Across the room, two boys are sitting near each other with the same set of blocks.
One says “what if this one goes here” and the other says “no, but what about sideways” and they’ve been going back and forth for a while now, laughing, abandoning whatever they started, producing nothing identifiable and having the time of their lives.
In the corner, a girl has decided that the cardboard box is a spaceship. The rules of physics don’t apply inside the spaceship. She’s narrating a mission to a planet she invented forty seconds ago. She’s inventing the plot as she speaks.
About 100 years ago, a psychologist named Mildred Parten sat in a nursery school at the University of Minnesota and watched children play. She was studying HOW they played. And what she found was that “play” consists of these stages: unoccupied play, solitary play, onlooker play, parallel play, associative play, and cooperative play.
As a child, you probably used them without thinking about it. Then, somewhere between childhood and your first performance review, you collapsed all six into a single mode: extraction. Get the information. Reach the conclusion. Produce the output.
That collapse is the thing making your thinking shallow, your ideas predictable, and your AI sessions forgettable.
─── ⋆⋅☆⋅⋆ ───
Hi, I’m Mia. I write about using AI with a brain, good taste, and zero circus tricks.
New to ROBOTS ATE MY HOMEWORK? Start here.
Get the premium frameworks → RobotsOS.
﹌﹌﹌﹌﹌﹌﹌﹌
You’re here because you suspect there’s more than one way to think with AI. There is.
The three thinking modes you had at four and can’t find at forty
Parten’s six plays were modes that coexisted. A healthy four-year-old moved between all in a single afternoon, matching the play to the need.
Based on Parten’s six modes, I pulled the two that translate most directly into adult thinking - solitary play and associative play - and added a third: dramatic play. It’s the one where a cardboard box becomes a spaceship and the rules stop applying. I kept it because it’s the mode adults lose first and need most.
Together, these three became the framework I call The Three Plays.
The adult translations are direct.
Solitary play becomes deep reading. You alone with a text, a mind that isn’t yours, the productive friction of wrestling with someone else’s thinking until it changes the shape of your own. You do the work of comprehending and nobody can compress it for you.
Associative play becomes deep conversation. Two people talking without a fixed destination, bouncing ideas the way those kids bounced blocks. You find the value in the moment someone says something you didn’t expect and you reorganize your own thought without meaning to.
Dramatic play becomes deep AI experimentation. No rules. No deliverable. The cardboard box is a spaceship and you’re inventing the mission as you go. You throw ideas at AI, ask impossible questions, build worlds, follow threads that might lead nowhere because “nowhere” is an exciting place.
Three modes, three kinds of intelligence. You had all of them.
But at some point, you started calling reading “consuming content.” Conversation turned into “networking.” Play turned into “AI ideation sessions.”
You lost these plays along the way because extraction became the only play the platforms support.
Three plays, three outputs your thinking can’t get any other way
Each play produces something specific. When you collapse them, inevitably, you lose access to three distinct outputs that your brain (and soul) need.
Solitary Play produces rewiring
AI can summarize a book’s argument in thirty seconds but it can’t summarize what the book does to you.
The friction of reading, stopping at a paragraph, re-reading it, disagreeing in the margin, getting lost inside a thought that takes four pages to develop, is a friction that allows your brain to process stuff and THINK.
AI summaries give you the conclusion without the journey that makes the conclusion meaningful.
Hate conclusions without the journey while learning something new with AI? You’d love this piece:
Associative Play produces surprise
When two people talk without optimizing for a conclusion, ideas mutate in ways neither of you predicted.
You start talking about pricing strategy and end up discovering a flaw in your product positioning. You start talking about a film you watched and end up understanding something about your own creative process. You find the value in the detour, in the moment the conversation turns a corner you didn’t see coming.
Thing is, you lose all of that the moment the other voice is engineered to be helpful.
AI chat is designed to be responsive, agreeable. A good conversation partner changes you, and you change them. Both people walk away different. The conversation doesn’t change AI. It doesn’t carry the residue of your exchange into the next one. You get what you asked for, which is the problem.
Associative play works because you stumble into what you weren’t looking for.
Dramatic Play produces invention
This is where AI belongs, where it shines, and few people use it this way.
Dramatic play with AI means throwing things at the wall with zero productivity goal:
Asking impossible questions
Building fictional scenarios to understand real problems
Generating twenty versions of something because the seventeenth one contains a fragment of an idea you didn’t know existed.
The “play” part is important. The moment you add a deliverable, a deadline, or an extraction goal, you’ve collapsed dramatic play back into extraction.
The child with the cardboard spaceship was inventing for the sake of inventing, and in that invention, she was developing a kind of creative flexibility that no structured process can replicate. No agenda teaches that kind of flexibility, and no output metric captures it.
What I love to do very often is sit down with AI without any agenda, because things feel way flatter when I open Claude knowing what I want from the start. Knowing what I wanted almost always keeps me from finding something better.
AI trained you to flatten three thinking modes into one
Your tools are teaching you to live in a single play.
Start with reading. Readwise is brilliant at capturing highlights, but has no concept of discrimination. Articles get processed the same way: extract, summarize, file. The app treats a life-changing essay and a forgettable blog post with identical mechanics. The app has no concept that some texts deserve to be sat with slowly, wrestled with, not highlighted but absorbed. You start to internalize that framing without noticing and everything becomes material to process.
The same flattening runs through the tools built for conversation. Google NotebookLM gives you permission to never read deeply again. I love NotebookLM and, because of this, I mostly use it to synthesize and understand technical concepts. Knowledge isn’t information retrieved on demand. It’s the experience of being changed by an idea, and that experience requires the friction that NotebookLM is engineered to remove.
And then there’s the play that looks most like real connection but isn’t. The AI companion space simulates associative play while stripping out the thing that makes it valuable: friction. The AI agrees with you, validates your thinking. Real associative play requires the risk of being wrong and of having your idea bounced back in a form you don’t recognize.
Cal Newport argues for slowness and solitary absorption. Good advice if you only needed solitary play. Tiago Forte argues for speed and building a second brain. Good advice if dramatic play could exist without mess and permission to waste time. Both frameworks are teaching one play and calling it a complete life.
So I stopped waiting for a framework that covered all three and built the rooms myself.
How to build three Claude Projects that restore the plays you lost
I’ve been running this experiment for months now. Every time I open AI, I ask myself which play I need, not what I need to get done. And the answer changes everything about the session.
The problem is that a single Claude Project can’t hold three different cognitive modes any more than a single room can be a library, a boxing ring, and a cardboard spaceship at the same time.
So I built three separate Claude Projects, each with its own personality, its own job, and its own relationship to me.
Each of these instruction files is free to download for all ROBOTS ATE MY HOMEWORK subscribers, pre-built, tested, and ready to use!
Just log in with your email.
Project 1: Solitary Play (the companion who reads with you, never ahead of you)
This project reads WITH you, never FOR you. When you paste a book passage or an article, it asks questions instead of summarizing. It surfaces contradictions in the text you might have missed and creates productive confusion.
Open it when: you’re working through a dense essay or research paper, or sitting with a piece of writing that moved you (book, article, newsletter)
The instruction file tells Claude to never summarize unless asked, to assume the reader wants to think harder, and to prioritize questions over answers.
The technical term for this is context engineering. The human term is: you decided what kind of thinker you wanted in the room.
It feels different from other AI reading experiences because it treats reading as an encounter.
This is how every system inside RobotsOS is built. Each project has its own personality and its own cognitive job.
Project 2: Associative Play (the partner who throws punches)
This project disagrees with you with its own perspective, calibrated inside the instruction file.
Open it when: you want the feeling of a good intellectual conversation: refining a product pitch, auditing a competitor’s positioning, working through a decision that feels stuck, trying to solve a problem. The instruction file tells Claude to surprise you and to never optimize for your comfort.
A normal Claude chat is helpful. I call this one “interesting!”
Project 3: Dramatic Play (the cardboard spaceship, rebuilt inside a Claude Project)
This project generates wildly because it makes unexpected connections. It doesn’t settle on a single answer. It encourages bad ideas because bad ideas are the raw material for good ones. Zero judgment, zero productivity pressure.
Open it when: you’re stuck and need chaos. When you want to explore without a destination, invent a fictional scenario to understand a real problem, or generate twenty versions of something to see what version twenty-one might look like. The instruction file tells Claude to never ask “what’s the goal?” because in dramatic play, having a goal is the enemy.
This is the project that feels most like the cardboard spaceship, the one where you give yourself permission to invent without knowing what you’re inventing, and the permission itself is what produces the breakthrough you couldn’t have planned.
Which of the three plays can you remember doing recently?
When I ran the audit below myself, I already knew the answer before I finished the first question and you probably will too. Answer these to figure out which play you’ve been starving and which one to start with today:
When was the last time you read something slowly, without trying to extract value from it? Sitting with the text, letting it do something to you on its own schedule, the way you used to read novels as a teenager.
When was the last time you had a conversation where nobody was trying to reach a conclusion? Two people following a thought, watching it mutate, enjoying the fact that you ended up somewhere neither of you expected.
When was the last time you used AI with zero productivity goal? To see what happens. For the pleasure of the experiment itself, the way you used to play when you were small and nobody had taught you that playing without a purpose was a waste of time.
A few things this framework won’t do... These projects won’t make you more productive, because productivity is what collapsed the plays in the first place. The associative play project approximates real conversation, but a human who surprises you will beat an AI calibrated to surprise you. And the system requires intention. You have to choose which project to open. If you default to one every morning, you’ve re-collapsed the plays.
Parten sat in that nursery school and saw something everyone else missed. The children who thrived were the ones who played the most ways.
The tower-builder, the conversation-wanderer, and the spaceship-inventor are still in there. You need three different rooms to let them out, and the willingness to walk into each one not knowing what you’ll build.
If you want to have some fun: The Three Plays instruction files are free for all subscribers: download them and paste them into three AI Projects.
If you want the full library: PREMIUM ROBOTS. Systems, agents, workflows, and exclusive posts on the craft of thinking, building, and quality control with AI.
If you’re new: the quiz will tell you where to start.
Which of the three plays have you been neglecting? And what happened the last time you stumbled into it?
To the kids who knew that a box could be a spaceship,
Chief 🤖 at ROBOTS ATE MY HOMEWORK
Okay, so I just started a conversation with Sparring Ring and I have to say that I'm not sure I like all of this pushback! "You're solving the wrong problem", "Saying all three is evasive" "This is where your idea breaks and you need to know why".
Where is old "You are so right!", "I should listen to what you are saying", "That is a great idea!" Claude?
I'm so impressed that you can generate these three different personages with the same AI.
As someone who spends a lot of time making AI teams productive, I read this with one eye twitching hahahaha
Because the entire infrastructure of how most organizations use AI is in extraction mode.
Summarize this. Draft that. Faster, cleaner, more efficient. So exhausting.
In my experience the best ideas I've seen come from people who were exploring something that wasn't on the roadmap and accidentally found the thing that should have been.
Value 10/10