19 Oct–25 Oct 2026
The pattern this week is that progress keeps coming from forcing a process rather than trusting good intentions, three-step sync disciplines, single-step workflows, and even asking the AI to explain its own defaults so I know what I’m actually choosing.
Everything I shared in Substack Notes last week, in one place. 11 notes.
Explain means two different things
A term like “Explain” needs to match what the reader expects — consumers want the business meaning of a number, not how it was calculated.
Navigation debt sneaks up on you
A flat layout that works fine with a handful of items becomes unusable once they accumulate, so navigation needs revisiting as content grows, not just at launch.
Configurable persona, not fixed voice
Letting customers set the avatar and language for an AI assistant means one capability can be tailored to whatever persona their market actually wants.
Disagreement needs its own space
Modelling a concept well sometimes means explicitly capturing where people disagree, not just the agreed definition.

Detailed feedback beats vague feedback
Specific, detailed feedback is more valuable than general comments because it is immediately actionable and often doubles as content ideas.

Blue Book step pages, iterated
Two Blue Book step pages got reworked this week — worth a look if you’re following the Modeling Business Concepts series.
Build the component where the data lives
A reusable component designed in isolation from real data causes problems, so building it in context first and syncing back to the design system works better.

Force the discipline, don’t rely on habit
A multi-step process you know you should follow still gets skipped under time pressure unless you build something that forces every step to happen.
Step 4 needs another set of eyes
A major rework isn’t necessarily right just because it’s done — some iterations need outside review before you trust they’ve landed.
Collapse unnecessary two-step flows
If creating something always requires an immediate second edit, that’s a sign the two steps should be merged into one.
Default AI styling converges on the same look
Relying on default AI-generated design means your product can end up looking like every other AI-generated dashboard, carrying decoration that signals nothing.





