The Wild 13

Michael Ende — author of *The Neverending Story* — writes in *Jim Knopf and the Wild 13* that the thirteen Wild 13 are the most dangerous characters in the world. They ride dragons. They shout. They arrive in formation, and your first instinct is to run. Lummerland nearly fell to them.

And then — slowly, page by page — you realise that is not the whole story. The Wild 13 are wild because no one has shown them any better. They are dangerous because no one has yet told them whom they could be fighting for. When Jim and Lukas give them that — a direction, a side, a person they can trust — they become the most loyal crew in the entire story. No less wild. But the wildness now serves someone.

That, in brief, is my experience with AI agents.

When I first began deploying agents, I had the headlines in my head: black box, uncontrollable, riding dragons through your processes, and if you are not paying attention your output is wrong or your client is confused. That picture was not entirely false. Uncalibrated agents without briefing discipline and without clear role separation are exactly that: the Wild 13 in the unflattering sense. They do a great deal — but not for you.

The question is not whether you tame agents. Taming is the wrong word — it implies suppression, and that is not the goal. The question is whether you can give them a direction. Whether you can understand what they can and cannot do. Whether you are prepared to be the person who decides.

That is the difference.

Over the past several months I have watched a handful of agents — named, with defined roles, with briefing discipline, and with the clear rule that a human always decides — begin to behave like a crew. Not perfectly. Not yet, by a long way. But recognisably coherent. The Designer asks about the user before answering. The Artist reads everything before writing. The Engineer checks before building. And when something goes wrong — and things occasionally go wrong — they say so.

This is not magic. It is briefing hygiene and a clear decision structure. But the result feels like something Ende described: a group that looks wild from the outside and that, once you understand the logic, develops a remarkable reliability.

The pattern has one condition, and it must be taken seriously.

In *Jim Knopf*, the Wild 13 do not fight for just anyone. They fight for Jim, because Jim is the person who is *there* — who listens, who has a direction, who does not run away. The agents I know do not fight independently for anything. They need a person who knows what they want. A person who makes the decisions that *are* the decisions — and does not merely nod while the agent charges ahead.

When the person is absent, or when they quietly hand the decision to the agent, the wildness returns. Then the dragons ride in circles again. That is not a flaw in the agent. It is a flaw in the structure.

A human always decides. At Apuna, that is not a phrase. It is the foundation on which everything else stands.

Ende sends the Wild 13 out from Lummerland at the end. They are needed — for a task they were built for, one that demands their wildness. And they return when Jim and Lukas call.

That is the image I carry when someone asks how we work with agents.

Not automation for its own sake. Not agents working away in silence while someone hopes the result will turn out fine. A crew equipped with a direction, with a role, with the knowledge that a person is waiting at the end who will review the result and decide. Wild, fast, capable — and reliably loyal to that one person.

The Wild 13, once you understand them, are not a threat. They are the best possible reason not to work alone.

*Michael Ende, "Jim Knopf und die Wilde 13", Thienemann Verlag, 1962.*