Writing
I write about what a product, a form, or an institution decides for the people using it: which options they see, whether it builds their judgment, and who fixes its mistakes.
Selected from The Crumple Zone, where the rest are · RSS feed
Start here
- The Option Nobody Asks About Some choices stay hidden because asking about them costs standing. Whether they stay hidden is partly an interface decision.
- 1,200 Physicians Trusted a Contract Tool and Needed It Once Medical training grades judgment. Products don’t.
- Staff Fix the Errors Your Dashboard Doesn’t Show An automation can report 94% straight-through processing while staff fix mismatched codes in shadow spreadsheets. The metric counts their work as zero.
The mark at the end of each row is that essay’s revision record. It has one column per passage, and a column stands up where the passage was rewritten after this copy was first committed: the later the rewrite, the higher. A flat line is an essay as it was first committed.
2026
- The Time AI Saves Has to Go Somewhere Documentation gets faster, and the saved minutes are absorbed before anyone decides where they should go.
- When Intelligence Gets Cheaper, What Becomes More Expensive? If generation and execution get cheap, the constraint moves to who may act, whether it worked, and what happens when it did not.
- Administrative Systems Removed the Stop Button Any Toyota worker can pull the cord that stops the line. Put it back.
- Treat Users as Threats and They Learn to Game You Zero Trust leaked from network security into how institutions treat people. When a fraud flag has no workable appeal, gaming the system is the rational move.
- A Decision Against You Is Instant. Fixing It Is on Hold. A card declines, a benefit vanishes, an account goes under review. What decides who gets hurt is how hard the decision is to reverse.
- Design for the Person at 15% Battery Stop designing for the idealized “Hero User.” Build resilient interfaces that work when your user is stressed, tired, and operating on 15% battery.
2025
The argument in practice
- Framework Approving and correcting AI systems A worked example, four stages of approval, and a six-step review for when evidence changes.
- Current work Software and evaluations Agent permissions, source-change detection, and draft tests of error recovery, alongside NextConsensus’s forecasts.
- Standards Ethotechnics technical standards Proposed standards for logging decisions, challenging them, timing the fix, weighing evidence, recording delegation, and taking permissions back while a system runs.
Revised