DesignAgents

A few weeks ago I read Patterns for Building AI Agents by Sam Bhagwat.

It made me rethink how we use AI in creative work.

Most tools today feel the same. You write a prompt, you get a result.
Sometimes good, often generic.

But real design doesn’t work like that.

In practice, it’s collaborative.
There’s someone thinking about visuals, someone about copy, someone about layout. Ideas evolve through interaction.

So I started exploring a simple question:

What if AI worked more like a design team instead of a single assistant?

I built a small experiment around this.

A canvas where different AI agents take on roles.
A “creative director” assigns tasks.
Other agents handle design, copy, layout.

What I found interesting wasn’t just the output, but the process.

Instead of forcing everything through one prompt, decisions are distributed.
The results sometimes feel more structured, sometimes more surprising.

Not always better. But different in a meaningful way.

Another unexpected insight:

If agents can create designs, then the way they collaborate becomes valuable.

So I added the ability to share not just designs, but the agent setups behind them.
Your “AI team” becomes something others can reuse and improve.

I also didn’t want to define everything myself.

Users can create their own agents, define roles, tools, and behavior.
Some setups stay simple. Others evolve into complex workflows.

Still early. Still learning where this approach works and where it doesn’t.

But it changed how I think about AI:

Maybe the real shift is not better prompts.
Maybe it’s moving from tools to systems.

Curious how others think about this.