LingoGhost

Inspiration

We realized that traditional language learning apps are often too rigid, making it hard for users to stay consistent. As developers, we asked ourselves: What if learning wasn't a chore, but an unexpected encounter? We came up with the "Mischievous Ghost" concept—a phantom lurking within your browser that stealthily swaps webpage text when you least expect it, forcing you into a state of natural language immersion via creative "MindGlitches."

What it does

LingoGhost is a Gemini 3-powered immersion tool that transforms the web into an interactive challenge. It turns passive reading into active, gamified learning:

Stochastic Immersion: It replaces fragments of webpage text into your target language at unpredictable intervals, mimicking real-world language exposure.

The "Catch the Ghost" Challenge: Users must identify correct translations amidst AI-generated distractors, turning every webpage into a potential learning encounter.

How we built it

Agentic Language Tutor (Gemini 3 API): We leveraged the Gemini model for its ultra-low latency, ensuring that "ghost" options and translations generate instantly.

Core Logic: Developed using Chrome Extension API (Manifest V3) for real-time DOM injection and mutation observation.

Performance Engineering: Drawing on our C++ development background, we optimized the text-matching algorithms to ensure zero lag during rendering.

Challenges we ran into

The primary challenge was balancing "Stochasticity" and "Readability." We spent significant time fine-tuning the probability distribution algorithms. Technically, we also resolved a critical UI bug where the first option would occasionally render as blank; we implemented a robust fallback validation layer for the AI's JSON output to ensure stability.

Accomplishments that we're proud of

Gemini 3 Driven Distractor Engine: Successfully engineered a prompt for Gemini 1.5 Flash that generates morphologically and contextually similar "ghost" options.

Seamless Visual & AI Integration: Integrated a high-fidelity "Ghost" visual identity 👻 that syncs with AI-triggered events.

Rapid Infrastructure Migration: Successfully moved our entire development ecosystem to the TeamFriday-Fj2 organization under strict Hackathon deadlines.

What we learned

We discovered that the best education is frictionless. We deepened our understanding of Gemini 3's prompting capabilities, specifically how to instruct the model to generate distractors that enhance pedagogical memory.

What's next for LingoGhost

Gamified Pts System: We are developing a feedback mechanism where users get +10 for a "Catch" and -5 for a "Glitch" (miss) to increase engagement.

Learning State Synthesis: We plan to implement an AI Mentor feature that analyzes interaction data to generate personalized summaries of cognitive progress.

WebAssembly (WASM): Rewriting core logic in Rust to achieve lightning-fast text processing for larger webpages.

Visual Learning: Integrating Gemini's multi-modal powers to label objects within images on the fly.