DEV Community: Sarvesh Dhaigude

I Built an AI That Wants You to Stop Using It 🌿 | GrassQuest AI

Sarvesh Dhaigude — Sun, 11 Oct 2026 17:47:30 +0000

GrassQuest AI: I Built an Open-Source AI That Wants You to Stop Using It 🌿

Less Scrolling. More Exploring.

The Problem

We have AI tools that help us write faster, code faster, and consume more information. But what if we built an AI application designed to make us close our screens?

That question inspired GrassQuest AI.

For the Touch Grass challenge, I wanted to explore a different relationship between people and artificial intelligence: one where technology becomes a starting point for real-world experiences rather than another source of endless engagement.

Introducing GrassQuest AI

GrassQuest is an open-source AI-powered outdoor adventure companion.

Users enter their available time, mood, interests, preferred environment, and difficulty level.

An open-weight AI model transforms those preferences into a personalized outdoor mission.

The application then encourages users to put their phones away, complete their mission, and return only to record their experience.

How I Built It

The application uses React, TypeScript, and Tailwind CSS for its frontend, with Python and FastAPI handling AI requests.

Its intelligence comes from Google's Gemma 3 model running locally through Ollama.

The AI generates structured outdoor missions, which the backend validates before displaying them to the user.

The project also includes a distraction-free mission experience, local mission storage, a nature journal, and an outdoor activity dashboard.

Why Open Innovation Matters

I chose open-weight AI because I wanted the project's intelligence to be something users could run and control themselves.

Rather than making every request dependent on a proprietary cloud AI service, GrassQuest supports local inference.

This approach creates opportunities for privacy, model customization, experimentation, and reduced dependency on third-party infrastructure.

It also means developers can inspect the application, modify its mission-generation logic, and experiment with different compatible models.

The goal is not simply to use open AI because the challenge requires it. The goal is to make local AI central to the product's purpose.

The Touch Grass Philosophy

GrassQuest is designed around a simple idea:

The best session is the one that ends with someone going outside.

Instead of encouraging endless interaction, it gives users a mission and then gets out of their way.

Testing It in the Real World

[Insert your actual outdoor testing experience.]

[Describe the mission Gemma generated.]

[Include what worked, what surprised you, and what you would improve.]

[Add authentic screenshots and outdoor demonstration photographs.]

Technical Challenges and Lessons

[Document a genuine challenge encountered while implementing or running local AI.]

[Explain how you investigated and resolved it.]

[Include an honest observation about local inference speed, model output quality, or resource usage.]

What's Next?

I want to explore richer nature-related missions, better accessibility, and additional local AI capabilities while keeping GrassQuest lightweight and privacy-conscious.

The long-term vision is simple: make AI a tool for discovering more of the world, not escaping further into a screen.

Project Links

👉 http://localhost:5173/ Just check it out once........

My GrassQUEST AI will really help you to get OFF THE SCREEN

I Built an AI That Wants You to Get OFF The Screen ,Using It 🌿 | GrassQuest AI

Sarvesh Dhaigude — Sun, 11 Oct 2026 17:28:59 +0000

GrassQuest AI: I Built an Open-Source AI That Wants You to Stop Using It 🌿

Less Scrolling. More Exploring.

The Problem

We have AI tools that help us write faster, code faster, and consume more information. But what if we built an AI application designed to make us close our screens?

That question inspired GrassQuest AI.

Introducing GrassQuest AI

GrassQuest is an open-source AI-powered outdoor adventure companion.

Users enter their available time, mood, interests, preferred environment, and difficulty level.

An open-weight AI model transforms those preferences into a personalized outdoor mission.

The application then encourages users to put their phones away, complete their mission, and return only to record their experience.

How I Built It

The application uses React, TypeScript, and Tailwind CSS for its frontend, with Python and FastAPI handling AI requests.

Its intelligence comes from Google's Gemma 3 model running locally through Ollama.

The AI generates structured outdoor missions, which the backend validates before displaying them to the user.

The project also includes a distraction-free mission experience, local mission storage, a nature journal, and an outdoor activity dashboard.

Why Open Innovation Matters

I chose open-weight AI because I wanted the project's intelligence to be something users could run and control themselves.

Rather than making every request dependent on a proprietary cloud AI service, GrassQuest supports local inference.

This approach creates opportunities for privacy, model customization, experimentation, and reduced dependency on third-party infrastructure.

It also means developers can inspect the application, modify its mission-generation logic, and experiment with different compatible models.

The goal is not simply to use open AI because the challenge requires it. The goal is to make local AI central to the product's purpose.

The Touch Grass Philosophy

GrassQuest is designed around a simple idea:

The best session is the one that ends with someone going outside.

Instead of encouraging endless interaction, it gives users a mission and then gets out of their way.

Testing It in the Real World

[Insert your actual outdoor testing experience.]

[Describe the mission Gemma generated.]

[Include what worked, what surprised you, and what you would improve.]

[Add authentic screenshots and outdoor demonstration photographs.]

Technical Challenges and Lessons

[Document a genuine challenge encountered while implementing or running local AI.]

[Explain how you investigated and resolved it.]

[Include an honest observation about local inference speed, model output quality, or resource usage.]

What's Next?

I want to explore richer nature-related missions, better accessibility, and additional local AI capabilities while keeping GrassQuest lightweight and privacy-conscious.

The long-term vision is simple: make AI a tool for discovering more of the world, not escaping further into a screen.

Project Links

project local host link : check out -> http://localhost:5173/

Built for the Touch Grass open-source AI challenge.

Close the tab. Start the adventure. 🌱

StudyMate — A Private AI Study Companion Built for a Friend 🤝

Sarvesh Dhaigude — Mon, 05 Oct 2026 06:50:12 +0000

This is a submission for the

This is my submission for the Hacktoberfest Weekend Challenge: Build for a Friend.

🤝 StudyMate — A Private AI Study Companion Built for a Friend

What I Built

I built StudyMate, a private AI-powered study companion designed for a college student who wants help understanding difficult concepts, organizing study material, revising through quizzes, and planning study sessions.

Instead of building another general-purpose AI chatbot, I wanted to create something more focused: a study companion that feels like a personal buddy while keeping privacy and open/local-first AI at the center.

The idea behind StudyMate is simple:

Your study buddy. Your data. Your device.

👤 Who I Built It For

I built StudyMate for a college student who sometimes struggles with organizing study material, understanding difficult topics, revising effectively, and staying consistent with study sessions.

Students often have notes, topics, assignments, and revision tasks spread across different places. I wanted to bring some of these study activities together into one simple and student-friendly experience.

The goal wasn't to build something huge.

The goal was to build something that could actually be useful to one real student.

🚀 What StudyMate Can Do

StudyMate is designed to provide several useful study features:

📚 Organize and work with study material
🧠 Help explain difficult concepts
❓ Generate quizzes for revision and self-testing
📅 Help plan study sessions
🤖 Provide AI-powered study assistance
🔒 Focus on privacy and a local-first approach
💻 Provide a simple and student-friendly interface

The idea is to make studying more interactive instead of simply reading notes again and again.

🔐 Why Private / Local-First AI?

Privacy was one of the main reasons I wanted to explore an open/local-first AI approach.

Students may work with personal notes, study plans, assignments, and other information that they may not always want to send to a remote service.

With open-weight and local-first AI approaches, developers have more control over where AI runs, which models are used, and how the system can be adapted.

For StudyMate, this makes the idea of a private study companion much more meaningful.

Instead of treating AI as a completely closed black box, I wanted to explore a direction where the technology can be more transparent, customizable, and closer to the user.

🛠️ How I Built It

StudyMate was developed as a modern web application using:

TypeScript
React
Vite
Tailwind CSS
Open-weight / local-first AI architecture
GitHub for source-code management
Bolt.new for AI-assisted development and deployment

I used AI-assisted development to move from the initial idea to a working application quickly, while still thinking about the product structure, user experience, privacy, and the problem being solved.

The project is publicly available on GitHub so that the implementation can be explored.

🌐 Live Demo

👉 Live Demo: https://studymate-private-ai-pt0e.bolt.host

You can open the live application and explore the StudyMate interface.

💻 Source Code

👉 GitHub Repository: https://github.com/SarveshDhaigude/StudyMate-Private-AI

The repository contains the source code and project configuration used to build StudyMate.

🧠 What I Learned

This challenge taught me that building a useful project doesn't always mean building something massive.

The most important part was starting with a real person's problem and then deciding what the simplest useful solution could look like.

During this project, I learned more about:

Building a complete web application from an idea
Using AI-assisted development tools effectively
Structuring a React and TypeScript project
Connecting a project with GitHub
Publishing a working web application
Thinking about privacy when designing AI-powered applications
Exploring open-weight and local-first AI concepts
Designing around an actual user instead of an imaginary audience

🌍 Why Open Innovation Matters

Open innovation matters because developers should have more control over the technology they build with.

For an AI study companion, this is especially important because the application may interact with personal learning material and study information.

An open-weight or local-first approach can provide opportunities to:

Run models closer to the user
Experiment with different models
Customize how the AI behaves
Have greater control over data
Reduce dependence on a single closed AI provider

For StudyMate, the open/local-first direction isn't just a technical choice.

It supports the main idea of the project:

A study companion should feel personal, useful, and respectful of the student's data.

❤️ Why I Built It

The Hacktoberfest theme was Build for a Friend, and that changed how I approached the project.

Instead of asking:

"What application can I build?"

I asked:

"What small problem could I solve for someone I know?"

That led me toward education and studying.

College students already have enough distractions and disconnected tools. I wanted StudyMate to be a simple companion that could help make studying a little more organized and interactive.

🎯 Final Thoughts

StudyMate started with a simple idea:

Build something useful for one real person.

It may not solve every problem a student has, but it demonstrates how open AI, modern web technologies, and AI-assisted development can come together to create a focused product around a real need.

This challenge also reminded me that good software doesn't always begin with a huge idea.

Sometimes it begins with one person, one problem, and the decision to build something that might make their day a little easier.

Thank you to the DEV Community and Hacktoberfest team for organizing this challenge. 🚀

🔗 Project Links

Live Demo: https://studymate-private-ai-pt0e.bolt.host

GitHub: https://github.com/SarveshDhaigude/StudyMate-Private-AI