Sandbox

These are experiments, half-finished tools, and weird ideas that helped me explore new interfaces, logic flows, and AI-powered tools.

These are experiments, half-finished tools, and weird ideas that helped me explore new ideas.

Check out my case studies →

Digital Quilt

Working! | Collaborative Art Tool | Replit

Test the app

→

Test the app

→

Test the app

→

Goal

Design a collaborative drawing interface where users could add unique “squares” to a shared digital quilt, like a crowd-made pixel mural, but softer and more playful.

Struggles

Tried multiple AI/no-code platforms to support async contribution and data persistence. Lovable wouldn’t connect reliably to Supabase, and Replit had layout issues—so I simplified the drawing interactions and re-scoped the upload logic based on platform limits.

Lessons Learned

Not every prototype needs to be polished, what matters is clarity of intent. I learned to work around technical limits, choose the right tools for the problem, and keep iterating without getting stuck.

Version 1: Replit build
What I explored: Structured guidance using geolocation, USDA zone data, and Farmer’s Almanac API integration.
What I learned: The flow felt calm and predictable, but too rigid for casual users. It needed more flexibility for one-off tasks.
Version 2: Lovable build
What I explored: Weather-responsive guidance and stronger plant database integration using Lovable’s AI prompting.
What I learned: This version felt more conversational and adaptable, with better suggestion quality, but it lost some of the long-term planning clarity from V1.
Version 3: Plants-first
What I explored: After focusing on time- and data-driven flows, I shifted to a plant-first model. Users started by creating their garden visually, then accessed guidance through a lightweight “Guides” section.
What I learned: This version felt the most personal and visual, great for casual users, but lacked the proactive momentum and nudging of earlier versions.
Version 4: Onboarding flow
What I explored: A guided onboarding experience to help users clarify their goals (garden type, vibe, preferences) and give the assistant the data it needs for better recommendations.
What I learned: This flow helped users feel more confident about where to begin and gave the system a strong foundation. It pairs well with timeline or suggestion-based models as a first-time entry point.
Version 1: Replit build
What I explored: Structured guidance using geolocation, USDA zone data, and Farmer’s Almanac API integration.
What I learned: The flow felt calm and predictable, but too rigid for casual users. It needed more flexibility for one-off tasks.
Version 2: Lovable build
What I explored: Weather-responsive guidance and stronger plant database integration using Lovable’s AI prompting.
What I learned: This version felt more conversational and adaptable, with better suggestion quality, but it lost some of the long-term planning clarity from V1.
Version 3: Plants-first
What I explored: After focusing on time- and data-driven flows, I shifted to a plant-first model. Users started by creating their garden visually, then accessed guidance through a lightweight “Guides” section.
What I learned: This version felt the most personal and visual, great for casual users, but lacked the proactive momentum and nudging of earlier versions.
Version 4: Onboarding flow
What I explored: A guided onboarding experience to help users clarify their goals (garden type, vibe, preferences) and give the assistant the data it needs for better recommendations.
What I learned: This flow helped users feel more confident about where to begin and gave the system a strong foundation. It pairs well with timeline or suggestion-based models as a first-time entry point.
Version 1: Replit build
What I explored: Structured guidance using geolocation, USDA zone data, and Farmer’s Almanac API integration.
What I learned: The flow felt calm and predictable, but too rigid for casual users. It needed more flexibility for one-off tasks.
Version 2: Lovable build
What I explored: Weather-responsive guidance and stronger plant database integration using Lovable’s AI prompting.
What I learned: This version felt more conversational and adaptable, with better suggestion quality, but it lost some of the long-term planning clarity from V1.
Version 3: Plants-first
What I explored: After focusing on time- and data-driven flows, I shifted to a plant-first model. Users started by creating their garden visually, then accessed guidance through a lightweight “Guides” section.
What I learned: This version felt the most personal and visual, great for casual users, but lacked the proactive momentum and nudging of earlier versions.
Version 4: Onboarding flow
What I explored: A guided onboarding experience to help users clarify their goals (garden type, vibe, preferences) and give the assistant the data it needs for better recommendations.
What I learned: This flow helped users feel more confident about where to begin and gave the system a strong foundation. It pairs well with timeline or suggestion-based models as a first-time entry point.
Version 1: Replit build
What I explored: Structured guidance using geolocation, USDA zone data, and Farmer’s Almanac API integration.
What I learned: The flow felt calm and predictable, but too rigid for casual users. It needed more flexibility for one-off tasks.
Version 2: Lovable build
What I explored: Weather-responsive guidance and stronger plant database integration using Lovable’s AI prompting.
What I learned: This version felt more conversational and adaptable, with better suggestion quality, but it lost some of the long-term planning clarity from V1.
Version 3: Plants-first
What I explored: After focusing on time- and data-driven flows, I shifted to a plant-first model. Users started by creating their garden visually, then accessed guidance through a lightweight “Guides” section.
What I learned: This version felt the most personal and visual, great for casual users, but lacked the proactive momentum and nudging of earlier versions.
Version 4: Onboarding flow
What I explored: A guided onboarding experience to help users clarify their goals (garden type, vibe, preferences) and give the assistant the data it needs for better recommendations.
What I learned: This flow helped users feel more confident about where to begin and gave the system a strong foundation. It pairs well with timeline or suggestion-based models as a first-time entry point.

Garden Helper

Exploratory Builds | Dashboard | Replit + Lovable

Goal

Design a gardening assistant to help casual users feel more confident planning, planting, and caring for their garden, using geolocation, weather, plant data, and seasonal insights.

Struggles

Tested four interaction models across multiple tools. Replit was great for structured logic, while Lovable offered stronger plant database access. Each version required trade-offs in flexibility, user initiative, and onboarding clarity.

Lessons Learned

Different users need different kinds of support. Some want structured planning, others just want a gentle nudge. Designing for both meant exploring multiple flows and learning how to balance user control with helpful automation.

← Check out my case studies