The value of a creator is shifting from technical mastery of complex software to directing AI agents. The core skill is no longer tool proficiency but the ability to articulate and guide an AI towards a creative vision or story. The human becomes the director, not the technician.
Researchers chase pushing technological boundaries (e.g., complex text generation), which often misaligns with customer needs. Successful AI products solve simple, high-value problems like background removal or lighting correction—tasks that may seem boring to researchers but are crucial for users.
Current AI interaction is a one-way command from user to model. The next generation of tools will behave more like human collaborators, asking clarifying questions to resolve ambiguity and better understand the user's intent, just as a professional at a creative studio would.
When generating personal content like headshots, users can be unhappy with a result that is perfectly accurate but unflattering. This shows that 'truth-seeking' and 'happiness-seeking' are different objectives. AI tools need to empower users to achieve a result they are happy with, even if it deviates from pure realism.
Simply building what users ask for can trap a product in old paradigms, like reinventing Photoshop's lasso tool for an AI context. A successful strategy involves staying slightly ahead of user adoption, introducing new capabilities that fundamentally change their workflow, and guiding them toward a more efficient future.
While AI lowers the barrier to entry for creating competent work, it also acts as a powerful lever for true visionaries. By removing technical constraints, AI enables the most creative artists to execute more ambitious ideas, potentially increasing the quality gap between 'good' and 'great' art.
Choosing between pixels, vectors, 3D scenes, or other representations isn't a purely technical decision. The best representation is the one that best facilitates the specific type of control a user needs. If they need to change text, a pixel-level representation is wrong. The control layer should define the representation layer.
Users don't use tools in isolation. They create complex, unexpected workflows, such as using an image identity-preservation tool to generate individual frames for a separate video model. This 'hacking' signals powerful unmet needs and new integration opportunities for product teams.