There is a game Paul Sheldon talks about in Misery that I find fascinating and think it is something we could use with AI.
If you haven't read it (first of all, fix that if you wanna be my friend), the basic setup is this: Stephen King's protagonist, novelist Paul Sheldon, is being held captive by his #1 fan, Annie Wilkes. Annie is unhinged in the most terrifying way possible.
While Paul is forced to find a way to rewrite a dead character back to life, he recalls a game from his childhood called "Can You?"
The question lobbed across the room, over and over, is: "Can you?"
Can you get out of this? Can you find the way through? Can you solve it within the rules? Can you find a new solution we haven’t considered?
I've been thinking about this game in the context of AI tools, and I think it's a useful mental model for how most people are failing to use them well — and why.
Someone opens an AI tool, asks it something, gets a response that isn't quite right, and... closes the tab. Done. "It doesn't work for me." "It couldn't do what I needed."
This is the equivalent of hearing "Can you?" and simply saying “no” and then walking away as if the question was never asked.
The tool pushed back: maybe it said it couldn't do something, or it gave you a mediocre first draft, or it misunderstood what you were asking and you accepted that as the final answer.
Almost nothing an AI tool tells you the first time is the final answer. The first response is the opening move. Your job is to keep playing.
This is the opposite problem, which just might be worse.
The tool produces something that is okay, not wrong, exactly, but it's just not you, and it's not great, but you published it anyway because it was faster (and we all know how much our society values faster, cheaper, better….).
This is the equivalent of being asked "Can you get out of this?" and your lazy answer being "Sure, I'll just walk through the wall" and nobody else playing points out that walls don't work that way.
The parameters of the game matter. The output has to actually be good. It has to sound like you. It has to be accurate. "Good enough" is a choice you're making consciously or unconsciously, and it's worth being honest about which one it is.
There are tasks you do every week (possibly even every day) that you haven't once asked "can I do this with AI?"
Not because you tried and it didn't work. Just because it never occurred to you to ask.
The "Can You?" game, applied to your own workflow, sounds like this:
The impossible scenario isn't actually impossible. You just haven't even tried.
The thing about the "Can You?" game that makes it genuinely useful as a framework is that it forces creative constraint.
You can't break the rules. You have to work within the parameters. That pressure is what produces creative problem-solving.
When you're working with AI tools, the parameters are real:
Most people treat those limitations as walls. The better move is to treat them like the rules of the game → things to work within, not obstacles to surrender to.
When an AI tool tells you it can't do something, the next move should be to ask:
That's the game. You keep playing until you find the way through.
If you're not sure where to start, here are some I find myself asking on a regular basis:
None of these are tricks, hacks, or shortcuts. They're just questions. The kind of questions you'd expect from a Questionologist like me.
The tools are waiting. The question is whether you're willing to keep playing.