The Magic Word: What Ursula Le Guin Taught Me About AI

I said one word to my AI last week. “Pipeline.”

No instructions. No explanation. No requirements document. Just the word.

It came back with a full analysis of my sales pipeline. Stockpot status, warm prospects ranked by engagement, outreach gaps, follow-up priorities, stale contacts that needed re-engagement. Connected systems I hadn’t mentioned. Surfaced problems I hadn’t asked about.

A week before that, I said “Domains.” One word. It ran the domain portfolio, cross-referenced against templates I’d already built, flagged expiring registrations, evaluated new drops against my acquisition criteria, and told me which ones matched existing revenue paths.

The Wizard of Earthsea

Ursula Le Guin understood this fifty years before anyone was prompting an AI.

In her Earthsea novels, magic works through knowing the true name of a thing. Not a formula you memorise. Not a technique you learn from a course. You study the thing itself - the wind, the sea, the stone - until you understand it deeply enough to know its name. Then you speak the name, and the thing responds.

The wizard who knows the true name of the wind can call it with a word. The wizard who doesn’t can recite every spell in the library and nothing happens.

Ged, the protagonist, doesn’t become powerful by learning more spells. He becomes powerful by understanding more deeply. The incantations get shorter as his knowledge grows. By the end of the series he barely speaks at all. The understanding carries the weight that words used to carry.

Incluing

There’s a word for what Le Guin does, and it’s the key to understanding why this works.

Jo Walton named it incluing. It’s the science fiction worldbuilding technique where instead of stopping the story to explain the world, you embed the world in the action. The reader picks up how things work from context, without ever being told directly.

Le Guin was the master. In The Ones Who Walk Away from Omelas, she builds an entire city through accumulating detail. She never stops to explain the social structure or the economy or the moral framework. The reader assembles it from what’s described. And by the time you reach the reveal - the child in the basement, the cost of the city’s happiness - you understand the world well enough to feel the weight of the choice.

The people who walk away from Omelas don’t explain why. They don’t write a manifesto. They just leave. And the reader understands, because Le Guin built the world thoroughly enough that the silence carries the meaning.

When I say “Pipeline” to my AI, I’m not being lazy. I’m incluing. The word works because the world has been built. My AI knows what pipeline means in this business - not a generic sales funnel, but the specific seven-stage flow from discovery to conversion, with the stockpot system, the drip sequences, the heat scoring, the audit emails. All of that is loaded context. The word “Pipeline” is the incantation. The context is the magic.

The Reliable World

Gene Wolfe does something different from Le Guin, and it matters here.

In The Book of the New Sun, Wolfe gives you an unreliable narrator inside a world that is completely reliable. Severian tells you what happened, but his understanding of events is partial, sometimes self-serving, sometimes simply wrong. The world’s physics are consistent even when the narrator’s understanding of them isn’t.

The reader who trusts Severian at face value gets one story. The reader who notices the gaps between what Severian says and what the world shows gets a richer, more accurate story. The truth is in the infrastructure, not in the narration.

My wiki, my codebase, my git history, my client documents - that’s the reliable world. The AI operating within it has a partial view, just like Severian. But the more of the world I make visible to it - the more context I load, the more patterns I establish, the more corrections I encode - the closer the narrator’s account gets to the truth.

An AI without context is Severian at his most unreliable. Confident, fluent, telling you things that sound right but aren’t. An AI with deep context is Severian at his most perceptive - still imperfect, but grounded in a world that holds the facts steady even when the narrator wobbles.

The Walk

There’s a deeper thread running through Le Guin that matters here, and it’s not about technique. It’s about choice.

The ones who walk away from Omelas walk away because they’ve seen the child. Not because someone explained the moral philosophy. Not because they read a manifesto about justice. Because the world was built thoroughly enough that seeing it produced a response that was beyond argument.

I walked away from WordPress. From complex stacks. From the agency model. From the way everyone else builds websites and runs consultancies. Not because I read a business book that told me to. Because I built a system, lived inside it, and saw what was possible when the context was deep enough. Once you’ve seen what a single word can produce when the infrastructure is rich, going back to writing requirements documents feels like going back to Omelas and pretending you didn’t see the basement.

That’s the thing about ingeniculture. It’s not a technique you can evaluate from the outside. It’s a practice you have to live inside before the results become self-evident. The people who try it for a weekend and declare it doesn’t work are the ones who stayed in Omelas. They saw the brochure but never met the child.

The Practical Discovery

I’ve been building context around my AI for seven months. The first month, every prompt was a paragraph. Detailed instructions, specific requirements, careful guardrails. The AI needed all of it because the context was thin.

By month three, the prompts were getting shorter. Not because I was being efficient. Because the documents had grown, the patterns were established, and the AI had enough world to work from.

By month seven, a single word produces a thorough examination of any part of the business. Not a summary. Not a generic overview. A specific, connected, actionable analysis that draws on patterns, history, client data, and methodology I never mentioned in the prompt.

This is incluing in practice. The worldbuilding is done. The story can move at the speed of the world’s own logic, without stopping to explain itself.

“Pipeline” produces a pipeline review because the AI knows my seven-stage sales process, my stockpot system, my heat scoring methodology, my outreach cadence, my client capacity, and my conversion patterns. It doesn’t need me to specify which pipeline, which metrics, which format. The context carries all of it.

“Domains” produces a portfolio analysis because the AI knows my acquisition criteria, my template inventory, my pricing model, my registration strategy, and my flip-or-build decision framework. One word sits at the intersection of six systems. The response draws from all of them.

“Monday Service” triggers a complete client service workflow - briefings loaded, emails synced, rankings checked, work summaries prepared - because the methodology is documented, the client data is loaded, and the service rhythm is encoded in the infrastructure.

The less I say, the more the system reveals. That’s not minimalism for its own sake. It’s Le Guin’s principle: when the world is rich enough, the story doesn’t need exposition.

Why This Can’t Be Copied

Someone could read this article, download the same AI, and type “Pipeline.” They’d get a generic overview of pipeline management concepts. Something from a textbook. Plausible, confident, useless. The tool is the same. The situation isn’t.

The word is the same. The world behind it isn’t.

And here’s the Gene Wolfe part: even I’m an unreliable narrator of my own system. I can tell you how it works. I can describe the infrastructure. But the world itself - the wiki, the codebase, the git history, the thousands of corrections - that’s more reliable than my description of it. The truth is in the infrastructure. My account of it is partial.

Which means the best I can do is point you toward building your own. Not copying mine. Building yours. Your business, your context, your corrections, your world.

The Magic Word

I call this principle The Magic Word. Not because it’s mystical. Because Le Guin’s magic is the best metaphor I’ve found for what actually happens.

Find the word at the intersection of the most systems. Speak it into loaded context. If sparks fly - if the system produces dense, connected, actionable output from a minimal prompt - the context is deep. If nothing happens, or the output is generic, the infrastructure needs building.

The silence after a magic word is diagnostic. It tells you exactly where the world is thin.

The word proves the world. Not the wizard.

And that’s what Le Guin, Wolfe, and seven months of daily practice taught me. The power was never in the incantation. It was in the years of study that made the incantation unnecessary.

The model is a commodity. The world you build around it is the edge.

One word. The infrastructure does the rest.

Ursula K. Le Guin, A Wizard of Earthsea. Gene Wolfe, The Book of the New Sun. Jo Walton coined “incluing.” The infrastructure is plain text and discipline. The magic is real.