The Problem

Product teams optimize for the average user. Edge cases get discovered in production — by the users who suffer through them. Traditional persona work is helpful but tends toward sanitized archetypes that don’t challenge assumptions hard enough.

The Approach

Create a panel of AI agents, each embodying an extreme but plausible user persona. Not average users — the ones who break your assumptions. Run your product decisions past them and watch what falls apart.

What I Built

A configurable agent system where each persona has:

• A backstory — Not demographics, but motivations and constraints
• A worldview — How they evaluate products and make decisions
• A communication style — How they’d actually complain or praise
• Specific triggers — What makes them abandon a product vs. evangelize it

The agents debate each other, argue with your assumptions, and surface failure modes your team hasn’t considered.

Tech Stack

• Core: Claude API with structured persona definitions
• Orchestration: Multi-agent conversation framework
• Interface: Claude Code skill with configurable persona sets

How It Fits

Edge-Case Society is the product validation layer. Before investing in a feature, run it past the Society. The agents don’t replace real user research — they prepare you for it by surfacing the questions you should be asking.