We often think metaphor is something that explains a system after the fact.
Now I think metaphor is often how you know what you are building.
This has been especially true with AI because AI systems get abstract very quickly.
You start with, “I need help managing recurring work.”
Three hours later, you are still wrestling with what that means, trying to identify recurring work in your own life for realsies this time, and maybe even falling into a therapy session with chat about why you are like this.
A good metaphor brings the humans back.
It is an anchor for what is happening and the dynamics underneath it.
It tells people:
- What is this?
- What belongs inside it?
- How does it move?
- What does healthy look like?
- Where does the human sit?
That is not fluff. That is usability.
The right metaphor becomes an interface.
Not because people are simple. Because complexity needs handles.
We are living in wildly complex times on a steep learning curve. I have been trying to keep up too. I am not an engineer by trade, and I have a wildly creative mind. The structure and systems thinking of AI is very necessary to my functioning well, but conversely, it is one of the hardest things for me to set up myself.
One day I was in my office, staring at the shelf where I keep a few nostalgic things, and my eyes landed on my collection of My Little Ponies.
Each pony is a unique color combination, and each “cutie mark” denotes their gift or identity.
What. A. Perfect. Analogy. For. An. Agent.
And each stable the ponies live in? Their main workstream.
This works really well for creating a visual map of what you are building and what is going on. If you cannot come up with one cutie mark for an agent, it has too many tasks.
Also, I am not sure if you have ever seen Friendship Is Magic, but those ponies are helpful.
That is how the Stable System started.
Each stable is a workstream, and eventually I created a central committee stable that coordinates the data. I have downloads and examples on this site.
It does not have to be ponies.
But having names and very clear tasks — “cutie marks” — is key to really seeing your AI infrastructure as you build it.
