📰 Key Highlights

A content creator and startup founder named Ben Guez used the open-source AI agent OpenClaw with Claude Code to set up an automated script that mass-posts “trial Reels” on Instagram, racking up over a million views and 200 DMs in just a few days.

Here’s what he actually did: OpenClaw tracks World Cup match results in real time, and after each game ends, it automatically triggers Claude to generate nearly identical videos — footage of Guez sitting on a train, staring out the window, looking bummed out, with subtitles reading “Can’t believe {country of the losing team} lost… if any girls from {that country} need emotional support… my DMs are open.” Only the country names change; everything else is a complete template reuse.

Since Instagram’s “trial Reels” don’t show up on the creator’s public profile, he posted the same content over a dozen times and nobody outside could tell. The key kicker: Guez clearly states on his personal profile that he only replies to DMs through his AI language-learning app “Canary,” effectively forcing anyone interested to download his product first — turning traffic directly into app installs.

Jeff Weisbein, founder of a South Florida tech PR firm, also used OpenClaw to research restaurants and dating spots in various neighborhoods, letting the AI compile everything into a document with links, saving him the hassle of looking things up manually. Both openly admit to using AI-assisted dating planning. Guez thinks that as long as you disclose it, there’s no issue — but TechCrunch says they can’t independently verify how the women actually responded.

💬 JudyAI Lab Take

The real point of this case isn’t how gimmicky AI-assisted dating is — it’s that one person used an open-source agent plus template-replacement logic to funnel a million views straight into app installs in a matter of days, at near-zero cost.

Breaking it down from an AI builder’s angle, Guez’s moves are actually pretty plain: an external-event trigger (World Cup results), a near-static content template (only country names change), a distribution mechanism that keeps the platform’s algorithm from noticing (trial Reels don’t appear on the profile), and a forced funnel (DMs only answered through the app). These four layers together are what make the actual product — the content itself doesn’t matter; what matters is the “trigger → generate → distribute → convert” pipeline. The trend we’ve been seeing is: once AI agents automate repetitive distribution, a creator’s core competitive edge shifts from output volume to the ability to design this pipeline.

If you have a product that requires installation before people can experience it, it’s worth asking: which external event could become your automated trigger?

📅 Source Info

🔗 Further Reading