Inspiration
Modern life is app overload — there’s a tool for every micro-task, download for every checklist or timer, and extensive set up before you can even get to using them. We wanted to challenge that. What if you could start from nothing and build exactly what you need using your own words? That’s how Nothing App was born: a Swift-based, AI-powered interface that turns plain English into instantly usable micro-tools. Whether it’s a hiit workout with timed intervals, a one-off checklist, or a just a simple counter, Nothing App is a canvas — a “void that becomes anything.”
What it does
Nothing App lets users describe small tasks in plain English and instantly generates lightweight, functional mini-apps tailored to their intent. It's more than a productivity tool — it's a prototype for a new programming paradigm where human expression becomes software. From void to vision, users can build and use tools like timers, checklists, or counters in seconds, without ever touching code.
How we built it
We built Nothing App using three core components:
- Smart Prompt Detection – Users describe what they want, and our hybrid detection system routes requests to optimized prompts (counter, timer, checklist) or flexible creative prompts for complex combinations.
- AI Schema Generation – Using Perplexity's Sonar API with carefully tuned 300-400 token prompts, we convert user input into structured TaskSchema JSON with robust error handling and fallback mechanisms.
- Dynamic SwiftUI Rendering – The JSON schema drives a custom ComponentViewModel system that renders functional UI with real-time state management, animations, and interactive controls.
This architecture allows users to go from intent to usable interface almost instantly.
Challenges we ran into
The biggest challenge was API reliability and prompt tuning — responses often included explanatory text instead of pure JSON, requiring us to build sophisticated parsing logic with code block extraction, brace counting, and comment removal. We also had to identify and account for common problems through thorough testing with multiple sample tasks. With our wide range of use cases and dynamic rendering, we had to carefully balance between hardcoding specific logic and overgeneralizing, ensuring the system remained flexible yet reliable across varied inputs. Further, as we are relying on API calls, we worked to organizing the prompting logic to minimize the number of tokens as much as possible without compromising functionality. Prompt engineering was crucial — we discovered that 85% success rates required explicit parameter type enforcement.
Accomplishments that we're proud of
We're proud of building a schema-driven UI engine that interprets natural language and delivers functional tools in real time, but also creating a bulletproof user experience with SampleTask fallbacks that ensure 100% reliability even when AI fails. The hybrid prompt system achieves 85% API success while maintaining sub-4-second response times and efficient token usage (300-800 tokens per request).
What we learned
We learned how to:
- Build a schema-driven UI system using AI as the translator from language to layout
- Craft precise prompts and handle AI inconsistencies with multi-stage JSON parsing
- Debug complex API integrations using comprehensive logging and error recovery
- Balance token efficiency with prompt effectiveness through optimized vs. flexible prompt routing
- Embrace simplicity in component design while supporting complex combinations
What's next for The Nothing App
We're planning to expand support for more complex UI patterns and enable users to modify or chain together generated components through natural language edits like "make the break longer" or "add a water reminder." We're also exploring deep integrations with iOS ecosystem features — calendar events that auto-generate countdown timers, notification-driven habit trackers, and HealthKit integration for fitness components — moving from quick tools to fully personalized task environments that understand your schedule and patterns. These integrations require robust backend infrastructure for user data persistence, cross-device sync, and intelligent suggestions based on usage patterns. Long term, we want Nothing App to be a foundational step toward a new programming paradigm where natural language becomes the primary interface for software creation: conversational development where "my timer needs to be louder" triggers automatic audio enhancement, intent-driven architecture that learns user patterns and proactively suggests tools, and cross-platform generation where the same natural language creates native experiences across iOS, web, and desktop. This isn't just about making apps easier to use — it's about democratizing software creation itself, working toward a future where anyone can think of a digital tool and have it exist, where the barrier between human intention and functional software dissolves entirely.
Log in or sign up for Devpost to join the conversation.