- Problem
- This wasn't a problem of too few tools — it was too many. Documents here, diagrams there, math somewhere else, and prompts buried in chat history. Context broke at every handoff, and a prompt I got right one week had vanished by the next. The question I sat with: why do prompts disappear? Because they live in conversations, not in files.
- Role
- I did everything myself — problem definition, screen and data design, TypeScript implementation, release. No team, no stakeholders: every design decision's reason had to stand on my own judgment.
- Process
- The hypothesis: if prompts live in files, they become assets. Everything stays plain text (compatible with coding agents and version control), and a JSON Prompt Builder structures reusable fragments into finished results. Instead of imagining users, I validated by dogfooding — running my actual planning work through this tool alone — against one criterion: did the repeat work disappear? The temptation to add features was constant; I kept cutting against a single standard, that an unbroken flow outranks features.
- Result
- Rewriting prompts from scratch — that recurring labor — disappeared. TextViz is public and still in my daily toolkit. What it confirmed is simple: user research is allowed to start from your own pain. The problem that hurt most made the best first product.
Personal Project · TypeScript
TextViz
A unified web editor that handles Markdown, Mermaid diagrams, and LaTeX math in one workspace — with a JSON Prompt Builder that assembles reusable prompt fragments into structured results.
