My role

Product design, UX, information architecture, AI-assisted build

Impact

A working studio app in one day, then one home for 14 studio tools

Challenge

Most project tools count work: tasks, due dates and hours. Early design work is different. Its real product is understanding. I needed a tool that shows what a team knows, what it assumes and what it does not know yet. I also needed one home for the many small tools I build for design, making and music.

Process

The house

Each room is one practice. The entry court takes new inquiries. The hall shows the next step for every active project. The studio holds service design and strategy. The drafting room holds visual design, type, colour and site themes. The workshop holds furniture design. The music room holds practice and teaching tools. The den holds time.

Measure understanding, not work

  • Show one next step at a time.
  • Record an unknown as progress. Uncertainty is data.
  • Give a disproved assumption the same message as a supported one.
  • No tasks, due dates, streaks or leaderboards.
  • Keep the phases open. A team can start at any phase.

How I built it with AI

I wrote the requirements and the product principles first. Claude Code wrote most of the code and the tests. I made the decisions and reviewed the result.

Every session follows the same protocol. At the start, the AI reads the rules, the decision log and the last session note. At the end, it runs the tests, records new decisions with a reason, and adds a session note. The decision log now holds 143 decisions. When a decision changes, the old one is marked as replaced, not deleted.

Each screen change is checked against the Laws of UX. When a law and a product principle disagree, the principle wins, and the session note records why.

Key insights

  • An unknown is progress. Naming what you do not know is the first useful step.
  • A clear model makes AI a better partner. The rules, decisions and principles live in plain files, so every session starts from the same place.
  • My part is judgment. The AI drafts fast. I decide what to build, what to leave out and when it is right.

Outcome

The first working version came together on 30 September 2026: eight milestones, a discovery questionnaire and 104 logged decisions in one day. That same week it became the studio hub. Every tool I build now opens from the house.

The vault of Markdown files is the source of truth. The same project opens in the app and in Obsidian, and nothing is ever deleted, only moved to a trash folder.

Reflection

The fastest part was the code. The slowest and most valuable part was deciding what not to build.
The GamutWorkshop studio plan: seven rooms laid out like a house

The studio plan. Each room is one practice, and each room opens its own tools.

Project cards that show confidence, explored phases, unknowns and the next step

Projects show confidence, how much is explored and the open unknowns. They never show a task count.

One next step at a time, and the six open phases below it. Sample data.

One next step at a time, and the six open phases below it. Sample data.

FlowMap: every note shows its type, its status and where it came from.

FlowMap: every note shows its type, its status and where it came from.

← All projects

Working on something similar?

I like to compare notes with people who design, build and make. Let's talk.

Let's connect