The secret of great explainers is not knowledge but their metaphor library. Today we unpack this through explaining AI models as car engines.
Let me share the secret of people who explain well. You'd think it's the people who know the most. Actually, no. Explanation quality scales with the number of metaphors you carry, not with the amount you know.
Today's example is explaining AI models as car engines. But the principle has nothing to do with AI. Three years from now, even if "AI" isn't the word we use, the skill of making the hard easy still applies.
When you "understand" a new concept, what does your brain actually do? Simple: it attaches new information to something you already know. Something completely new is unreachable. It has to bond with prior knowledge. That bond is the metaphor.
When I didn't know this, I explained with jargon. "Claude Opus is a 175-billion-parameter large language model." Faces stiffened. Every word was new, and understanding each new word required yet more new words. I call this the vacuum of understanding — nowhere to stand.
Swap it for this:
"Claude Opus is like a big car engine. Powerful but thirsty. Haiku is a small engine. Fast and fuel-efficient, but it lugs on hills."
Same content. The listener gets it. Car engines are already in their head. New information (AI model) stuck to prior knowledge (engines).
Another example. Say you need to explain OpenRouter. Technically: "an API gateway service unifying multiple LLM APIs into a single endpoint." Four new words in one sentence. Three questions before the listener understands.
With a metaphor:
"OpenRouter is like a broker. Just like a delivery app lets you order from many restaurants in one place, OpenRouter lets you use many AI companies' models in one place."
Ten seconds to understand. Delivery apps are universal. The new info (OpenRouter) attached to known knowledge (delivery app) and found its seat.
Understanding arrives when new information attaches to something already known.
So what's the difference between an expert and an expert-who-explains-well? Not knowledge volume. It's the size of the metaphor library.
Thirty-year experts who can't explain share a pattern: they only use their own world's vocabulary. They've never practiced translating to another world. Great explainers, in contrast, carry metaphors from cooking, architecture, exercise, music, gardening. They read the listener and pull the right metaphor from the right domain.
How do you grow the library? Practice. One drill:
Each day, take one concept from your expertise and explain it in a completely different domain's language.
It'll feel awkward. Stretched metaphors will come out. That's okay. Stacked practice builds a metaphor dictionary in your head. Next time someone asks "what is this?", a word from their world pops out of your dictionary.
I personally carry five metaphors for AI models — car engines, head chefs, doctors, company orgs, train engines. Chef for chefs, medical for doctors, org for office workers. The same Claude Opus wears five different faces.
A concrete scenario. Tuesday morning, you have three minutes to explain to a non-technical executive why the system got slow. Saying "the DB query is missing an index so it's doing a full scan" freezes the face. Instead, try this:
"Imagine a library with a million books and no catalog cards. The librarian has to flip through every book. With an index, they find it in three seconds. Our system right now is the library without a catalog."
Same fact, but one is a wall and the other is a door. Before the meeting ends, the executive starts asking "how much does the catalog cost?" A single metaphor brings the budget approval.
A caution. Metaphor is a tool for crossing the first threshold of understanding, not a replacement for the real thing. Once you understand AI model selection via the car engine metaphor, you have to move on to the actual numbers — cost per token, context length, response speed. Staying at the metaphor means surface understanding only.
I've seen misuse. One person kept insisting "AI is an intern." Fine at first. Easy to grasp. But months later, that person was only assigning intern-level work to AI. The metaphor had become the ceiling of capability. Metaphor is a foothold, not a roof. Once up, jump to the next metaphor.
A simple drill you can do right now. I call it 3-2-1.
Five minutes a day, two weeks. You'll stack 14 new entries in your metaphor library. Three months: 90. You'll catch up in three months to what great explainers accumulate over a lifetime.
Explanation quality comes not from knowledge but from metaphor count. Stop stacking jargon. Stack everyday metaphors instead. That's a more valuable asset than knowledge itself.
Three words to carry — Metaphor. Connect. Understand. One metaphor connects new information to prior knowledge, and that connection creates understanding.
This isn't only about AI explanation. Teaching, writing, persuading a business partner, explaining something to a child — all the same. The skill of making the hard easy is a lasting asset. Technology changes. The principle of understanding doesn't.