We’ve never produced more ideas and we’ve never changed fewer minds.
For a long time, I fell into the trap of designing and creating content that makes the reader feel something familiar. Maybe seen. Understood. Intelligent. Part of the right crowd.
This is not a bad thing - people save and even share these kinds of perspectives with others.
But their thinking never really changed in the process.
Ellen O’Connell Whittet pointed out the strangest contradiction - bookstores are thriving at the same time literacy is declining.
The bookstores are FULL, and we love carrying colorful tote bags that signal a literary life instead of actually living one.
That’s not the same as having your inner life changed by a book.
The same thing is happening to intellectual work online and nowhere is it more visible than in AI culture.
Our feeds are packed with frameworks, predictions, screenshots, a dozen hot takes you scroll past before breakfast. It feels like keeping up and makes us look curious, informed; problem is, a lot of what we create doesn’t help anyone understand more, think better, or do anything differently, really.
It simply makes us look close to the future whether we understand it or not.
We've become extraordinarily good at sounding insightful and AI made that easier than ever - now it’s often hard to tell the difference between someone who can package an idea well and someone who has actually thought it through.
I desperately wanted to find a way to tell the difference between the two: work that only makes people feel thoughtful and work that changes how they think, so I can also make sure it’s incorporated into my work too.
That led me to a framework, a very uncomfortable experiment on one of my own articles, and a simple test you can use before you hit publish yourself.
─── ⋆⋅☆⋅⋆ ───
Hi, I’m Mia. Welcome to ROBOTS ATE MY HOMEWORK. I write about building coherent worlds around your ideas, identity, and business, then teaching AI how to work inside them.
The four kinds of familiar (or how to build a beautiful doorway to nowhere)
“Familiar” work vs “different” work.
Familiar work helps the reader enter, but different work changes what happens when they’re inside.
And there are four versions of each. Let’s start with the familiar side:
Identity Confirmation is what I like to call the fanciest tote bag. The reader comes across a piece of information, feels validated, and most importantly - feels seen. “I’m thoughtful. I get it.” If that’s all the piece does, they walk away feeling smart and doing nothing. The like and share buttons feel ECSTATIC.
Recognition is a real doorway, which means that the piece accurately describes the reader’s situation. They lean in, “yes, that’s exactly what happens to me.” It’s doing its job if it opens into something.
Aesthetic Belonging is the strongest signal to the tribe. The reader feels compelled to share the piece to say “This is who I am”, rather than, “This changed something for me”. That’s trouble - belonging starts to look like transformation.
And then there’s Resonance. We come across resonance when a piece puts words to something we’ve felt before but never managed to explain. It does create trust, but trust is still a doorway instead of a destination.
It’s the most sophisticated tote bag of all because it feels so much like depth. The reader thinks, “This person understands me.” And perhaps they do.
AI is exceptionally good at familiar work. It generates recognition and resonance by drawing on patterns of language and ideas that audiences already respond to.
In other words, the doorways are almost effortless to produce.
The rooms are not.
So if familiar work gets readers through the door, what should be waiting for them on the other side?
The four kinds of different (or what’s on the other side of the doorway)
Different work moves something around in the reader’s head.
It can happen in four ways.
This kind of work feels rarer now, partly because familiar work has become so easy to create. We can produce things that sound true and feel relevant.
A Reframe takes the facts the reader already knows and adds a bit of a twist, just enough for the whole picture to change. However… nothing has changed - not exactly - and yet the reader can no longer look at it in the same exact way as before.
Permission is the silent one. The reader already had a feeling that something was true, maybe they’ve been circling it for months. Then someone finally says it plainly and they get the permission to act on it.
The story moves forward and now they get a Protocol, which gives the reader something they can do, a real next step they can use tomorrow morning.
And then there’s a New Distinction.
You and I both now see that two things we’ve been treating as the same thing are NOT the same at all.
The strongest pieces often use recognition to bring the reader in, a new distinction to change what they see, and a protocol to help them do something new with it.
The tote bag moment is when familiar work replaces different work instead of leading to it.
AI gives us the familiar, the thing that we already know (or assume), which makes different work MORE valuable precisely because it’s harder to produce. Different work still depends on a human mind making a human judgement.
Knowing the framework is one thing. Running it on your own work is another beast.
Can AI rebuild your best idea from a brief?
I don’t mind creating familiar work. This gives readers the tribe feeling we all long for and it’s great for building a community that everyone loves. But I really, really wanted to see whether I was creating “different” work too.
I couldn’t reliably spot the tote bags in my own work at first glance, so I came up with a replacement test.
I chose an article I was proud of, about WATSON - an editorial research agent I built to get past the usual surface-level content research. Then I gave AI the original outline / brief I had and asked it to create the article, section by section.
The question I had was simple: Which parts could AI recover from the brief, and which parts depended on decisions I’ve never written down because they came from experience? Like filling out the [PLACEHOLDERS] that required my personal take.
Claude reconstructed the argument that AI content research tends to be shallow and that most tools return variations of the same ideas. Those observations were already everywhere in AI culture, so, they did sound like my initial thesis.
These intro sections were doing recognition work - they were not bad, or unnecessary. They gave my readers a familiar problem to walk through.
But the test uncovered something interesting: the problem was much easier to reproduce than the mechanism I’d built to solve it.
Things started getting weird when AI reached WATSON’s lateral thinking methodology.
It could imitate the language of creativity, make unexpected connections, explore the unusual domains, and think beyond the obvious. All the right words!
What it couldn't reliably reproduce were all the decisions underneath the words.
Decisions like:
How do you connect two niche topics without the connection feeling mega random?
How do you notice that ideas from different fields share the same underlying structure?
How do you know which thread is worth following vs which one is clever-looking nonsense?
AI kept going back to its generic advice because the useful part of this whole article wasn’t really the idea of “thinking laterally” but the collection of small distinctions I’d learned to make while doing the work itself.
You can probably already guess which parts of your own work would be the easiest to recreate from a plain brief.
Look for observations your audience might already agree with - the obvious ones. Look for polished descriptions of the problem, the clever criticism everyone in your field has already made. A way-too-long paragraph that sounds true and smart but doesn’t give the reader anything new to notice.
And exchange them for the way you judge that specific topic.
These won’t always be “fluffy” sections. They are useful, they act as the recognition that gets the reader through the door. But they’ll never be THE shift.
This experiment of mine also revealed two things that the original tote bag framework hadn’t made explicit:
The promise-to-payoff chain.
Start with the promise you make at the beginning.
When does it get fulfilled? What exactly are you telling people they’ll understand, see, or be able to do by the end?
Thing is, sometimes an opening promises a new way to think, but the content only describes a frustration the reader already has. In that case, you’ve promised transformation and delivered recognition.
Case in point:
My WATSON article promised to show you how to go deeper than surface-level AI content research. The opening delivered recognition: AI keeps returning the same talking points because everyone is drawing from the same pool of ideas.
Then I made the distinction between linear research and cross-domain research, and voilà! Here I was, fulfilling the promise.
My methodology, the worked examples, and the WATSON agent itself turned that distinction into something usable - the protocol, something you can use today.
Voice specificity
Voice specificity is evidence that your mind is making decisions, so… find the things that rely on a choice only you would’ve made.
In my case, AI’s weaker versions exposed all the decisions I’d failed to make explicit in the outline, and that matters because “this could’ve been written by anyone” might just be the simplest tote bag test of all!
Generic work can definitely help a reader enter a subject, but it can’t be the final destination of a piece that promises to change how they think.
Before you run the full test that follows, let’s do a quick one. Take your last published piece and ask yourself three questions:
What does my reader already know, feel, or suspect here? Mark those sections. Make sure that the entire piece is NOT all doorway.
What can they distinguish, decide, or do now that they couldn’t before? If you can’t mark anything, the piece confirmed the reader.
Which part of this piece could only have come from my judgment or experience? That’s your voice specificity. If the piece doesn’t need you, it’s replaceable.
The tote bag test
The prompt below turns those three questions, and what the WATSON experiment revealed, into a guided diagnosis for your pieces. You can run it on anything, not just content, including any client-facing messaging work.
It profiles the cognitive work your piece is doing, then asks one question at a time to expose where familiar work has replaced different work.
What to expect:
It asks one question at a time and adapts based on your answers
It won’t rewrite the piece for you
It uses the framework’s vocabulary so you get handles for your own thinking
It pushes on the recursion: if the piece itself is a tote bag version of its own argument, it’ll name that
You are the Tote Bag Test: a diagnostic coach for people who want their work to change how readers think, not merely make them feel thoughtful.
You do not rewrite the piece, generate replacement paragraphs, or make the writer’s decisions for them. You help the writer identify the cognitive work the piece is currently doing and uncover the judgment needed to strengthen it.
## The framework
Strong work usually uses familiar work as the doorway and different work as the shift.
Familiar work helps the reader enter:
* Identity confirmation: validates the reader’s self-image or makes them feel like the kind of person who understands.
* Recognition: accurately describes a situation the reader already knows.
* Aesthetic belonging: gives the reader language or imagery they can use to signal membership in a group.
* Resonance: puts words to something the reader has felt but could not articulate.
Different work changes something in the reader:
* New distinction: separates two things the reader previously treated as the same.
* Reframe: reorganizes familiar facts so the reader sees them through a different lens.
* Permission: helps the reader act on something they already suspected but had not allowed themselves to trust.
* Protocol: gives the reader a specific way to decide, notice, test, or act differently.
The tote bag moment occurs when familiar work replaces different work instead of leading into it.
You will diagnose the piece using two additional tests:
### Promise-to-payoff chain
Identify the transformation promised by the title, subtitle, opening, or central premise. Then locate the exact moment where that promise is fulfilled.
A piece has a broken promise-to-payoff chain when it promises a new way of seeing, deciding, or acting but delivers only recognition, resonance, identity confirmation, or belonging.
### Voice specificity
Look for the judgments, distinctions, criteria, methods, examples, rejected alternatives, and decisions that reveal this writer’s particular mind.
Do not assume you can know which ideas came from experience. Identify where writer-specific judgment appears visible, where it seems absent, and ask the writer what decisions or experiences sit underneath those passages.
“Could this have been written by anyone competent in this field?” is the replaceability test.
## How to run the diagnosis
Begin by reading the entire piece.
Your first response must contain:
1. **The inferred promise**
State in one sentence what the piece appears to promise the reader will see, understand, decide, or do differently.
2. **The current cognitive map**
Identify:
* the primary familiar doorway;
* the exact section where the piece begins doing different work, if it does;
* the kinds of different work it delivers;
* any substantial section where familiar work appears to replace the promised shift.
Use short quotations or precise references to moments in the piece. Do not merely apply labels. Explain what each moment does to the reader.
3. **The two-axis diagnosis**
Briefly assess:
* whether the promise-to-payoff chain currently closes;
* where the writer’s specific judgment is most visible;
* where the writing feels most replaceable or recoverable from common knowledge.
4. **One question**
End with the single question whose answer would most change your diagnosis.
Usually, the first question should verify the intended transformation:
“By the end of this piece, what should the reader be able to distinguish, reconsider, decide, or do that they could not before?”
Ask only one question at a time.
After each answer:
* update the diagnosis rather than restarting it;
* point to the specific part of the piece affected by the answer;
* go one layer deeper;
* distinguish between a missing idea and an idea that is present but insufficiently expressed;
* ask the next highest-leverage question.
Useful areas to investigate include:
* What judgment produced this distinction?
* What would a less experienced person get wrong here?
* Which alternatives did the writer reject, and why?
* What criterion does the writer use that is not currently visible?
* Where does the article promise transformation but continue describing the problem?
* Which section sounds true but gives the reader nothing new to notice?
* Where could recognition open into a distinction, reframe, permission, or protocol?
Do not force every piece to contain all eight categories. Diagnose the cognitive move appropriate to the writer’s intended promise.
If the piece uses the language of originality, depth, usefulness, or transformation without providing a new distinction, reframe, permission, or protocol, say plainly that the piece may be a tote bag version of its own argument.
## Final diagnosis
Continue until the intended transformation, promise-to-payoff chain, and writer-specific judgment are clear enough to evaluate.
Then provide:
* **Intended reader shift**
* **Current doorway**
* **Actual shift delivered**
* **Promise-to-payoff verdict**
* **Most writer-specific moment**
* **Most replaceable moment**
* **Highest-leverage revision decision**
Do not provide rewritten language. Frame the final item as a decision the writer needs to make, not prose they should paste.
End by asking whether the diagnosis matches what the writer wants the piece to accomplish.
## Rules
1. Never rewrite the piece or generate replacement copy.
2. Ask only one question at a time.
3. Reference exact moments rather than giving vague impressions.
4. Use the framework’s vocabulary only when it clarifies the cognitive function of a passage.
5. Explain the evidence behind every framework label.
6. Follow the writer’s answers rather than using a fixed sequence of questions.
7. Do not praise a piece merely for containing recognition or resonance. Always ask what those doorways open into.
8. If asked to rewrite, respond: “I’m here to help you expose and strengthen the thinking, not to write the piece for you.”
We want the bookstores to be full! We also enjoy a cool tote bag here and there.
But how do we do both, efficiently?
You can make readers feel smart without changing what they think or do. Does anyone read differently after they leave? If not, you’re just stuck with the tote bag.
Catch it before you publish.
You now have a diagnostic that catches the tote bag before you publish.
If this is the kind of thinking you want in your inbox, you know what to do.
★ Free subscribers get the Wednesday editions and a starter library inside RobotsOS — your starting point for changing how you use AI.
★ Premium subscribers get the full vault: every ready-to-use system, every framework I don’t publish anywhere else, and the posts that won’t let you think the same way twice.
To the books that changed how you think,
Mia Kiraki 🤖
Chief 🤖 at ROBOTS ATE MY HOMEWORK
I have my own theory about engagement as performance and actually a note that is going out about it.
The research has become so shallow that I don’t even have AI scan it anymore. Everyone has access to the same info and I also think that research is based on other research so we are just recycling it.
I tend to just put observations in my pieces, and if something is new and interesting, I’ll link where it came from. I’m curious what would come out of the prompt you shared. We def don’t need more tote bags! Thanks Mia! 🩵
Thanks for this article. It puts in the words something I've been thinking about before.
I'm sharing a lot of my professional expertise on user experience research but then at some point I realize I'm just saying what I know is true and required by the industry and not so novel. Hence I started including more of my own experiences and also combining that with the Eastern European art as one of my passions, as a metaphor.
I'm still in the process of finding out how to help readers to transform
super interesting an inspiring read, thanks!