Flint: Standing in the loop of thought

As Generative AI evolves into an agent capable of complex reasoning, traditional "chatbot" interfaces often promote automation over engagement. By delivering instant answers, these systems can lead to cognitive offloading and a loss of human agency, reducing users to passive supervisors of automated output.

Flint is a custom software prototype designed to restore human agency by shifting the AI from an "oracle" that provides answers to a "methodologist" that supports co-thinking. The project traces an evolution from linear chat to a spatial "co-thinking canvas”. Unlike standard AI tools that generate finished content, Flint utilizes Generative UI to diagnose a user’s cognitive task and dynamically provide structural scaffolding. A core design principle of Flint is the separation of concerns: the AI is structurally prevented from writing on the canvas, ensuring that while the machine shapes the environment, the human remains the sole architect of the reasoning and ideas.

By externalizing the reasoning process into editable, manipulable objects, Flint reintroduces the "productive struggle" necessary for deep understanding. This research demonstrates that staying "in the loop" of thought is not about passive viewing, but about active collaboration and control over the structure of the reasoning process itself.

***

I’m an AI product designer and entrepreneur based in Toronto. I come from a Computer Science and UI/UX background and have spent the past two years building and launching AI products. I love learning new things and making cool stuff, and I’m excited to be designing at a moment when AI is reshaping how we build, think, and create.