Two phones showing the plant planning for 2026 and the task list on the home screen

2026 | Personal project

Saatbox

Garden planning app

This project shows how I use AI to prototype, build and improve a product over time.

Challenge

Plan a garden season that keeps changing

and keep track of it while it runs.

Paper seed packets spilling beans and seeds onto dark soil

Solution

Planning with seeds

and managing tasks flexibly

Learnings

Vibe coding allowed me to gain insights faster and iterate quickly.

  • Fast iteration

    Vibe coding made it easier to build, test and adjust ideas in a flow state. This noticeably sped up my design and decision process.

  • Designing with data

    Working with realistic data revealed new perspectives on flows and system design, reducing complexity through hands-on exploration.

  • Build to test

    When a working flow takes hours instead of weeks, I can put a design decision in front of real use instead of theorising about it.

Development

Workflow

Build, test, improve

Object map of sticky notes grouping objects, attributes and calls to action

Object map & user flow

I defined the main components, their attributes, actions and relationships in an object map.

From there I outlined the core user actions and a first page flow and information structure.

Wireframe sketch of the collection page beside the finished collection screen

Wireframes & mockups

I worked out layout and early flows as wireframes first, then turned them into app screens and a clickable Figma prototype to get a first feel for the product.

Learning: with the AI options available today I would invest more in clickable prototypes instead of working with mockups.

The first plant detail screen beside the reworked one with calendar, spacing and sowing data

Working Android app

I used ChatGPT and Cursor to structure epics and build them. For the first version I chose to build it in React with a simple database.

I built the design system directly in the app and turned the first wireframe look step by step into a usable app.

Board of app screens from the season test, annotated with yellow post-it notes

Testing flows in the real world

I tested the app in my own garden for a few months. It showed me which flows I needed to improve and what already worked well.

AI enabled me to make small changes instantly. Bigger concept pivots I collected until the end of the season and formed them into new epics.

The earlier task screen beside the reworked one with tasks grouped by month

Improving the app

I switched to Claude to build on bigger concept changes and extend the data catalogue.

After the test my focus moved to improving task management, making the plan more flexible and extending the 2D editor.

Context

Project type
Personal exploration project
Project duration
April 2026 - today
Tools
Miro, ChatGPT, Figma, Cursor, Claude Code, VS Code, GitHub, DaVinci Resolve
Technology
React Native, TypeScript, SQLite