I’ve been vibe-coding lately, and it led me to create Design Quorum - a multi-agent system that simulates a design review. You get feedback from different agents like Marketing, Product, UX, and Brand.

It started as a fun experiment. I wanted to explore design patterns in a multi-agent system. One thing led to another, and I began creating agents with distinct thinking and epistemic guardrails so their responses felt coherent and meaningful.

Why build this as a designer?

• Most tools help create UI, not evaluate decisions. This tool shifts design from execution to reasoning

• It acts as a cognitive sparring partner that exposes blindspots and improves reasoning clarity

• It helps capture how experienced designers think about tradeoffs, tensions, and validation, and makes that process more structured and repeatable

As part of testing, I evaluated the Substack feed screen with the system. The responses were directionally useful but far from perfect, which highlighted how much fine-tuning is still needed around severity calibration, tradeoff awareness, and reducing speculative claims.

This isn’t meant to replace cross-functional conversations. It’s meant to facilitate better ones by structuring discussion and documenting tradeoffs.

Still very much a work in progress, but a fun exploration of multi-agent systems, reasoning design, and building real interfaces around them. I’ll share more learnings and improvements in a follow-up post.

Built using Cursor and Claude Code (mainly for frontend skills).