The manifesto
The hackathon opened with a manifesto that shaped how I thought about the project.
A few positions stuck with me:
Static is a tool limit, not a law
We designed identical screens for everyone because our tools could only draw one. That constraint is changing.
The UI should learn the user
An interface can reshape itself around a person's context and intent instead of asking the person to adapt to fixed software.
Generation needs judgement
A UI that changes without purpose is just noise. Every change has to earn its place by being clearer, faster or more useful than what came before.
Constraints are the genome
You're not just designing individual screens.
You're designing the rules and guardrails that generate them. The brief gave me a few concrete requirements to keep coming back to:
Show at least two distinct generations of the same UI
Keep the human in control of adaptation
Make adaptation a proposal, not a trap
Be transparent about what is mocked or hard-coded versus what is real
The problem
Save ≠ Act
I save exercise Reels because I want to try them but saving doesn't guarantee I'll act.
The problem isn't a lack of fitness content. It's the gap between:
“I want to do this.”
and
“I actually did it.”
I wondered:
Could a UI turn that existing intention into a small, achievable action — and then learn what helps me follow through?
Figure 1: Storyboard of the problem and potential solution
Exploring the ideas
I sketched three directions before deciding what was worth building.
Idea 01: Save a Reel → Generate a Workout Video
Input: Saved Reel (some Reels contain multiple exercises, others just one)
System: Understands the exercise/content → generates a complete workout
Output: A personalised workout video
GenUI question: Can a saved piece of inspiration become a generated experience rather than just saved content?
What it explores: Generative content + UI.
Figure 2: Idea 1 save a reel to generate a full lenghth workout video
Idea 02 — Goal → Voice Note → Analyse → Generate Plan
Input: “I want to be able to do 8 pull-ups.”
System: Analyses the goal, potentially context/preferences → generates a structured plan
Output: A personalised list
GenUI question: What happens when the interface is generated from intent rather than from a predefined flow?
What it explores: intent → interpretation → generated interface/content.
Figure 3: Idea 2 Set goal via voice note to generate a workout plan
Idea 03 — Form Corrector
Input: Voice note / exercise selection
System: Sets up camera → observes movement → analyses form
Output: Live audio guidance
For example:“Keep your chest up.”
“Slow down the descent.”
“You’re shifting your weight to the left.”
GenUI question:Can the interface adapt itself in real time to what the person is doing?
What it explores: real-time contextual adaptation rather than just one-time generation.
Figure 4: Set up phone for voice form correction feedback.
