Pseudoglot

Pseudoglot - Learn a language by living in it. Passive multilingual immersion, one word at a time.

Comment

Inspiration

Learning a new language is tough — especially for busy people who don’t have time to sit down with an app every day. But what if you could learn passively, while doing what you already do — browsing the web?

We wanted to create a tool that brings natural, contextual language exposure into everyday browsing. Inspired by the way people pick up words when immersed in a new country, our idea makes this experience accessible, AI-smart, and personal — directly in your Chrome browser.

What it does

Our Chrome extension transforms ordinary web pages into a dynamic language-learning environment, intelligently powered by Google Gemini AI:

✨ Smart in-page translation: When you visit any website, the extension uses Gemini’s advanced language reasoning to read the entire page, understand its context, and select the most relevant words you’re ready to learn next — based on your level and the surrounding grammar. Instead of random translations, Gemini considers word frequency, contextual clues, syntactic patterns, and semantic relationships to pick vocabulary that’s easy to guess and useful to reinforce.

It then swaps these carefully chosen words into your target language — so you naturally acquire vocabulary as you read.

🔍 Interactive hover popups: Every translated word is highlighted. Hover over it to instantly see its meaning, part of speech, pronunciation, and example usage — all driven by Gemini’s understanding of how the word fits into the sentence.

🔊 Click-to-pronounce: Click the speaker icon to hear how the word is pronounced, helping you build accurate listening and speaking skills alongside reading.

🧠 Personalized word tracking: Every new word you encounter is saved to your personal learning record. Gemini uses your learning history to recommend words that build on what you already know, gradually introducing new vocabulary in a way that matches natural language acquisition. After each browsing session, you can review what you’ve learned or test yourself with quick, Duolingo-style quizzes to strengthen retention.

TLDR Our extension makes language learning effortless, smart, and naturally integrated into your everyday browsing. With Gemini’s linguistic intelligence, it’s like having a personalized tutor guiding you — one word at a time.

🚀 How We Built It

🏗️ Project Architecture

Our language-learning Chrome extension is built as a full-stack application with three main parts:

  • Frontend (/frontend): A modern Next.js React app using TypeScript, Tailwind CSS, and Radix UI for a clean, responsive dashboard and settings panel.
  • Backend (/backend): A lightweight Flask API server in Python that connects to Google Gemini AI for intelligent word recommendations and handles user progress storage.
  • Chrome Extension (/extension): A Manifest V3 Chrome extension that interacts with webpages in real-time, replacing selected words with smart translations and showing hover tooltips with pronunciation.

⚙️ Core Technologies & APIs

🎓 AI & Translation

  • Google Gemini AI: Integrated using the google-genai Python SDK. We use the gemini-2.5-flash-lite-preview-06-17 model to generate contextually appropriate vocabulary lists based on the user’s learning level and webpage content.
  • Custom Prompting: Prompts guide Gemini to leverage linguistic insights — word frequency, syntax, and semantic context — to select words that match the user’s proficiency and foster natural language acquisition.

💻 Frontend Stack

  • Next.js (15.2.4) — Server-side rendering for fast performance.
  • React (19) with TypeScript — Type-safe, modern component architecture.
  • Tailwind CSS — Utility-first styling for rapid UI prototyping.
  • Radix UI & Lucide React — Accessible UI primitives and icon sets.
  • React Hook Form + Zod — Robust forms and validation.

🗂️ Backend Stack

  • Flask (2.3.3) — Lightweight Python server framework.
  • Flask-CORS & dotenv — Secure cross-origin requests and environment variable handling.
  • CSV-based storage (MVP) — Simple file-based storage for word tracking (easily upgradable to a database).
  • Custom REST endpoints — Handle translation, stats, and health checks.

🧩 Extension Tech

  • Chrome Manifest V3 — Modern, secure extension structure.
  • Vanilla JS — Lightweight content scripts for precise DOM manipulation.
  • Chrome APIs — Storage, runtime messaging, scripting.
  • DOM TreeWalker & MutationObserver — Replace text nodes in real-time without breaking page layouts.

Challenges we ran into

🤯 After an hour of pushing Gemini to its limits—analyzing every new webpage and chunk of past data—our API key finally tapped out. That’s when we hit a hard truth: AI at the client level isn’t cheap.

🔧 So we restructured the entire system, slashing Gemini calls by ~95%. The result? A leaner, faster setup that we can now playtest freely without worrying about hitting quota walls. 🧪🚀

Accomplishments that we're proud of

🧠✨ A self-learning AI model that smartly adapts to your language journey. It recognizes patterns in your learning data, nudges you toward more trivial vocabulary when needed, and sprinkles in rare words to keep things fresh and challenging. Every day, it kicks off from a different point on your personal learning path to avoid repetition and burnout.

🌍 Thanks to Gemini’s powerful data curation, the system scales effortlessly — adding a new language is as simple as picking a new font family. 🔤⚡

What we learned

🧠⚙️ AI and vibe coding are becoming the new foundations of the future — not as replacements, but as amplifiers of a developer’s vision.

💡 Today, learning tech before using it is optional. You can explore frameworks, tools, and languages freely — test-drive first, master later. This shift unlocks a new era where curiosity leads, and intuition shapes innovation. 🌐🚀

What's next for Pseudoglot

🧪✨ Right now, Pseudoglot is just a word-swapping sidekick. But to truly evolve into a powerful language-learning companion, it needs more than that.

📊 It’ll grow to include testing modules, 🧠 progress tracking, and 📤 sharing features. Even deeper: support for context-aware verbs, 🌍 cultural idioms, and smarter translations that understand what you mean, not just what you say.

Built With

Share this project:

Updates