I Built an AI That Turns “I’m Bored” Into Real-World Side Quests 🌿

lovely sharma — Fri, 09 Oct 2026 14:12:23 +0000

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

Your neighborhood has side quests. Go unlock one. 🌿

What I Built

POV: You open your phone to check one notification. An hour later, you're watching a raccoon steal someone's lunch and wondering where your life went.

We've all been there. 💀

I wanted to build something that uses AI to get people off their screens instead of giving them another reason to stare at one.

Meet OUTLAND: TouchGrass, an AI-powered IRL side quest generator.

You tell it your mood, energy level, available time, budget, social preference, and how chaotic you're feeling. It turns that information into a small adventure you can actually do outside.

Think of those days when you want to go somewhere, do something, touch grass, become the main character for five minutes, but your brain has absolutely no plans.

That's where OUTLAND comes in.

The goal isn't to gamify every second of your life. No streak anxiety. No guilt for staying home when you need a rest. Just a little nudge toward doing something different.

Demo

🌱 Live app: https://outland-touchgrass.onrender.com/

Open the app, choose your current state, and generate a quest. When location information is available, OUTLAND can discover real places and ground the quest in a location that passes its checks.

During testing, it generated a quest called THE GREEN LUNG AUDIT at Leisure Valley Park. Instead of simply saying “go for a walk,” it framed visiting a green space as a quest with a time estimate, difficulty, and XP.

It's a small twist, but that's the idea. You don't need to book a trip, spend a lot of money, or coordinate six people's calendars just to do something different.

Give it a try when you're bored, restless, or tired of doing the same thing every weekend.

Your next side quest might be closer than you think.

Code

💻 Open-source repository: https://github.com/LovelySharma-dev/Outland---TouchGrass

The code is open for you to explore, learn from, and contribute to.

If you have ideas for better quests or ways to improve the experience, I'd love to hear them.

How I Built It

The stack is Node.js, Express, JavaScript, Gemma, and SerpApi.

Here's how the pieces work together.

📍 SerpApi discovers real places

I didn't want the AI inventing a park with a suspiciously convenient name. SerpApi helps discover actual places so the generation pipeline has real location data to work with.

🧠 Gemma generates the quest

The app combines structured user preferences with relevant place information. Gemma then generates a quest intended to fit the user's time, budget, energy, and social preferences.

🛡️ The app checks the result

A response isn't automatically a good quest just because an AI generated it.

OUTLAND checks safety, location grounding, quality, and personalization. Weak results can be rejected, with a bounded retry path.

🧺 When AI has a bad day, there is a fallback

Free inference can be unreliable. Sometimes the model produces a generic quest, and sometimes the inference route is unavailable.

OUTLAND includes a curated fallback pool for those situations. Fallback quests are labelled honestly rather than pretending they came from the model.

I'd rather show a useful fallback than serve AI-generated nonsense with confidence.

🧪 What I learned

The biggest lesson was that getting AI to generate text is one thing. Getting it to generate something a person might genuinely want to do is another.

A quest can sound creative but still be too vague, ignore the user's energy level, or fail to point to a useful real-world destination.

Free inference adds another challenge because model availability and response quality aren't always consistent.

So I had to build around the model instead of treating the prompt as magic. Quality checks, bounded retries, location validation, and fallback behavior became important parts of the application.

The takeaway? A clever AI demo and a useful AI product are not the same thing.

The model can suggest an adventure. The system still has to make sure that adventure makes sense.

Why Does Open Innovation Matter?

I wanted to explore an AI use case beyond another chatbot or productivity assistant.

Gemma powers quest generation, while SerpApi helps connect generated ideas to real places. The surrounding application handles validation, personalization, and failure cases.

Using an open-weight model makes the generation layer more open to experimentation. Developers can inspect the project, change prompts, improve validation, and explore alternative model setups.

I'm not claiming the whole app runs locally or without an internet connection. Instead, this project explores how open AI building blocks can power a practical experience that encourages people to spend less time on their screens.

For me, the exciting part is not just what the model can generate. It's what we can build around it.

🏆 Prize Categories

I'm entering these partner categories for the Hacktoberfest Open-Source AI Challenge:

Best Use of Gemma: Gemma powers OUTLAND's AI-generated side quests, turning user preferences and real-world place information into personalized outdoor activities.
Best Use of SerpApi: SerpApi helps OUTLAND discover real-world places, grounding quests in actual locations instead of made-up destinations.
Best Use of Render: OUTLAND is deployed on Render, hosting the live application so people can try generating their own outdoor side quests.
Best Use of MongoDB Atlas: OUTLAND uses MongoDB Atlas for its data layer.

👤 Team Submission

Solo submission: Built by @lovely_puff.

🌿 Your Turn

OUTLAND was built for Hacktoberfest Open-Source AI Challenge Week 1: Touch Grass.

If you try it, tell me what quest you get. Did it give you a reason to explore somewhere new, or did you immediately return to scrolling?

Be honest. 😂

🌱 Try OUTLAND: https://outland-touchgrass.onrender.com/

💻 Explore the code: https://github.com/LovelySharma-dev/Outland---TouchGrass

🔗 Connect with me on LinkedIn: https://www.linkedin.com/in/lovely-sharma-dev/

If you're into AI, open source, or building fun things that solve real problems, let's connect.