Research & writing
Evidence before claims.
We publish the work behind our products: field studies of people learning with our tools, papers on legible AI and human pace, and notes from prototypes we retired. Read the methods, results, and negative findings — not just the polished conclusion.
2026
On Building Slowly in PublicWhy we treat writing as part of the lab — not marketing after the fact — and what we refuse to publish.Learning at the Speed of UnderstandingSix months alongside forty learners: what happens to comprehension when a tool refuses to move faster than its user.Legibility as an Interface PropertyA framework for machines that can be asked why — and what honest, complete answers cost in model design.
2025
Quiet Hardware: Intelligence Without Attention CostDesign notes from our first ambient devices: how much intelligence fits in an object that never asks to be looked at.Measuring Capability Retention After Tool RemovalA method for the test we hold ourselves to: is the person more able once the tool is put away?The Human Pace: Latency, Comprehension, and TrustResponse latency shapes trust more than accuracy does. Evidence for slowing systems down on purpose.Tools That Teach: Notes from the First PrototypesEarly lessons — including two prototypes we retired, and why spectacle kept beating usefulness in testing.
Studying human-paced learning, legible AI, or attention-aware hardware? We collaborate with a small number of labs and universities. Tell us what you are working on and where our methods overlap at[email protected].