi-bin/browse-edgar?action=getcurrent
| beardedwizard wrote:
| It's this. Just doing basic analytics on SEC data for free
| provides a lot of value to the niche looking for it. A few
| players have realized that ingesting and normalizing the data
| is half the battle and charge for very small subsets of data,
| for example cybersecurity breaches.
|
| Most of the data is unstructured, or worse Edgar xml, adding
| yet another barrier.
|
| I suspect this case is mostly about sentiment analysis and
| summary using LLM. The claims about reacting before the
| market seem egregious.
| cortesoft wrote:
| These releases usually happen after trading hours, so instant
| analysis isn't super helpful for making trades.
| bartread wrote:
| Plenty of after hours trading happens around these releases
| so there's certainly the potential for them to be useful to
| people engaging in that trading.
| xivzgrev wrote:
| Sell pick axes to gold miners. More predictable!
| rmah wrote:
| Typically, you won't be able to make any trades for hours after
| the info is released anyway.
| bartread wrote:
| Not true. I can't speak to the situation in the US but, in
| the UK, there are certainly platforms that enable after hours
| trading on US markets even for retail traders.
| mike_d wrote:
| It might kick out 2 or 3 "good idea" trades a year that you
| could milk, but the real value in this data is getting it
| quickly into the hands of people who manage sector portfolios.
|
| Say for example a filing happens that a company will be
| impacted by trade sanctions. You could short the stock as a
| naive investor and maybe make a few bucks. Someone with sector
| knowledge will rebalance their portfolio knowing that customer
| demand for widgets will shift to sprockets which in turn
| require doodads. AI isn't there yet because it requires a lot
| of non-public knowledge to really make money off of.
| docdeltaneer wrote:
| The value is in scaling this beyond what any individual could
| do. Yes, the data is public, but building reliable tooling
| around it, continuously validating the analysis, and expanding
| coverage across all filings creates far more interesting
| opportunities than solo trading.
|
| We've already seen how just having basic sentiment and risk
| analysis has helped our users catch significant changes they
| would have missed - like subtle shifts in supply chain
| disclosures that preceded market moves. As we add features like
| more filing types, earnings call transcripts, and cross-company
| competitive insights, the system becomes more valuable for
| everyone.
|
| The best trade isn't always the obvious one.
| bko wrote:
| Unfortunately much of earnings responses is written after the
| fact. If the stock price goes up you can point to some positive
| parts. If it goes down, you can point to some negative parts. I
| see that all the time. "Earnings beat estimates but analysts
| were worried about declining growth in app related earnings and
| looming regulatory threats". Sometimes they even switch! The
| stock goes up, reporters write their pieces, then it goes back
| down and switch over to another explanation.
|
| I'm reminded of the headline "More than 16M Americans have lost
| jobs in 3 week, dow's best week since 1938"
|
| https://www.reddit.com/media?url=https%3A%2F%2Fi.redd.it%2Fr...
| mschuster91 wrote:
| > I'm reminded of the headline "More than 16M Americans have
| lost jobs in 3 week, dow's best week since 1938"
|
| That's likely a significant contributor behind the November
| election results in the US and across Western countries in
| general. People aren't dumb, they spot through politicians
| and media claiming "good economy numbers" while their own
| paychecks don't come close to keeping up with inflation.
|
| The economy numbers may be at record highs, but way too much
| of the wealth ends up in the pockets of a very select few.
| z0r wrote:
| Who is Doc Delta?
| whiplash451 wrote:
| You would probably increase your subscription rate by not putting
| a signup page before the demo.
| JSDevOps wrote:
| Your trading way after the fact in some cases.
| carabiner wrote:
| It's priced in. I guarantee every major hedge fund had a
| prototype running within hours of GPT API being made public, and
| they're probably running it on real money right now.
| ezero wrote:
| You are charging way too much and offering too little. $29/month
| to track 15 companies, $89 to track 50, and "Contact sales" for
| unlimited. The "AI Analysis" you have is too brief and doesn't
| tell me anything useful as an investor.
|
| Compare this to what https://quartr.com/products/quartr-core is
| offering for $20/month.
|
| I know how difficult it is to build a tool like this and you've
| done a decent job. Just doesn't make sense to me to price it like
| this.
| gregdoesit wrote:
| As someone who used Quartr: they do not offer the ability to
| look or search actual filings at the $20/month plan. Full text
| search starts from $500/month, and is an annual contract (so
| $6,000/month.)
|
| Pricing for these services is not cheap, given it can be very
| helpful for professional traders.
|
| (I'm no professional trader, and not even a trader: I would pay
| for a decent service offering full text search for transcript
| search, to use it a few times per month. Still not found a
| product that does it at a sensible price point for my use
| case.)
| myhau wrote:
| This has similar vibes to The Infamous Dropbox Comment:
|
| https://news.ycombinator.com/item?id=9224
| luxpir wrote:
| I like seeing trading tools released. Kudos.
|
| Barriers are falling, fees and such. APIs popping up and interest
| in markets growing. I'd be bullish on retail trading and its
| healthy future.
| rurp wrote:
| If you mean that retail trading volume and opportunities to
| make money from it have a healthy future then I agree. But if
| you mean that retail traders themselves have a good outlook, in
| the sense that they will end up with more money overall than if
| tools and access were more limited, I couldn't disagree more.
| Retail stock traders have a _terrible_ track record overall,
| and bits of additional market information here or there will do
| nothing to change that. Increased access to trading results in
| amateurs having less money.
| vouaobrasil wrote:
| I think the world would be a better place if there were less
| efficiency in the market, not more.
| bdefig wrote:
| Have you tried selling to high-frequency traders / hedge funds?
| If they bite, it's probably not even worth having a low-
| cost/self-serve tier.
| richardw wrote:
| I'd guess they're already using this signal and if not they'd
| be more likely to do it at competitively high speed and
| internally, where they can control and iterate? Maybe crypto
| HFT since that's a big more sluggish.
| instalabs wrote:
| I'm in the business of selling AI tools to high-frequency
| traders / hedge funds and the bad news is that, most of them
| are so risk averse and/or technically illiterate that sometimes
| we joke that there's big alpha to be had in starting a hedge
| fund that simply embraces AI.
| GoldenMonkey wrote:
| Let's do it :)
| Etheryte wrote:
| I think you might be slightly misunderstanding why you're
| struggling to sell. In finance, correctness matters. A tool
| that's mostly right most of the time, but radically incorrect
| sometimes is a very hard sell when there's battle tested and
| well understood boring tools out there that do the same
| thing.
| addicted wrote:
| > there's big alpha to be had in starting a hedge fund that
| simply embraces AI.
|
| Maybe hedge funds are completely out of step with the rest of
| the finance world, but much of finance has been using AI for
| a long time. AI is old news at this point.
| Etheryte wrote:
| Everyone who is anyone in the finance space has been doing
| stuff like this since way before LLMs were a thing, and they've
| put a lot of time and effort into their tools. This suggestion
| is kind of like saying you should try to sell your lemonade
| stand to Coca-Cola.
| Kiro wrote:
| This is not possible without LLMs. The interesting stuff is
| not the numbers but the soft values and insights. Go check
| out the example reports.
|
| With that said, I agree that hedge funds probably already
| have their own internal versions of this.
| Etheryte wrote:
| This is a good example of Dunning-Kruger, no offense.
| Sentiment analysis has been used as a tool to guide
| investments since the late 90s and there are companies out
| there who have been doing it for decades. LLMs didn't
| really break any new ground here, they just made this more
| accessible for the masses.
| petesergeant wrote:
| > Sentiment analysis has been used as a tool ... since
| the late 90s
|
| Yes!
|
| > LLMs didn't really break any new ground here
|
| I don't think this is true. LLMs are a gigantic leap
| forward from any traditional sentiment analysis I've
| seen. Pre-2013, bag-of-words and TF-IDF were still state
| of the art. Call me out if you think I'm wrong, but hand-
| waving away the transformative power of LLMs here because
| we had some sentiment analysis approaches before seems
| unfair.
| bdefig wrote:
| That's exactly what I'm saying. If you've got something
| novel/hard to replicate, it's a no-brainer for them to pick
| up at any price. If not, that's good to know too.
|
| Btw, Coca-Cola bought Minute Maid Lemonaide in 1960. :)
| simian1983 wrote:
| Is there documentation for how you calculate the Risk Score?
___________________________________________________________________
(page generated 2024-12-03 23:00 UTC)