DEV Community: Aarav yadav

Phantom AI: A Ghost in the Machine

Aarav yadav — Tue, 27 Oct 2026 08:34:04 +0000

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 4: Ghost in the Machine

What I Built

👻 Phantom AI — A Ghost in the Machine

I built Phantom AI, a spooky interactive terminal inspired by Halloween, mysterious computer messages, and the idea of a ghost hiding inside a machine.

Users can enter commands such as help, story, haunt, and whoami to explore mysterious messages and generate spooky stories. I built the project using HTML, CSS, and JavaScript, with an optional integration for running an open-weight AI model locally through Ollama.

My goal was to combine a fun Halloween experience with the idea that AI can be more private, customizable, and accessible when it runs on your own computer.

Demo

Run locally: Open index.html in your browser to try the offline story mode.

GitHub repository:

https://github.com/aaravyadav192007-crypto/Ghost-in-the-Machine

Live demo: Add your GitHub Pages URL here after deployment.

Code

🔗 View the Phantom AI source code on GitHub

The project uses three main files:

index.html — The terminal interface and website structure.
style.css — The spooky theme, neon effects, animations, and responsive layout.
script.js — Interactive terminal commands, story generation, and optional local AI integration.

How I Built It

I used HTML, CSS, and vanilla JavaScript to build a responsive, haunted-terminal experience.

The project has two modes:

Offline story mode: Uses built-in spooky responses and stories without requiring an API key or internet connection after the files are available.
Local AI mode: Can connect to Ollama running on the user's computer with a compatible open-weight model, such as Llama 3.2, to generate original responses.

This approach let me build the interface with familiar web technologies while exploring how local AI can make the experience more interactive.

Why Does Open Innovation Matter?

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

Phantom AI explores three important benefits:

Privacy: With local inference, prompts can remain on the user's computer instead of being sent to a third-party AI provider.
Customization: Developers can experiment with different compatible open-weight models and change the system prompt to shape the AI's behavior.
Accessibility and independence: The built-in story mode works without a paid AI API, and local inference can avoid per-request API charges.

Unlike a closed, hosted AI API, a locally running model gives developers more control over model selection and how their application uses AI. Local AI still requires installation, suitable hardware, and an initial model download.

The offline story mode is available even when local AI is not configured.

The Spirit of It 👻

What if an abandoned computer still had one last story to tell?

That question inspired Phantom AI. The terminal creates a playful feeling that something forgotten is waiting to communicate. Beneath the spooky atmosphere is a more serious idea: open technologies help us build experiences that we can customize, understand, and keep running on our own machines.

Sometimes, bringing something back to life means giving an old idea a new purpose.

Prize Categories

Open-source AI / local inference — where applicable.

Final Thoughts

Phantom AI is my experiment in combining creative web development with the possibilities of open-weight AI. It is a small project, but it demonstrates how accessible web technologies and locally run AI can work together to create something fun, customizable, and privacy-conscious.

Built with HTML, CSS, JavaScript, and a little bit of mystery. 🕯️

LazyMate AI 🦥 – Your Personal Lazy Assistant | Work Less, Live More

Aarav yadav — Sun, 25 Oct 2026 12:23:23 +0000

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 3: Slow and Steady

What I Built

I built LazyMate AI 🦥, a personal productivity assistant designed to help people save time and effort by letting AI handle everyday tasks.

LazyMate AI includes three main features:

📝 Smart Summarizer: Converts lengthy notes into short, easy-to-understand summaries.
✅ Lazy Task Planner: Organizes tasks into simple, manageable steps.
🌙 Relax Mode: Generates calming break ideas and short bedtime stories.

The idea behind LazyMate AI is simple: work smarter, reduce unnecessary effort, and make more time for the things that matter.

It is especially useful for students who want to organize their study notes, plan their day, and take a break without overcomplicating everything.

Demo

Project: LazyMate AI — Your Personal Lazy Assistant

Local Demo: Open index.html in your browser through VS Code Live Server after setting up Ollama.

Live Demo: A public deployment link is not available yet.

Code

Check out the source code on GitHub:

LazyMate AI — GitHub Repository

The project is built using HTML, CSS, and JavaScript, with a local AI integration.

How I Built It

I built LazyMate AI using three core web technologies:

HTML: Creates the structure of the application and its three productivity tools.
CSS: Provides a clean, responsive interface with a relaxing green theme.
JavaScript: Handles user input, communicates with the AI model, and displays the generated results.

For the AI component, the project is designed to use Ollama with the Llama 3.2 3B model, running locally on the user's computer.

Instead of relying on a hosted AI API, the application sends prompts to the local Ollama service. Custom system prompts guide the model to summarize notes, organize tasks, or generate relaxing content.

This approach makes the project a practical example of combining a simple web interface with locally running AI.

Why Does Open Innovation Matter?

Open innovation makes LazyMate AI more flexible, accessible, and privacy-conscious.

1. Local AI inference

Once the model is downloaded, AI generation can work without an internet connection, provided the local service is running and the computer has sufficient resources.

2. Better control over personal data

Users can send their prompts to their own local AI service instead of a remote AI provider. This is particularly useful when working with personal notes or study material.

3. Freedom to experiment

The model and prompts can be changed to experiment with different responses and workflows.

4. Lower recurring costs

Running a local model avoids per-request charges from a hosted AI API, although hardware, electricity, and setup still have costs.

5. Learning by building

As a student learning web development, I wanted to understand how AI can be integrated into a real application rather than treating it as a black box.

Open innovation makes it possible to experiment, customize, and learn without depending entirely on a closed AI platform.

My Agent Session

I developed LazyMate AI by working through the application structure, interface design, JavaScript functionality, and local AI integration.

I have not included a DevRelay agent-session link because I do not have a saved session URL to share.

The Lazy Payoff

LazyMate AI aims to reduce the small, repetitive tasks that take time throughout the day.

Instead of manually shortening lengthy notes, students can request a summary. Instead of figuring out where to start with a long task list, they can generate a manageable plan.

The goal is not to avoid meaningful work. It is to spend less effort on repetitive work and more time on learning, creating, and relaxing.

Prize Categories

Main challenge: Hacktoberfest Open-Source AI Challenge — Week 3: Slow and Steady.

I am submitting LazyMate AI as a local-AI productivity tool. I have not listed additional partner prize categories because I have not verified their eligibility.

Built with HTML, CSS, JavaScript, and local open-weight AI.

*LazyMate AI 🦥 — Let the machine do the busywork, while you take it slow.

GreenQuest AI: Turning Open-Source AI into Real-World Adventures | #Hacktoberfest #TouchGrass

Aarav yadav — Fri, 09 Oct 2026 11:20:49 +0000

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

What I Built

I built GreenQuest AI 🌿, an AI-powered outdoor activity companion that encourages people to spend less time scrolling on their phones and more time exploring nature.

Users can select their mood, available time, and interests to generate personalized outdoor quests. Activities include exploring plants and trees, observing birds, walking outdoors, and gardening.

Key Features

🤖 AI-powered nature quest generation
🌳 Personalized outdoor activities
🏆 Points for completing quests
💾 Progress saved using browser local storage
📱 Responsive design for different screen sizes
🌱 Built-in fallback activities if AI generation fails

The goal is to use technology to encourage real-world experiences rather than endless screen time.

Demo

🌐 Live Website: https://aaravyadav192007-crypto.github.io/GreenQuest-AI/

Try GreenQuest AI and generate your own nature quest!

Code

💻 GitHub Repository: https://github.com/aaravyadav192007-crypto/GreenQuest-AI

The project source code is available on GitHub.

How I Built It

I built GreenQuest AI using HTML, CSS, and JavaScript, with Hugging Face Transformers.js for browser-based AI inference.

Technologies used:

HTML5 for the website structure
CSS3 for styling, animations, and responsive design
JavaScript for interactivity and application logic
Hugging Face Transformers.js for running AI models in the browser
Qwen2.5-0.5B-Instruct as the intended open-weight language model for generating nature quests

The user selects their mood, available time, and interests. JavaScript creates a prompt, the AI generates a nature quest, and the website displays the result. Users can then complete the quest and earn points.

The application is designed to run AI inference in the browser without requiring a paid AI API key. An internet connection is needed to download the model initially unless the required files have already been cached or hosted locally.

Why Does Open Innovation Matter?

Open innovation makes it easier for students and developers to experiment with AI, customize applications, and learn without depending entirely on proprietary services.

For GreenQuest AI, an open-weight model offers several advantages:

Flexibility: Developers can experiment with compatible models and customize prompts.
Cost efficiency: No paid AI API is required for each quest-generation request.
Privacy: User preferences can be processed locally in the browser.
Local inference: The model can generate activities on the user's device after its files are available.
Learning: The project demonstrates how open-weight AI can be integrated into a practical web application.

I believe AI should not only keep people engaged with screens. It can also encourage them to disconnect, explore their surroundings, and reconnect with nature.

My Agent Session

No agent-session recording is included yet. I may add a DevRelay session link after recording one.

Prize Categories

Touch Grass — Week 1

What's Next?

🌿 Plant identification using images
🐦 Bird identification and nature challenges
🗺️ Outdoor activity recommendations based on location
🏅 Daily challenges, achievements, and badges
📶 Improved offline functionality

This project helped me explore how open-weight AI can be combined with frontend development to create something useful beyond the screen.

Built by Aarav Yadav 🌿

🔗 Live Demo | GitHub Repository

Hacktoberfest #OpenSource #AI #JavaScript #TouchGrass