Knowing three categories beats memorizing ten names. Today we unpack this principle through the story of finding free AI models on OpenRouter.
Most people entering a new field do the same thing. They start memorizing names. "What was that tool called, what was that company called." Twenty names pile up in a notebook, and strangely, as time passes you still don't know what to do. Why? Simple. You skipped the categories.
This essay shows you a faster way. Even if you've never tried this approach, you can follow along. Today's example is finding AI models, but the principle works for any new field — travel, cooking, investing, any unfamiliar world.
You've had this experience. "AI is hot these days" and you open the search bar. Ten names cascade. ChatGPT, Claude, Gemini, Llama, DeepSeek, Mistral, Qwen, Perplexity, Cursor, Copilot. Memorizing them makes your head hurt. Even after memorizing, "so what do I use?" has no answer.
Names only matter to someone who already knows the terrain. Memorizing ten tree names without knowing the structure of the forest is like writing "oak, pine, ginkgo" in the air. The words pile up. The forest doesn't appear.
Flip the order. First grasp "this forest has conifers and broadleaf trees." Two categories. Now each name you learn locks into a slot — "ah, this one is a conifer." The same name only sticks when it lands on a category frame.
Let's get concrete. There's a service called OpenRouter. As of April 2026, it hosts 28 free AI models. "Wow, 28!" and you try them one by one — weekend gone.
But pause and ask something first. What kind of thing is OpenRouter? Answering this requires a category. OpenRouter doesn't make AI. It aggregates models from many companies in one place. This kind of thing has a name — a broker.
This one word matters. Grasp "broker" as a category and the AI terrain suddenly becomes legible.
These three categories cover the whole AI world. Any name you hear, 30 seconds and you know which of the three it belongs to. That's what "seeing the terrain" actually looks like.
Think about landing in a city you've never been to. You need a place to sleep. How do you find one?
First way — memorize ten hotel names in advance. Hilton, Marriott, Shilla, Hyatt. But that city might not have those hotels. Or they might be two hours from the airport. Your ten memorized names are wasted.
Second way — learn the lodging categories first. Hotel, motel, B&B, guesthouse, hostel, Airbnb. Six categories. Just know how each differs in price and feel. Then when you land and open a search, you narrow instantly — "I'm alone and low budget, so guesthouse." You find the name then.
Same 30 minutes, very different outcomes. The category-first person adapts in any city. The name-memorizer starts over every time the city changes.
The day after I grasped "broker" as a category, I found OpenRouter in three minutes. Not three minutes of search — a day of reading made those three minutes possible.
Here's the key sentence.
You cannot search for something whose existence you don't know about.
The search bar finds what you know. It can't find what you don't. If you don't know "brokers" are a thing, no matter how many times OpenRouter appears in Google, you can't search for it. You might search "cheap AI models" and OpenRouter does show up in the results, but without the eye to recognize "this is what I was looking for" — it slides past.
The most expensive mistake lives here. "I didn't know that kind of thing existed." This is the most frequent regret in the tech space. The tool wasn't missing. The tool was there, but you didn't know that category of tool existed, so you couldn't find it.
How do you actually learn categories? One thing. Entering a new field, find three categories before ten names.
The question is simple — "In this world, what kinds of players are there?" Ask ChatGPT, or ask a person. This one question is enough. Most of the time, the answer clusters into a few categories. Write them down in three lines. Those three lines are your map.
Here's the AI world example again.
| Category | Role | Examples |
|---|---|---|
| Provider | Builds AI | OpenAI, Anthropic |
| Broker | Aggregates providers | OpenRouter |
| Interface | Delivers to user | Claude Code, Cursor |
One table, the whole AI world. Every new name you hear, check which cell it lives in. Cells first, names inside the cells.
Memorizing names first is the wrong order. Grasp categories first, lay names on top. Someone with categories can recover from forgetting a name. Someone with only names starts over when the name goes.
Build one habit — entering a new field, ask "what kinds of players exist here?" first. The two or three categories in the answer are your map.
Three words to carry — Category. First. Name later. Don't flip the order. Today's example was OpenRouter as a broker. In three years, OpenRouter may be gone, some other broker in its place. Fine. Someone who knows "broker" as a category finds the new name in 5 minutes. Technology changes. The principle doesn't.