Bubble Over
A soundless calendar event alarm system
A calmer way to never miss a meeting. Gentle visual reminders that are subtle when you have time, but become impossible to ignore as the clock runs down.
Product
UX
UI
AI
Project Overview
Client: Self-launched App
Project Type: Product Design, UX/UI
Tools: Claude Code
My Role: End-to-end Designer
Time: 2 Weeks
Status: Launched, in testing







CONTEXT
A Visual Alarm Without Sound
An alarm system that gets you to your meeting on time. Bubble Over connects to your calendars and launches bubbles for each scheduled event. Through four timed intervals, they increase in intensity the closer you get to your meeting and become impossible to ignore.
PROBLEM
Why Still Use Traditional Alarms?
Traditional alarms and alerts are the go-to tools for reminders. We've accepted them as the best option, but are they really?
Alarms need to be manually set and are stressful to constantly hear.
Small notifications are easy to overlook.
People are overloaded with meetings and can accidentally miss one.
Urgency is a tedious manual setting for alarms. They do not automatically adapt.
PROCESS
AI as a Tool
For this project I completed the end-to-end process within Claude Code.
1
Research
I drew inspiration from individuals with hearing impairments or ADHD, resulting in a solution that benefits everyone, the ˝curb-cut effect˛.
2
Plan
Using Claude Codeˇs plan mode, I mapped out how to design and successfully launch the product.
3
Design/Prototype
With that foundation, I built in Claude Code and refined the UX/UI with prompts.
4
Launch/Test
By week 2, I launched a final product and put it in front of user testers for feedback.
ROADBLOCK
Design Judgment Forms a Strong Prompt
In my prototyping phase, I identified a key problem that could ruin the usability of my app.
LAUNCHED
Designed, Built, and Launched
Measuring Success
Reflection
This was my first vibe-coded project that I launched, and it wasnˇt the traditional product design process. Using AI bypassed some of the typical flow, but there are some things I would consider changing:
1
Pay closer attention to how AI is building
the features.
2
Bring wireframes back into the process
AI gave me polished screens fast, but I landed on a solid answer without exploring the alternatives. I never got to ask 'what about a layout like this?'
3
Work within the target coding language faster
Refining within HTML and converting into Swift was more of a setback than it was helpful. I spent days fixing what broke in translation.