Small apps, built in public.
Working experiments built on Google AI Studio, Claude, and custom stacks. Each one started as a question I wanted to test. Some are stable. Some break weekly. All of them are live. Click through to run the demo.
Frame Critique
Upload any still image. Get a 300-word color-science critique written in the voice of a working colorist, not an AI. Built to test whether the critique voice can be preserved through a prompt chain.
Brief Grader
Paste a creative brief. Get it scored on six axes of clarity — objective, audience, specificity, constraint, deliverable, and tone. Rubric is mine. Critique is also mine, translated through an API.
Sycophancy Detector
A small classifier that rates an LLM response on how much it is performing agreement versus engaging with your actual point. My ongoing experiment in measuring the one thing AI is getting steadily worse at.
Article Pipeline
The actual automation that produces the AI-Assisted pieces on this site. Open-sourced so you can see exactly how much human vs machine work goes into each piece.