Project Relay
An AI-assisted product-building workflow that turns a structured idea into implementation, evaluates the result, and revises its own work before human review.
Innovation Lab / The Research Greenhouse
Questions I became curious enough to build something around.
Some questions are easier to understand once I can touch them.
The Research Greenhouse is where I test ideas in motion — building with AI agents, designing decision systems, challenging assumptions, and paying attention to the places where technology still needs judgment.
All of them teach me something.
We experiment. We learn. We evolve.Human-led. AI-enabled. Responsibly built.
On the bench
Three questions in motion.
Real experiments, abstracted only where privacy or responsibility requires it.
Project Signal
An AI-assisted decision agent that continuously evaluates a fast-moving environment, verifies candidates against fixed constraints, and escalates only the strongest possibilities for human review.
What happens when an AI agent has to make decisions inside a fast-moving, constraint-heavy environment?
Density + Clarity
An interactive decision-design prototype that changes the amount of information presented at once to explore density, clarity, and modeled decision effort.
When does more information stop making a decision clearer?
Open live experimentFeatured live experiment
Density + Clarity
A design question made manipulable — not an empirical research claim.
When does more information stop helping?
Move the slider to see how modeled clarity changes as more information is added.
4 / 10Strongest clarity
Enough context to support the decision without unnecessary overload.
Modeled decision effort34s
Strongest modeled zone: 3–5. The range where the model balances useful context with lower overload.
About the model. A deliberately constructed design experiment, not participant research.
Lab notes
What the experiments left behind.
Project Relay“More capability does not make direction less important. It makes ambiguity more expensive.”
Project Signal“A useful agent needs permission to return nothing.”
Project Signal“Not every source that can answer a question should be allowed to authorize a decision.”
Public / private principle
Some real experiments can be discussed at the level of the question and learning while the underlying project details remain private.
Archive — Nothing has been put away yet.