DeadWave 13: A Haunted AI Experience Powered by Open-Weight Models

Syed Mohammed Sazid Ullah — Thu, 05 Nov 2026 18:35:51 +0000

DeadWave 13: Tune Into the Dead, Break Their Curses 👻

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

What I Built

What if you could tune into mysterious frequencies and communicate with ghosts through AI? 👻

Introducing DeadWave 13, a Halloween-themed AI experience inspired by supernatural mysteries, haunted radio signals, and Día de los Muertos.

Instead of building another traditional AI chatbot, I wanted to create something interactive, mysterious, and fun using open-weight AI.

DeadWave 13 brings two supernatural experiences together:

📻 Ghost Radio: Tune into haunted radio frequencies between 88.0 and 108.0 MHz. Different frequencies connect you with different fictional ghost personalities, each with its own backstory and mysterious conversations.

🕯️ Curse Decoder: Enter a mysterious object, generate an AI-powered supernatural puzzle, uncover hidden clues, and try to break the curse within three attempts.

Both experiences are presented through a Halloween-inspired interface featuring violet shadows, marigold-orange accents, spectral effects, and vintage radio aesthetics.

Demo

🎨 Figma UI Preview:

https://chisel-happy-21041407.figma.site/

The Figma preview showcases the visual concept, including the haunted radio interface, supernatural puzzle cards, and Halloween-inspired design.

The Python application can be run locally using the instructions in the GitHub repository.

Code

💻 GitHub Repository:

https://github.com/md-sazid9089/DeadWave-13

The project is built using Python, Streamlit, SQLite, and open-weight AI through Ollama, with optional Grok integration.

The repository includes application source code, setup instructions, automated tests, and verification documentation.

How I Built It

I built DeadWave 13 using a Python-based architecture designed to support both local and cloud AI inference.

Tech Stack:

Python: Application logic and AI orchestration
Streamlit: Interactive web interface
Ollama: Local open-weight LLM inference
Qwen2.5: Primary local language model
Grok: Optional cloud-based AI provider
SQLite: Conversation history and game persistence
Pydantic: Structured AI response validation
Pytest: Automated testing
Figma Make: UI design and visual exploration

Ghost Radio

When users select a radio frequency, the application maps it to a fictional ghost persona.

The selected language model generates mysterious transmissions using the ghost's personality, background, and conversation context.

Users can reply to the ghosts and continue their supernatural conversations.

Curse Decoder

The AI generates fictional haunted case files containing mysterious objects, puzzles, clues, and solutions.

To keep the experience reliable, puzzle answers are validated using deterministic Python logic instead of depending entirely on an LLM to judge correctness.

Each puzzle includes three attempts and progressive hints.

Testing

I also focused on reliability.

The project passed 49 automated tests, covering AI provider handling, ghost personas, frequency mapping, puzzle validation, session storage, and Streamlit rendering.

Live Ollama inference requires a locally installed model and running Ollama service. The optional Grok integration requires API credentials.

Why Does Open Innovation Matter?

The most important part of DeadWave 13 is its support for local, open-weight AI.

Using Ollama, the application can run supported language models directly on a user's computer without requiring a paid cloud inference API.

That enables several benefits:

Privacy: Local prompts and conversations do not need to be sent to a third-party AI service.
Offline capability: Once the application, Ollama, and model weights are installed, local inference can operate without an internet connection.
Model flexibility: Users can experiment with different compatible open-weight models.
Customization: Ghost personalities, prompts, and supernatural experiences can be modified freely.
Cost control: Local inference avoids per-request hosted AI charges.

Grok is available as an optional cloud provider, but local Ollama inference remains central to the project's open-weight AI approach.

Open innovation makes it possible to create experiences that users can inspect, customize, and run on their own machines.

The Spirit of It

Halloween is about mystery, imagination, and the stories that remain after someone is gone.

With DeadWave 13, I wanted to explore that atmosphere through AI.

A simple radio frequency becomes a doorway into a fictional ghost's story. An ordinary antique object becomes the beginning of a supernatural mystery.

The goal wasn't just to generate spooky text. It was to make AI feel like part of an interactive Halloween adventure.

From haunted transmissions to cursed puzzles, every interaction is designed to create a little curiosity, suspense, and fun.

The dead have a frequency. Are you brave enough to tune in? 📻👻

🎨 Figma: https://chisel-happy-21041407.figma.site/

💻 GitHub: https://github.com/md-sazid9089/DeadWave-13

Open-weight. Private. Yours. 🎃

DeadWave 13: A Haunted AI Experience Powered by Open-Weight Models 👻

Syed Mohammed Sazid Ullah — Thu, 05 Nov 2026 18:35:06 +0000

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

DEV Community: Syed Mohammed Sazid Ullah

DeadWave 13: A Haunted AI Experience Powered by Open-Weight Models

DeadWave 13: Tune Into the Dead, Break Their Curses 👻

What I Built

Demo

Code

How I Built It

Ghost Radio

Curse Decoder

Testing

Why Does Open Innovation Matter?

The Spirit of It

DeadWave 13: A Haunted AI Experience Powered by Open-Weight Models 👻

What I Built

Demo

Code

How I Built It

Why Does Open Innovation Matter?

My Agent Session

The Spirit of It

Prize Categories