05
All tests green. Why does it still look horrible?
Technical correctness kept a mobile effect safe; human perception decided whether it communicated the intended idea.
- Frontend craft
- Visual QA
- Product judgment
Field Notes / Building with AI
Concrete lessons from work in progress.
Product pivots, architecture tradeoffs, agent behavior, and the evidence behind what changed. Written to be useful before it becomes social content.
17 / Published NotesPage 02 of 02
Newest first. Each Note starts with the decision and keeps the technical detail in service of the lesson.
04
Moving a file should not publish it
A useful content automation needs permission boundaries as much as it needs a writing prompt.
- Content systems
- Human approval
- AI automation
03
The agent platform I decided not to build
A technically credible prototype was not a differentiated product. The useful move was to preserve the machinery and replace the claim.
- Product strategy
- Agent systems
- Differentiation
02
A live AI demo can be less honest than a replay
Interactivity is attractive, but a controlled replay can reveal more about cost, evidence, failure, and product boundaries.
- AI products
- Trust
- Portfolio design
01
Three interfaces, one source of truth
Why this portfolio can change its visual personality without creating three versions of the truth.
- Frontend architecture
- Design systems
- Product communication