A developer built a Pomodoro todo app for their Amazfit GTR4 smartwatch using Cursor (an AI coding tool that generates code from natural language descriptions). The point isn't the app itself—it's what the app represents: needs that were once ignored because "the market is too small" are, for the first time, satisfiable at low cost by a single person.

What This Is

The need was simple: the author manages tasks in TickTick, but Pomodoro reminders on the phone are too disruptive; they wanted to push tasks down to the watch for focused check-ins. The Amazfit Zepp OS app store had nothing suitable, so they decided to build it themselves.

The full workflow: set up the Zepp OS dev environment (Node.js + watch simulator) → use Cursor's prompt engineering (a prompt is the text instruction you give AI to get it working) to generate code → run it on the simulator → debug on real hardware → commit in small steps. The article publishes the complete prompt template—a hands-on AI Coding record worth referencing.

Worth noting: the author's prompts were highly disciplined—have AI produce a plan first, discuss it through, then start work; write constraints into README.md and AGENTS.md so Cursor reads the rules before each session. This "set the rules first" approach is itself a new literacy for the AI Coding era.

Industry View

Supporters see this as a sample of AI Coding democratization: a non-professional developer shipped a complete watch app. Needs that were ignored because "the market is too small, no vendor will do it" can now be met by individuals at low cost.

But the author's own war stories also draw a few boundaries worth noting:

  • They flatly wrote "AI Coding isn't as awesome as you think" and strongly recommend "commit in small steps, don't wait until everything's done"—meaning AI still can't reliably produce controllable code in one shot
  • Prompts must ask AI for a "development plan" first; otherwise AI will run wild and write whatever
  • Zepp OS is a niche platform with very little content in AI training data, so results depend heavily on the author first organizing the official docs before handing them to Cursor
  • A working project isn't a maintainable project—ongoing bug fixes and version updates still need a human

Our take: AI Coding expands what individual developers can do, but it hasn't reduced "knowing how to code" to zero. The ability to break a need down clearly and set the rules upfront—these "soft skills"—are becoming more valuable, not less.

Impact on Regular People

For enterprise IT: For small teams and edge-case internal tools, the chain "business user defines the need + AI produces the first version + IT finishes the job" is becoming viable. But code security, compliance auditing, and long-term maintenance remain unavoidable gates.

For individual careers: "Knows the business + can break down needs for AI" will be scarcer than pure coding skills. The next role companies fight over may not be the engineer, but the product or operations person who knows how to collaborate with AI.

For consumer markets: Custom apps for niche hardware like watches and smart-home devices will proliferate, but the mainstream app market won't be rewritten by this—professional teams still hold the advantage on large-scale products.