Audit and iteration loop
I move between Claude and GPT to audit prototypes, question weak decisions and improve each round. I add my own direction wherever product judgment is needed.
Repeated until the decision holds up
AI helps me move
faster. I keep the
direction human.
I use AI across product design, research and development—not as a shortcut to a finished answer, but as a way to explore, test and improve the work.
My working system · always evolvingProcesses I use
Variant exploration
I create several versions of a journey, flow or interface, then compare them and select the strongest direction. When useful, I personalize the result for a specific user or context.
SQL, Supabase and auth
I generate and review SQL scripts, run them in Supabase, and build authentication and update flows. Each iteration improves the schema, permissions and product behaviour.
Git to production
I use Git to keep changes traceable, test work in small commits and push production-ready code without losing earlier decisions.
Code back into design
I use Codex to read working code and create a structured prompt with exact layout, component, state and token details. I bring it into Figma Make, extract the design tokens and continue from the existing product instead of starting again.
Research synthesis
I turn scattered research into useful frames such as personas, SWOT analysis, competitor reviews and feature benchmarks—only when they help answer the product question.
Living knowledge base
I connect Drive, Docs and Sheets into a shared knowledge base. New findings update the source sheet, so future analysis starts with better context.
Breaking the template feel
I use design skills, plugins, mood boards and personal references to remove generic AI patterns. The goal is work that feels specific to the product, not generated by default.
Prompt refinement and review
I pass prompts and outputs between Claude Code and Codex to find gaps, sharpen instructions and review the result from a second perspective.