Build Notes
Hikari Brandan’s workflow: problem, research, think, PRD, build with AI, test, break it, audit and fix, ship. Factual notes and examples from his products.
Role & attribution
Written by Hikari Brandan from his product development and debugging experience.
Development workflow
- Problem
- Research
- Think
- PRD
- Build with AI
- Test
- Break it
- Audit & fix
- Ship
Problem
Producing more code does not answer whether the product should exist or whether it works for the person using it. My workflow starts with the problem and the customer context.
Decisions
- Research the task, constraints and existing behavior before choosing an implementation.
- Write requirements and direct coding agents with a defined outcome.
- Inspect the implementation, test the experience and deliberately look for failure cases.
- Debug, audit and verify changes before shipping; keep known limitations visible.
What I learned
The UGC Camera and MenuTap story grew from an earlier, broader FoodSpot exploration. The Mac projects also show why fast implementation needs a feedback loop: dictation can lag, file automation needs review and display streaming remains experimental.
I want to keep improving this judgment alongside stronger engineers, close to both the product and its customers.
Limitations
These are authored summaries of my supplied workflow and project notes. They are not a test log or a claim that every project has the same verification history.
AI tools listed here are development tools. They are runtime dependencies only when a product explicitly integrates a provider, as iBrain can.
Debugging example
When login appeared to complete but downstream operations failed, I inspected the callback URL and browser console, then traced Supabase Edge Function logs and database access. In some cases, row-level security or data access was blocking the operation.
Workflow tools
- ChatGPT / Codex
- Claude
- Gemini
- Perplexity
- Git / GitHub
- Real-device feedback