Show HN: I lost 15% to Congress' lag, so I built a trade-sniping tool
A couple years back, my girlfriend and I jumped on the congressional trading bandwagon. You know the drill—politicians like Nancy Pelosi disclose their trades (eventually), and some of them have suspiciously good timing. We copied a few, and it worked. In months, we were up +38%. Even now, with the market tanking, we’re at +6%. Not bad, right?But here’s where it fell apart: Congress gets a 30-day grace period to file trades, and I had to manually check sites like QuiverQuant to catch them. I’m no robot—I’d forget, get busy, or just miss the window. One time, Pelosi bought a tech stock before a big announcement. I didn’t see the filing until it was up 15%. I could’ve been in at the ground floor, but nope—too late. That loss hit me hard, and I knew I couldn’t keep doing it by hand.So, I decided to build something: a tool that tracks Congress members’ trades and emails you the second a filing drops. Pick your politicians, set it, forget it—no more missed gains. Simple idea, brutal execution.the first issue was OCR, filings come from disclosures-clerk.house.gov—mostly PDFs, some handwritten. I tried Tesseract (free OCR tool), but it barfed on anything not typed. Handwritten trades? gibberish. So I moved to openai’s gpt API to decode the mess. It’s not perfect, but it can spot “50 shares of AAPL” in a scrawl. For the worst ones, I just send the raw PDF—figure it out yourself!I’m not a pro dev—day job keeps me busy—so this was all after-hours. Think 10 p.m. to 2 a.m., fueled by coffee and spite. My girlfriend deserves a medal for not dumping me. Anyhoo, after building it, I’ve been using it myself, and it’s clutch. Last month, it pinged me about a Congress member buying a biotech stock. I got in early—two weeks later, up 12%. Without it, I’d have missed it again. So I'm really really advocating its usefullness as it solves my problem!I am a number guy, and I really appreciate when people are transparent, and I really love to see and read when other people openly discuss costs. Her we go: I have 20 beta testers (me, friends, a few randos), I'm getting ~$50/month from paid tiers. Surprisingly I'm not quitting my job anytime soon. The real issue here are the costs: $70/month (AWS) + $10/month (gpt API) = -$30/month profit. Yep, I’m in the red, but it’s a passion project for now. The pricing i worked out is: Free tier (1 random politician, delayed), Bronze ($2.99/month, 1 choice, real-time), Silver ($5.99/month, 4 choices), Gold ($12.99/month, 20 choices). The break-even point isn’t too far off, though. If everyone subscribed to the Bronze tier at $2.99, I’d need around 27 paying users, which means I’m short by just seven. If they all picked Silver at $5.99, I’d only need 14 in total—six more than I have right now. And if every single person went for Gold at $12.99? I’d break even with just seven, meaning I’d only need two more. Of course, reality isn’t that clean, and the mix will be all over the place. But the numbers tell me that if I can get to around 20-25 paying users, this thing starts sustaining itself, the architecture (aws fargate with lambda function to send the emails) will be more than enough to bear 500 users. The gpt cost won't increase since no matter the users i'll be processing the same number of pdfs filings. Comments URL: https://news.ycombinator.com/item?id=43368488 Points: 9 # Comments: 0
A couple years back, my girlfriend and I jumped on the congressional trading bandwagon. You know the drill—politicians like Nancy Pelosi disclose their trades (eventually), and some of them have suspiciously good timing. We copied a few, and it worked. In months, we were up +38%. Even now, with the market tanking, we’re at +6%. Not bad, right?
But here’s where it fell apart: Congress gets a 30-day grace period to file trades, and I had to manually check sites like QuiverQuant to catch them. I’m no robot—I’d forget, get busy, or just miss the window. One time, Pelosi bought a tech stock before a big announcement. I didn’t see the filing until it was up 15%. I could’ve been in at the ground floor, but nope—too late. That loss hit me hard, and I knew I couldn’t keep doing it by hand.
So, I decided to build something: a tool that tracks Congress members’ trades and emails you the second a filing drops. Pick your politicians, set it, forget it—no more missed gains. Simple idea, brutal execution.
the first issue was OCR, filings come from disclosures-clerk.house.gov—mostly PDFs, some handwritten. I tried Tesseract (free OCR tool), but it barfed on anything not typed. Handwritten trades? gibberish. So I moved to openai’s gpt API to decode the mess. It’s not perfect, but it can spot “50 shares of AAPL” in a scrawl. For the worst ones, I just send the raw PDF—figure it out yourself!
I’m not a pro dev—day job keeps me busy—so this was all after-hours. Think 10 p.m. to 2 a.m., fueled by coffee and spite. My girlfriend deserves a medal for not dumping me. Anyhoo, after building it, I’ve been using it myself, and it’s clutch. Last month, it pinged me about a Congress member buying a biotech stock. I got in early—two weeks later, up 12%. Without it, I’d have missed it again. So I'm really really advocating its usefullness as it solves my problem!
I am a number guy, and I really appreciate when people are transparent, and I really love to see and read when other people openly discuss costs. Her we go: I have 20 beta testers (me, friends, a few randos), I'm getting ~$50/month from paid tiers. Surprisingly I'm not quitting my job anytime soon. The real issue here are the costs: $70/month (AWS) + $10/month (gpt API) = -$30/month profit. Yep, I’m in the red, but it’s a passion project for now. The pricing i worked out is: Free tier (1 random politician, delayed), Bronze ($2.99/month, 1 choice, real-time), Silver ($5.99/month, 4 choices), Gold ($12.99/month, 20 choices). The break-even point isn’t too far off, though. If everyone subscribed to the Bronze tier at $2.99, I’d need around 27 paying users, which means I’m short by just seven. If they all picked Silver at $5.99, I’d only need 14 in total—six more than I have right now. And if every single person went for Gold at $12.99? I’d break even with just seven, meaning I’d only need two more. Of course, reality isn’t that clean, and the mix will be all over the place. But the numbers tell me that if I can get to around 20-25 paying users, this thing starts sustaining itself, the architecture (aws fargate with lambda function to send the emails) will be more than enough to bear 500 users. The gpt cost won't increase since no matter the users i'll be processing the same number of pdfs filings.
Comments URL: https://news.ycombinator.com/item?id=43368488
Points: 9
# Comments: 0