Fragments
A public notebook. Thoughts, not essays.
Unpolished by design. Written when something is worth saying, not on a schedule.
On details nobody notices
Mar 13, 2026
craftdesign
The grain on this website took three hours to get right. Nobody will notice it. That's exactly why it took three hours.
There's a specific opacity - around 4% - where texture stops being a design decision and starts being a physical property of the surface. Below that it's invisible. Above it, it's a filter. At 4% it just *is*.
Most design decisions don't have a right answer. Some do. Finding the ones that do is the whole job.
Slowness as a competitive advantage
Mar 13, 2026
buildingprocess
Everyone talks about shipping fast. Few people talk about the things that only emerge from going slow.
Candor took six months longer than it needed to because I kept redesigning the feed. But the question I was really asking - what does a social product look like without engagement mechanics - only became clear in month four.
If I'd shipped in month two I'd have shipped the wrong thing faster. Not sure that's winning.
The edge case that broke the graph
Mar 13, 2026
engineeringhealthcare
Fever + fatigue + headache activates about 40 nodes in Aushadham's symptom graph simultaneously.
That's the problem. Those three symptoms are shared by everything from the common cold to early-stage dengue. The graph has no way to know which branch to eliminate first without more information - so it asks everything and the user experience collapses.
The fix isn't more questions. It's an urgency layer that runs before the graph - a separate check that asks: is this a "see a doctor today" situation or a "let's figure this out" situation.
Simple. Took me three weeks to realize it.
Variable fonts are genuinely weird
Mar 13, 2026
designtypography
A variable font isn't multiple fonts. It's one font with axes - mathematical dimensions you can move through continuously.
Fraunces has a wght axis (300 to 900), an opsz axis (optical size), a SOFT axis (how rounded the letterforms are), and a WONK axis (enables specific alternate glyphs).
That's just strange and I like it.
Why stat decay is the right call
Mar 13, 2026
buildingdesign
Early testers hated the decay mechanic in SuperSoul. "Why does my character get worse when I'm busy?"
Because you actually do. You don't maintain peak fitness during exam week. The character reflects reality - it just makes the cost of neglect visible.
The complaint isn't really about decay. It's about stakes. Some people want progress tracking without consequences. That's a different product. Hence: Explorer mode.
The mechanic was right. The product just needed two modes.
More fragments as they form. No schedule, no promise.