Sitemap

Margin Call Me: How I Misused AI to Deliver Cheesy Financial Pickup Lines

5 min readFeb 14, 2025
MarginCall.me — the only GPT for financial investing pick-up lines. 🩷
TL;DR: A Valentine’s Day experiment in AI-generated pickup lines 
combines nostalgic data from a 15-year-old social media project with
modern tools like GPT, DALL·E, and web development frameworks to create
MarginCall.me — a whimsical site where an overconfident cartoon bull
named Chad Cattle delivers cringe-worthy financial-themed pickup lines.

It began in 2010

Fifteen years ago, when social media was in its infancy, I worked with a team to test Twitter engagement with a Valentine’s Day project: sharing cheesy financial-themed pickup lines.

Digital ad to go to the very manual process of receiving an Investing Pickup Line from Feb. 2010

We built a small site around the concept. The test achieved modest success, generating positive but not overwhelming engagement. After the team disbanded, the original site was taken down. However, I had somehow managed to keep the Excel file from our brainstorming sessions — the one containing all those cheesy lines — and years later, I decided to use that data for a new experiment in today’s AI-driven world.

Feeding the GPT

I fed those original items into a GPT generator to teach it to create its own cheesy finance pickup lines. Using OpenAI, I built and trained a GPT on my old CSV data. The first attempts were rough — while it could recite back the provided lines, its new creations lacked wit, and some jokes veered into overly suggestive territory. I wanted to keep things fun and silly, not make anyone uncomfortable.

Through several rounds of fine-tuning, I taught it to include laugh lines while avoiding graphic content and excessive innuendo. Finally, I achieved the perfect balance: confident, cringey, and delightfully ridiculous.

MarginCallMe the GPT on OpenAI

Chad Cattle

Creating the right visual identity proved challenging. Human designs felt too realistic — straying too close to “that guy” we’ve all met and instantly disliked. He’s right there in the ads for the version over a decade ago. He’s cringe and to be avoided.

After abandoning human likenesses, I explored metaphorical financial animals — bulls and bears — using DALL-E. The bears oscillated between too cute and too ferocious. The bull started sheepish and shy — this first iteration, “Billy Bull,” was so endearingly hopeful that you couldn’t help but root for him, even while dreading his use of these terrible lines.

“Billy Bull,” was so endearingly hopeful that you couldn’t help but root for him.

I had DALL-E dial up unearned confidence and cringiness and I was gifted Chad Cattle — a buff, overconfident cartoon bull perfect for delivering these cringey lines. Moving away from people to cartoon animals and finding one as boorish as this bull transformed the experience from awkward to fun.

Chad Cattle — a buff, overconfident cartoon bull perfect for delivering these cringey lines.

Assembling the Site

With the “Chad Cattle” concept in mind, I created a high-fidelity mockup in Figma of all the graphic elements and layout. It went pretty fast, it’s a fairly basic layout and single-page website. I used ChatGPT to brainstorm names for the site. After narrowing the options to two, I let my wallet decide — choosing the $7 .me domain over the $50 .io option. Since this project had no business model or ROI expectations, it was the prudent choice.

Using Microsoft VS Code as my IDE, I set up a new repo in GitHub deployed for site hosting with Netlify. To expedite development, I chose a web-based approach over an app. I selected Astro as my lightweight framework — perfect for a small one-page project — and implemented Tailwind for styling. With GitHub Copilot’s help on overcoming linting issues in the basic structure, I refined the styling, adjusted the Flexbox layout, and added pseudo-classes to denote the thought bubble. I ensured the static version of the site worked flawlessly across all devices, then got to work plugged it in.

Integrating the GPT

Needing to pull out my credit card and set up a wallet in OpenAI was an unexpected step, but necessary to have a funding source to make the API available for use in my website. I seeded the account with $10 and the dynamic pickup line generation began to work. Then came the most important step: knowing when to stop, call it done, and ship it — just in time for Valentine’s Day. After all, when better to deliver a cheesy pickup line?

The entire build took just over four hours and cost me $17 of new money (I already had subscriptions to ChatGPT, Figma and CoPilot). This fun project let me experiment with DALL·E (which I hadn’t explored much at all) and learn the OpenAI API, comparing it to other APIs I’d used on other small silly sites.

I welcome your feedback, come visit the working live site — MarginCall.me

Disclaimer: As always, all thoughts are my own and not my employer. While my employer may provide financial advice, they are not in the business of providing financially themed romantic pick-up-line advice. Nor is anyone as far as Iknow. If you would like to be in that business and have a pitch to turn this silly site into a lucrative business, reach out and let’s talk. Any likeness of images on the site to people in your real life is unfortunate, but is also just a coincidence. With assists from AI, all the design and development was lovingly hand-crafted by Greg Robleto. Happy Valentine’s Day.

Greg Robleto — robleto.com

--

--

Greg Robleto
Greg Robleto

Written by Greg Robleto

Creative leader exploring design, AI, and product strategy. Writing on UX, CSS, leadership, and the evolving role of technology in design.

No responses yet