Nest — Calm Planning + Savings Companion
What Inspired Nest
Nest was inspired by a simple problem: daily life feels heavier than it should.
Busy households juggle food, tasks, and money every day.
Grocery bills keep rising. Planning feels overwhelming.
Most tools either overcomplicate things or solve only one small piece.
I wanted to build something calm.
A companion that reduces cognitive load instead of adding to it.
Nest focuses on the two most exhausting loops:
- “What should I do today?”
- “Why is my grocery bill so high?”
If those two loops improve, trust follows.
What I Learned
I learned that clarity is more valuable than features.
Users don’t want dashboards.
They want direction.
They don’t want raw data.
They want easier decisions.
I also learned that AI must feel safe.
That’s why Nest uses a confirm-before-execute system:
- The assistant suggests actions.
- The user approves.
- Nothing changes without consent.
Trust compounds over time.
Mathematically, the value of Nest can be simplified as:
[ \text{Daily Stress} = f(\text{Decisions} \times \text{Uncertainty}) ]
Nest reduces both variables.
And when uncertainty drops, confidence rises.
How I Built It
Nest is built using:
- iOS (SwiftUI) for a clean, responsive mobile interface
- Next.js for backend APIs and orchestration
- Supabase (Postgres + Auth) for secure user data
- A streaming grocery comparison engine
- A multi-specialist AI assistant with strict JSON schema validation
The grocery engine streams store results progressively, so users see updates in real time instead of waiting on one large response.
The assistant validates structured outputs before executing actions to ensure safety and predictability.
Estimated savings framing:
If a user saves \$15 per grocery trip and shops weekly:
[ 15 \times 52 = 780 ]
That’s up to \$780 per year in potential impact.
Challenges I Faced
1. Grocery Data Reliability
Retail sites change frequently.
Latency and blocking create instability.
Solution:
- Caching
- Normalization
- Graceful partial streaming results
2. AI Reliability
Large language models are powerful but unpredictable.
Solution:
- Strict output schemas
- Validation layers
- Idempotent action execution
- Confirm-before-run flow
3. Avoiding Feature Overload
It’s easy to keep adding features.
It’s harder to simplify.
Nest is intentionally constrained to reduce cognitive load, not increase it.
Final Thought
Nest isn’t trying to be everything.
It’s trying to make everyday life lighter.
When planning is clear and food costs are transparent,
confidence replaces chaos.
Log in or sign up for Devpost to join the conversation.