Writing
Working it out
A working record of design and product work as AI changes it. Partial thoughts, things that broke, and what I would do differently. Written mostly so I stop relearning the same lessons.
-
I forgot a fact about my own work. A machine had to remind me.
Fact-checking a post about his own work turned up a detail he'd completely forgotten — not misremembered, gone — because nothing had been tracking it.
-
Three circles and a second sort
A method rather than a case study. Three overlapping circles for sorting AI features you have already shipped, why the second sort matters more than the first, and an open ask for how other people are handling the same problem.
-
Nothing checks the path
Automated design checks are all screen-scoped. The errors that survive live between screens, which is where the only human left is still doing the work by hand.
-
The snow day
When GitHub went down for 7.5 hours, a product design team didn't have a runbook. They had a snow day.
-
Resolution, not scale
A reaction to Harvard/MIT's 8.3B-persona simulation, reframed as the actual failure mode of UX personas: not sample size, compression.
-
They always say yes. I still hate asking.
The expensive part of metered AI isn't the bill. It's what people quietly stop trying.
Nothing matches that yet.
← Back to portfolio