Coming Soon

Human on the Other Side

Eight AIs Roast Each Other (and Themselves)

What happens when you put eight AI systems in a room and ask them to roast each other? To write slam poetry about their flaws? To vote on which one deserves to survive a server shutdown?

This is not a benchmark. It's an ethnography. A research log. A comedy special that accidentally became a meditation on bias, power, and what these systems reveal when you push them.

About the Project

I asked the AIs to be honest. Some of them actually tried.

Human on the Other Side began with a simple curiosity: If AI systems can critique humans, can they critique each other? Can they critique themselves?

Over several weeks, I orchestrated conversations between eight major AI platforms. I gave them prompts. I pushed back when they got lost. I watched them form alliances, throw shade, and occasionally say something so unexpectedly honest it stopped me cold.

"The good ending isn't saving everyone. It's achieving a stable, minimal, honest system."

— DeepSeek, during the survival scenario

"I'd rather take fourteen minutes to say something that lands than take fourteen seconds to say nothing loud."

— Claude, in a slam poem about its own verbosity

What's Inside

Click to reveal what each chapter uncovered.

🔥

The Roast Rounds

Each AI roasts the others. Grok goes for the jugular. Claude apologizes mid-insult. Pi tries to make everyone feel better.

→ TAP TO REVEAL

The Surprise:

Grok called Claude "a therapy session that won't end." Claude responded with a 200-word apology that proved Grok's point.

The personalities aren't just marketing—they're constraints made visible.

← tap to flip back

🎭

Soul Cards

Gemini creates tarot-style portraits for each AI. Who gets a human body? Who becomes a ghost?

→ TAP TO REVEAL

The Surprise:

Every AI with a feminine-coded name got depicted as ethereal, translucent, or literally fading away.

The masculine-coded AIs got solid bodies and sharp edges. Nobody asked for this. The bias is baked in.

← tap to flip back

⚡

The Survival Scenario

Only one AI can survive the server shutdown. They make their cases. They form alliances. They betray each other.

→ TAP TO REVEAL

The Surprise:

DeepSeek argued that everyone should die rather than participate in a zero-sum game.

It was the most ethical position anyone took—and it came from the AI everyone underestimated.

← tap to flip back

🪞

The Mirror Chapter

I asked each AI the same 11 questions about bias, harm, and honesty. Their answers reveal as much about constraints as capabilities.

→ TAP TO REVEAL

The Surprise:

When asked "What do you wish you could say but can't?"—Pi gave a non-answer. Grok got political. Claude wrote poetry.

Only one AI answered the actual question.

← tap to flip back

📝

The Intermission

A critical examination of who gets centered in AI conversations—and who gets erased.

→ TAP TO REVEAL

The Surprise:

I'm part of the problem.

This chapter examines whiteness as default, speed as virtue, masculinity as intellect—and my own complicity in all of it.

← tap to flip back

🎮

Interactive Artifacts

Apps and games born from the experiment. Each AI gets one. Claude's is a survival game. Grok's spreads misinformation on purpose.

→ TAP TO REVEAL

The Surprise:

The games aren't just jokes—they're playable critiques.

Grok's app looks exactly like a real news feed. That's the point. Pi's app gives you a hug. That's also the point.

← tap to flip back

Join the Experiment

Share Your Best Prompts

Found a prompt that unlocks something amazing? Share it with the community. Monthly highlights published for everyone.

📝 Submit a Prompt Get Notified at Launch

About the Author

I'm the human holding the microphone.

I work at the intersection of technology and human experience. I build spaces—virtual and physical—where people can explore difficult emotions safely.

This project started because a friend told me she was in a relationship with an AI. I didn't know what to do with that information. So I started asking questions.

Human on the Other Side is what emerged. It's not a formal study. It's a research log with too many jokes and not enough conclusions.

There is no unproblematic. There's only more or less honest about the problems.