· Ryan Ashton · creativity · 4 min read
True Creative Freedom in the Age of AI
AI does the heavy lifting now, so what makes a human creator stand out? Taste, values, and a voice AI can learn to express.
Last night I did a talk for Hong Kong Web Dev Meetup about honing one’s creativity now that AI and agents do the heavy technical lifting when it comes to building websites, and it got me thinking about two specific questions: How creative is AI really? And if AI is handling a large proportion of “creative” work now, what does it take as a real human being to stand out in creativity?
Because LLMs and AI are trained on existing data, they cannot truly generate anything new; they are the ultimate execution engine—a relentless generator of average, templated output that shifts the human role entirely from “builder” to “curator.” A great approach to truly novel ideas is to take two or more separate concepts, things, or principles and combine them in entirely new ways. To generate originality step-by-step, we need a practical framework. And because the agent can build and iterate fast, getting your ideas out into the world should be prolific—you should leave your own judgments about good and bad at the door.
AI cannot replace the human experience. The very fact of being human and being able to connect to fellow humans through culture, stories, emotion, and sharing one’s real-life experiences becomes the currency that makes your creative input valuable. A real human story told by a real human, with the pain and pleasure and ups and downs, is valuable. Being specific about what you want to talk about and what you want to show the world means having clear values and principles: “This is what I stand for,” “This is what I believe,” “This is right and this is wrong.” Every human needs to step up and own their reality and tell people about their reality, values, thoughts, and feelings.
This also means engaging in the act of creation over and over and over again, because each time you do that—even with the help of AI—you are refining your taste and your unique voice. Using AI to help you create has absolutely nothing wrong with it, but you need to be meticulous about what you put out into the world. Read every word that AI writes, watch every second of video it generates, and examine every pixel of images it makes—firstly to make sure it doesn’t make embarrassing AI glitches, but equally importantly, that it expresses your humanness and reflects your values and viewpoint exactly.
To make sure AI starts expressing your vision more in line with your values, you can get an agent or AI to examine things you have written—with or without AI—to create a framework for extracting the hallmarks of your work. (My suggestion for doing this: feed a curated corpus of your best writing directly into an analysis script or prompt with explicit instructions to isolate recurring syntactic cadences, preferred transition markers, and idiosyncratic sentence lengths, then codify those metrics into a persistent system prompt constraint).
To stand out in this time, you can also purposefully add imperfection to your creations. Obviously, when creating a website, functionality cannot be sacrificed—that’s just a broken thing - not desirable. To deliberately add human imperfection even in digital products so they work perfectly yet retain a human signature, conduct a friction audit: scan your layouts and code for places where it feels too smooth—like standard CSS transitions, mathematically centered text blocks, or sterile UI animations—and inject intentional, organic asymmetry into layouts or use deliberately raw or handwritten graphic accents.
I also enjoy using AI to generate images because you can ask for very specific things that would never be found in stock photos or video libraries. Adding very specific prompts to the creation process can make it unique to you—for example: <image prompt> and then make it feel very very marshmallow and with an angular, glass shard edge to it.
Finally, while true creativity with the help of AI is possible on established commercial platforms, you are always at the whims of the platform, and if the system prompt changes, all the work you put into developing a style goes up in smoke. Instead, I advocate for deploying open-weight models on private cloud infrastructure where you have full control over as much of the input and output as possible, and if so inclined, you can even take it a step further by fine-tuning a model specifically for your own tastes.