The AI That Helped You Text Your Crush

And how it broke every product rule and went viral anyway.

See the full clip on LinkedIn Post

Most projects try to solve a problem. This one set out to go viral.

You’ve probably seen this project before! If so, you already get the hype, which proves the point. If not, don’t worry. Here’s the catch-up:

In February, four students at Columbia DevFest built a hackathon project that felt like a thought experiment come to life: an AI that sees five moves ahead in your iMessage flirtation and tells you the most optimal thing to say to your crush.

They called it Social Stockfish.

Here’s how it worked:

• It scrapes your iMessage conversation for context

• One LLM (Large Language Model) plays as you, another as your crush

• A third model evaluates each response for how much it advances your goal (like landing a date)

• Then it picks the best first move, using Monte Carlo simulations

In their own words: "So bad at rizz we built an AI to optimize it."

Was it expensive to run? Yes.
Plug-and-play ready for users? Not really.
Did it go viral? Absolutely.

Eddy Xu’s post on X got over 12,000 likes. Madhav Rapelli’s LinkedIn post pulled nearly 800 reactions.

People laughed — then imagined actually using it.
That’s what made it stick: the joke rested on a sliver of truth, and behind the spectacle, the idea of AI-enhanced messaging carried real intrigue.

It looked amazing… in a screenshot (or a gif, or a clip).

And that’s the point.
This wasn’t a product. It was a performance, by design.

From the name to the user interface to the outrageous premise, it was built to land like a punchline.

Built for the Screenshot

There’s no shortage of advice telling builders to focus on product-market fit, tight feedback loops, and solving real problems. All good principles.

But sometimes, the smarter strategy is this: Optimize for virality.

Grab attention first. Figure out the growth strategy second.

Social Stockfish wasn’t meant to scale. It ran hundreds of API calls in parallel using multiple TogetherAI accounts just to bypass rate limits. It racked up $150 in API costs in just two days. It was duct-taped together for demo day.

This wasn’t built to serve real users in a sustainable way. It was built to stop the scroll, and in a timeline-driven world, memorability is leverage.

Most builders obsess over product and treat distribution as an afterthought. They’ll spend weeks polishing a feature and barely a day thinking about how anyone will even discover it.

But Social Stockfish flipped that. It started with distribution.

The whole build was shaped by the question: would someone share this? (Yes, obviously. If it helps you get that date, of course you’d send it to your friends.)

Every technical and UX decision bent around that axis. And when people are already talking about what you made, you’re not just launching a project, you’re seeding a distribution channel.

A waitlist for the app. A demo funnel. The starting point of a SaaS startup.

Or maybe even a product in a category of its own.
“If we continued,” Madhav told me, “we’d probably try putting it in smart glasses [another one of Eddy’s viral builds]. But tech isn’t there yet—for the latency and the glasses themselves.”

You don’t scale into attention. You start with it and scale from there.

What You Can Steal

If you're building something right now, here are some questions to pressure-test your idea:

• Would someone screenshot it and send it to a friend?

  (Social Stockfish had a clean UI, a clear payoff, and an absurd promise.)

• Would someone brag about trying this?

  (“AI that helps you text your crush” lives right between a joke and a flex.)

• Can you tell what it does in 10 seconds with no explanation?

  (The demo clip said it all: AI reads your texts, simulates the convo, and picks the best move. Done.)

Social Stockfish didn’t win because it solved a hard problem.
It won because it looked like it solved something absurd — with flair.

And in an attention economy, optimizing for virality is strategy.

P.S. Huge credit to the builders: Eddy Xu, Caden Li, Andrew Cheng, and Madhav Rapelli. Big thanks to Madhav as well for taking the time to speak with me.

Until next time,
Khai