AI Skills Through Play

2026-04-14

~540 Words | ~2min Read

The more things change, the more they stay the same. And the age of AI is no different. Software engineers learned new languages, frameworks, and techniques the same way for decades. They built a toy project. Want to learn Python? Write a small app. Want to understand design patterns? Implement them yourself. It’s how we’ve always learned. So when I wanted to understand AI “Skills”, I decided to play a game.

I’m fond of solo journaling RPGs. The whole mechanic is simple: roll some dice to get a prompt, then write about it. That’s easy enough to explain to an AI, so I wondered if I could teach the AI to facilitate the game. Not play it for me, mind you! But teach it to handle the drudgery. In this case: rolling the dice, looking up results in tables, and managing game state. That way I could focus on the part I actually wanted: writing the story!

Good news: it worked. I found you need something that’s almost entirely not work to really get a sense of what these tools are capable of. That said, it did take some work. AI is great at transformation verbs. This time it was turning a description of a process (the game rules) to the sequential execution. In short, I had to transform the written rules, into a workflow. And the context and tables into skills the AI would use while following the workflow. The AI invokes Skills only when needed. This keeps the context window clean while providing deep expertise on demand. For example “How do I find out which scenario to prompt the user with?” That skill loads only when I need it, not while I’m writing my response to the prompt. And the AI doesn’t invoke it while trying to look up the setting from a different results table.

We learn more by doing than by theorizing. At some point, you have to bite the bullet and just go. But there’s something wonderful about learning through play! You’re freeing yourself from the expectation of “I’m a professional.” You’re acknowledging “I don’t know,” and giving yourself permission to tinker. It takes time, yes, so you have to balance that. But that freedom to experiment is where the learning happens.

The new core skill in the Age of AI isn’t coding or analysis. It’s clearly articulating your intent! With AI skills specifically, there are two parts. You have to articulate when that skill applies, so the AI can find the skill at the right time. And you have to describe how to do that skill well, or else the skill doesn’t help the AI do the work consistently! AI excels at transformation tasks. The trick is learning to structure the context. You have to know when to load what information. Then design interactions that remove drudgery while preserving what matters.

A solo RPG is certainly an unusual approach. But don’t be afraid to literally play around with the new tools to learn them. Or consider starting with skills that others have built at skills.sh, then refine them for yourself. The pattern that’s worked for decades still works. You just have to give yourself permission to play.

The game I used: The Wandering Tea Garden