StepOut: The Invisible Mission - An OpenSource AI That Gets You Off Your Screen

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 1: Touch Grass.

What I Built

StepOut — The Invisible Mission: AI designs the experience. You put your phone away. Real life becomes the experience.

StepOut is a different kind of AI-powered outdoor experience. Instead of keeping users engaged with another app, it uses AI to give them a reason to leave their screens behind.

The idea is simple: a CAPTCHA you can't solve from your screen.

Users create a personalized physical mission based on their available time, surroundings, preferred experience, energy, and accessibility constraints. StepOut generates a real-world challenge that requires observation, listening, exploration, or reflection rather than interacting with a device.

The experience follows four steps:

• Create a Mission: Choose a duration, surroundings, experience style, and physical or accessibility constraints. Missions can be tailored to places such as your home, neighbourhood, park, or urban surroundings.
• Disappear: Receive a concise mission briefing with clear steps and an observation goal. Lock your phone and experience the world around you.
• Return & Reflect: Answer a mission-specific question, followed by an adaptive AI-generated follow-up that explores what you actually noticed.
• Read Your Field Report: Get a grounded insight synthesized from your own observations, a personal takeaway, and an entry in your Field Journal Archive.

What makes StepOut different?

• Context-aware AI missions: Challenges adapt to the user's circumstances instead of relying entirely on a fixed list of activities.
• Phone-free by design: Missions don't require photography, GPS, continuous phone interaction, purchases, or dangerous navigation.
• Adaptive reflection: AI asks relevant follow-up questions based on the mission and the user's actual observations, rather than generic mood questions.
• Honest by design: StepOut doesn't pretend that a timer proves someone went outside. Time is reported using elapsed timestamps or user-provided information, without surveillance.
• Privacy-conscious experience: No camera or GPS tracking is needed to verify the experience.
• Graceful AI fallback: When the AI service is unavailable, deterministic fallback logic keeps the experience usable and clearly identifies fallback-generated content.
• Persistent missions: Mission state survives refreshes, browser closure, and phone locking through timestamp-based session persistence.
• Field Journal: Users can revisit completed missions, observations, AI insights, and personal takeaways.

StepOut is built for people who want to spend less time scrolling and more time noticing the world around them. It doesn't try to make people addicted to another app; it helps them leave it.

Demo

Live Demo: https://stepout-1vha.onrender.com

The demo will show the complete journey: creating a personalized mission, receiving the briefing, putting the phone away, returning for an adaptive reflection, and viewing the final Field Report.

Code

GitHub Repository: https://github.com/ankit9241/StepOut

Built with TanStack Start, React 19, TypeScript, Tailwind CSS, and an open-weight Gemma model served through Ollama.

How I Built It

StepOut is built around Gemma, an open-weight model from Google, with Ollama providing the model-serving interface.

Rather than using AI as a decorative chatbot, the project uses it for specific tasks throughout the experience:

1. Mission generation: Gemma creates contextual, actionable, real-world missions based on the user's available time, surroundings, preferences, and constraints.
2. Adaptive inquiry: The AI generates a follow-up question grounded in the user's first reflection, encouraging more meaningful observation.
3. Insight synthesis: The AI turns the user's actual observations into a concise, personalized field note and takeaway without inventing details about their experience.

Technical architecture

• Frontend: TanStack Start, React 19, TypeScript, and Tailwind CSS.
• AI inference: Open-weight Gemma through Ollama.
• Backend: Node.js server with API endpoints for mission generation and reflection processing.
• State persistence: A centralized session layer with local storage and timestamp-based session recovery.
• Resilience: Deterministic fallback generation and reflection logic for AI outages.
• Testing: Automated unit and integration tests covering AI services, API endpoints, fallback behavior, routing, security, and database functionality.
• Deployment architecture: Designed for a unified Render web service, with the AI inference service kept separate from the web application.

The interface follows an editorial field-journal aesthetic rather than a conventional gamified productivity dashboard. It uses Instrument Serif and Geist typography, a warm cream background (#F7F4EC), and a restrained burnt-vermilion accent (#C84B31).

I also built explicit guardrails into the mission-generation process. A challenge should fit the user's real circumstances, remain accessible and safe, and never require them to keep looking at their phone to complete it.

Engineering verification: The latest implementation passed TypeScript checks, linting, production builds, and all 36 automated tests across six test suites.

Why Does Open Innovation Matter?

StepOut's purpose would be undermined if escaping screens required dependence on a closed, opaque AI service.

Using open-weight Gemma through Ollama makes the AI layer more inspectable, gives developers flexibility over model serving, and creates a path toward running inference locally for greater privacy and control.

Open innovation also makes the project's underlying approach reproducible. Developers can examine the prompts, experiment with different models, improve the fallback mechanisms, and contribute better ways of generating safe, meaningful offline experiences.

Most importantly, the model is not the product by itself. The value comes from how the model is integrated into a complete experience that encourages people to disconnect.

StepOut demonstrates that AI doesn't always need to maximize engagement, session length, or screen time. It can be used to help people spend less time with technology and more time in the physical world.

My Agent Session

[Insert DevRelay agent session URL, if available]

Prize Categories

• Overall Hacktoberfest Open-Source AI Challenge
• Best Use of Gemma
• Best Use of Render