Perspectives on
applied AI and product design.
Short pieces on strategy, analytics, and design — across decision tools, AI products, dashboards, and consumer software.
- 01
Make extraction failures visible.
Most "AI extracts X from documents" demos quietly hide their failures. The design move that turns an extraction agent into something a domain reviewer would trust is per-cell source attribution.
May 07, 20263 min read - 02
AI as thinking aid, not calculator.
Two failure modes dominate AI-augmented decision tools — the calculator that produces output the user cannot defend, and the vanilla tool that doesn't improve on guessing. The middle path is AI inside the user's reasoning.
May 07, 20263 min read - 03
What aggregators dedupe away.
News aggregators optimize for volume. What the serious reader actually wants is what each source emphasizes that the others omit — exactly what aggregators dedupe away as duplication.
May 07, 20263 min read - 04
Why analytics dashboards fail to drive decisions.
The cause is rarely the underlying data. It is the absence of a clearly defined decision the dashboard is intended to support.
Apr 22, 20263 min read - 05
Structured discovery in ambiguous projects.
Projects that begin in ambiguity benefit from a discipline most teams skip: structured discovery before hypothesis formation.
Mar 14, 20264 min read