Research

Work that outlives the deadline.

Some questions are worth studying past the point a project requires. This is where that work lives: a long-running program on readable type, field notes from autonomous survey hardware, and technical writing on building software people can understand and trust.

A plot of contrast ratio against text size, 8 to 48 CSS pixels, showing where text stops being readable. WCAG 2.2's minimum-contrast rule is drawn as the hard staircase it is: 4.5:1 for normal text, stepping down to 3:1 once text reaches 18 point, with a dashed early step at 14 point bold. A blue curve above it marks the contrast a low-vision reader actually needs; it rises steeply as text gets smaller and only meets the standard at 48 pixels. The grey region between the two is labelled THE GAP: every point inside it passes 1.4.3 and is still unreadable. Three specimens of the word readable, set at 16 pixels in real inks, are plotted at their computed ratios: 18.58:1 clears the curve, 2.55:1 fails the standard outright, and 5.92:1 passes the standard yet falls inside the gap. That middle ink is traced as one horizontal line at a constant 5.92:1 across to the same word at 32 pixels, where it crosses the threshold and becomes readable — the ratio never changed, only the size. An axis below runs 2005 to 2026, the two versions of the research, marked at 2008, when WCAG 2.0 set 4.5:1.

01Typography & accessibility

The Universal Web

A field guide to readable type on screen, begun at the Rhode Island School of Design and revised for the modern web across 21 chapters. It retests twenty-year-old low-vision-legibility findings against high-density displays, variable fonts, and WCAG 2.2. The historical record is preserved and clearly separated from present-day commentary.

Read the guide2005 archive

02Autonomous & geospatial

Field & survey notes

A drone-borne magnetometer survey program in western Colorado: sensor and airframe integration, flight-log analysis, and turning raw magnetic data into interpreted maps. The write-up covers the systems thinking and hardware/software integration without publishing sensitive locations or proprietary data.

Read the field reportThe survey platform

03Technical writing

A Field Kit in PETG

Eighty-odd printed parts hold a flying magnetometer together: clamps, cradles, couplers, and a towed sensor pod, all printed overnight on a desktop machine. Notes on 3D printing as a field-engineering practice, revision by revision.

Designing for an Invisible Constraint

A drone that measures magnetic fields cannot be built like other drones: every screw is a design decision. What a survey aircraft taught me about constraint-first design — a lesson twenty years of typography had already tried to teach.

The Agent in the Workshop

I gave Claude three specialized subagents — a mount designer, an OpenSCAD author, and a print checker — and let them iterate real drone parts through eighteen logged revisions. Notes on what AI-assisted mechanical design actually looks like.

The Performance Budget of a Typeface

If the brand face earns its place, make it cheap: one variable file instead of four statics, ruthless subsetting, honest font-display, and a fallback engineered so the swap is invisible. Font loading as an engineering discipline.

All writing →