I designed the digital systems that shape behavior at scale — now I study how they fail.
Current work
An independent research program on how complex digital systems — the ones that consume their own output as input: markets, models, recommenders, more — can invisibly lose the ability to make progress before they suddenly break.
The Loopzero benchmarking framework I built is falsifiable and the directional evidence consistent. That said, no detector — neither the standard methods nor my own — has yet reached an accepted operating point under the locked test. So the framework stands, and creating a working detector is still an open problem.
The research program → loopzero.orgThe throughline
I came to this sideways.
I studied art history at Northwestern, then found film through post-production. Cubism had shown me the goal: an image that refuses a single fixed viewpoint. The Royal College of Art trained me as a designer; the pursuit carried on, bringing that eye to the moving image as a director at Ridley Scott Associates.
Virtual reality finally exploded the frame — not just past the screen but all the way around you, embodying what I’d spent years circling. I made cinematic VR and founded a studio to build it. Then the pursuit turned from the image to the system: I co-founded Pupil, a machine-learning and spatial-data company, and went on to Meta Reality Labs as Lead Product Designer for Engagement & Growth, building immersive digital products at scale — where I began to see how systems shape attention itself.
A few years ago I started to suspect that the breakdowns around us — radicalization, polarization, outrage, eroding trust and attention spans — shared a common mechanism: digital systems feed on their own output until they consume themselves, and us with them. I spent the next two years taking the question apart, convinced there had to be something formal underneath — a claim specific enough to fail. That search is Loopzero.
Design teaches us that the problem you’re given is rarely the problem — I think I’ve found the one worth naming.
Contact
I’m looking for critical readers and collaborators — people who’ll push on the claim, especially from recommender evaluation, statistics, or formal methods. The harder the scrutiny, the more useful.
Email → d@loopzero.org