ai. working with multiple agents

One of the interesting things I have been learning through experimentation with different AI image systems is that the differences between them become visible surprisingly quickly. While I am convinced that most modern AI agents are capable of producing very similar results when given the right prompts and sufficient guidance, they do not all arrive at those results in the same way.
Some agents seem naturally inclined towards certain visual languages. Others require far more direction before they begin to understand the destination. During my recent experiments, I found that reproducing the Kandinsky 1 visual language with another AI was far more difficult than expected. The agent was capable, but without the communication protocols and visual understanding that had been developed over time, it struggled to reach the same outcome.
This led me to an interesting simply do not have the time, nor the desire, to become an expert in the strengths, weaknesses, quirks and prompting requirements of every available AI image system. The number of agents is growing rapidly, and each comes with its own preferred methods of communication and image construction.
What I discovered instead was a workflow that feels remarkably effective.
Rather than expecting a single AI to perform every task perfectly, I can allow one agent to create the foundation of an image in the style it naturally prefers. Once the core concept exists, that image can then be transferred to a master agent that understands my personal artistic language and visual preferences. The master agent is then used to reinterpret, refine and complete the work.
The process feels surprisingly similar to a traditional creative studio. One artist produces sketches, another develops colour studies, and a lead artist brings everything together into the final piece.
The recent compass artwork demonstrated this perfectly. The original image contained strong symbolism, good composition, and an impressive painterly quality. However, it remained largely illustrative in nature. Once transferred and reworked using the Kandinsky 1 principles of fragmented cohesion, colour flow, emergent structure and interconnected visual rhythm, the image transformed into something that felt much closer to my own artistic interests.
The image was no longer simply explaining an idea.
It was expressing one.
Perhaps the most important lesson from this experiment is that the future may not belong to a single AI agent. Instead, it may belong to artists who understand how to combine the strengths of many different systems into a unified workflow.
The challenge is no longer finding the perfect AI.
The challenge is learning which AI should perform which part of the journey.

Popular posts from this blog

A Quiet Return It’s been a few years since anything new appeared on this blog. No dramatic exit, no crisis, no grand story — just life unfolding in its own quiet way while I wandered off into a long stretch of creative exploration. I didn’t spend that time chasing expectations or trying to produce anything impressive. In fact, it was the opposite. I let myself make things without pressure, without anticipation, and without any need for an audience. It was a good kind of absence — the sort that gives you space to breathe and rediscover what you enjoy. Along the way, painting slipped out of my hands. Not abandoned in frustration, just… released. It had carried me through many chapters, but eventually it felt like a language I’d said enough in. So I turned to sculpture, almost by accident, and found myself on a long, meandering tour through every material within reach. Papier‑mâché came first — messy, forgiving, and strangely liberating.Foam followed, stubborn but full of possibility.Cardboard had its moment too, proving far more architectural than I expected. Each material taught me something different about form, weight, and patience. And then, gradually, I settled into polymer clay. It felt right — structured but flexible, expressive without demanding perfection. It opened a new path, and with it came another journey: exploring genre, tone, and character in ways paint never quite allowed. So yes, I’ve been away. But not idle.Just quietly occupied in the kind of creative work that doesn’t need announcing. Now the blog returns — not as a relaunch, not as a promise, but simply as a continuation. A place to share what I’ve been shaping, learning, and enjoying along the way. If you’ve found your way back here too, welcome. Let’s see where this next chapter leads.

Polymer Clay 01

grounding