I build, test, and write about how human judgment shapes AI.
Recent work includes interfaces for expressing taste, experiments in steering image models, and essays on values and power.
Latest essay / Sep 2026 AI Swarms: Perception, Values, and Power Whose judgments should an AI swarm inherit?Tools, games, and visual experiments.
These hackathon builds explore what happens when people can shape the output, inspect an interpretation, or make decisions with an AI companion.
OpenAI hackathonAstralogy: how far can we go with AI?
A playful space adventure where you find your bearings, shape a route, and work with an AI companion to get home.
Read the story & watch the demo ↗
Claude hackathonGenerative Remix
A workbench where AI builds the controls and you shape visual effects by eye, guided by your own taste.
Explore the prototype ↗
Figma hackathonIntent Configuration Interface
An editable map of how AI reads your preferences, so you can adjust its interpretation before it generates an output.
Explore the interface ↗
Cursor hackathonHolovibes
Digital souvenirs from San Francisco, with holographic foil and depth that respond to your movement.
Explore the project ↗What AI learns, and how we judge it.
AI Swarms: Perception, Values, and Power
When AI agents learn from one another, whose judgments do they inherit - and who gets to challenge them?
Read the essay ↗ Research pilot / June 2026Steering by recognition: can an image model learn a quality we struggle to describe?
I tested whether visual examples could define a quality and become a control inside an image model. The outputs changed, but establishing whether they captured the intended quality proved harder.
Read the experiment and its limits ↗These questions also shape my work at Thirdplane Labs, where I design and review evaluations for AI products. I investigate where AI and human judgments diverge, and what those differences mean for the task.
Explore evaluation work at Thirdplane Labs ↗