John Rector · September 8, 2026 · AI & Creation
Stop Consuming and Start Creating with AI. Then Consume What You Create.
Make something you wish existed. Become its first audience.
You have watched enough demonstrations of what AI can do. Make something you want to experience.
A short story with the ending you have been looking for. A visual explanation of the idea you keep trying to articulate. A song for a particular mood. A tiny tool that removes one irritation from your day. An essay that finally puts your own question into words.
Then read it. Listen to it. Look at it. Use it.
Stay long enough to discover what you actually made.
That second instruction matters. Creating with AI can become another form of scrolling: prompt, generate, glance, discard, repeat. The screen keeps changing, but you never develop a relationship with anything on it. You have become the operator of your own endless feed.
Give one creation your sustained attention. Let it become an experience.
You are allowed to be the audience.
There is a peculiar pressure attached to the word create. It seems to demand a public result: build an audience, launch a product, publish the work, prove that other people want it.
Begin closer to home. Make something because you want it.
You can write a story without becoming a novelist. You can make an illustrated field guide to the birds outside your window without starting a media company. You can build a little game and play it on a quiet evening. A creation can earn its place in your life before it earns anyone else’s attention.
AI makes it possible to attempt forms you might previously have admired only from the audience. You still have to decide what is worth making, what belongs, what feels wrong, and when the result deserves another pass. Those decisions give the work direction.
Your taste becomes more useful when it has something specific to work on.
The important generation happens in you, too.
There is research behind the value of active participation, although we should be precise about what it establishes. In their 1978 generation-effect experiments, Norman Slamecka and Peter Graf found better memory for words participants generated themselves than for words they simply read. These were controlled word-memory tasks. They did not test AI collaboration.
The practical inference is modest: doing some of the mental work can matter. A model generating an answer is not the same event as you generating an understanding.
A later field experiment involving nearly a thousand high school mathematics students in Turkey makes the distinction concrete. GPT-4 assistance improved practice performance, but students using the unrestricted interface performed worse than the no-AI group when assistance was removed. A tutor designed to protect learning largely mitigated that harm. This was a particular educational setting, not a verdict on every creative use of AI.
It does suggest a useful question for your own process: where are you still thinking?
Supply the observation. Choose the example. Explain why a passage fails. Ask the AI to challenge an assumption. If understanding is the purpose, try explaining the result before asking for another explanation. Keep a meaningful part of the work inside your own attention.
Consumption becomes feedback.
While creating, you know what you intended. When you return as the audience, you meet what is actually there.
The paragraph you thought was clear takes three readings. The image is beautiful but tells the wrong story. The tool works in a demonstration and feels awkward when you need it. The song has a moment you want to hear again.
These are useful discoveries. They give the next version a reason to exist.
Make the experience complete enough to judge. Read the entire essay instead of inspecting the opening. Listen to the whole track instead of generating five more introductions. Use the tool through the last step. Enjoy the parts that work; you do not have to turn every private pleasure into an improvement project.
When you do revise, change something because you noticed something. An additional generation is easy. A considered revision requires an opinion.
Bring in what the model cannot know about you.
AI assistance does not automatically make a work distinctive. In Doshi and Hauser’s short-story experiment, access to AI ideas improved evaluations of individual stories, particularly for writers with lower measured baseline creativity. Yet AI-assisted stories were more similar to one another. The study concerns a specific writing task; it does not establish that every use of AI makes all art alike.
For your practice, take it as an invitation to contribute particulars. The odd detail you remember. The unresolved question. The sentence you disagree with. The place you know. The feeling you have trouble naming.
Ask for alternatives, but make the choice yourself. Keep what surprises you for a reason you can recognize.
And keep encountering work you did not commission. Read people who disagree with you. Listen to music outside your preferences. Follow evidence that complicates your explanation. Your own creations can become a comfortable enclosure if nothing unfamiliar ever gets in.
“Consume what you create” is a practice of returning to your work, not a promise that your work should become your entire world.
Start with one thing.
At your next moment of automatic browsing, pause. Name something you would like to read, hear, see, or use. Make a small version with AI. Bring enough of yourself to it that you have a reason to care.
Then close the prompt box and experience the result.
You may love it. You may discover that your taste is ahead of your ability to direct the tool. Either way, you now have something more useful than a vague intention to become creative: a work you can return to.
Stop consuming and start creating with AI.
Then consume what you create.