AI Practice

Building ideas, together with AI

AI is driving the cost of execution toward zero. An idea can become a usable product in a matter of hours — validated, discarded, rebuilt. Trial and error is no longer expensive.

But once execution stops being the bottleneck, judgment becomes the only one. AI can generate endless options; it can't tell me which one is right. Design and taste decide whether AI's speed leads to something good, or just a pile of mediocre output.

Tools keep changing; the method stays. The process from concept to launch is the capability that's actually stable. The tools are just the replaceable variable inside it.

So the ideas and designs that used to live only in my head, or stall at the prototype stage, now get built — one by one. That's what AI gives me, and it's my answer to it.

How I work

01

Conceive

Brainstorm and anticipate risks alongside AI, and define the problem.

My roleSee wider, think earlier

02

Define

Produce product design docs and development-flow docs, fast.

My roleTurn ideas into executable specs

03

Design

Figma-led, iterating with Claude Design and Figma Make; imagery from Midjourney and Google Flow.

My roleOwn the visual and interaction quality

04

Build

Cursor and Claude Code turn the design into real code.

My roleReview screen by screen, like inspecting a delivery

05

Refine

An ongoing dialogue with Claude and Claude Code, correcting each other.

My roleJudgment, at every step

With AI

Traditional · A team · Weeks to months
With AI · Just me · Days to a week or two
3–5 days
Research
~1 hour
3–5 days
Docs
1–2 hours
1–2 weeks
UI design
Hours – 1 day
Weeks (needs engineers)
Development
Days
Stops at prototype
Launch
Goes live

The difference isn't just speed — it's that I also complete the whole "build → launch" stretch.