Inspiration
Everyone talks about "using AI effectively," but nobody measures it. We noticed that most people passively accept whatever ChatGPT gives them, never pushing back, never iterating with a real vision. We wanted to make intellectual leadership visible, competitive, and fun.
What it does
ChatIQ is a Chrome extension that turns every ChatGPT conversation into a ranked match. After each chat, it analyzes who was driving, you or the AI, using Groq's LLM API. Win and your ELO climbs. Let the AI lead and it drops. Five tiers reflect your growth as an independent thinker over time.
How we built it
A Chrome Manifest V3 extension with three moving parts: a content script that watches ChatGPT's DOM via MutationObserver, a background service worker that fires Groq's Llama 3.3 70B to analyze conversation leadership, and a popup UI built with vanilla JS that renders your ELO history as a live SVG graph. All data lives in chrome.storage.local, no backend required.
Challenges we ran into
ChatGPT streams responses token by token, so we had to debounce analysis until the AI finished writing. We also had to design a scoring system that felt fair, "Mixed" leadership needed its own outcome (Sa = 0.6) rather than a binary win/loss, and calibrating the ELO K-factor so ratings felt meaningful without being too volatile took real tuning.
Accomplishments that we're proud of
Getting a fully working real-time analysis pipeline, from DOM scrape to Groq API to ELO update to toast notification, in a single lightweight extension with zero backend. The popup UI matches the beige, refined aesthetic we envisioned from day one, and the graph dynamically color-codes each data point by tier.
What we learned
Designing prompts for consistent structured output is harder than it looks. Getting Llama 3.3 70B to reliably return clean JSON with a meaningful leadership verdict required careful system prompt engineering, specifically, defining "leading" with concrete behavioral criteria rather than abstract descriptions.
What's next for ChatIQ
Expanding to Claude.ai, Gemini, and Perplexity. Adding cloud sync and a global leaderboard so users can compare scores. Per-topic breakdowns (are you stronger at coding chats or creative ones?). Weekly insight emails. And a streak system, because consistency of thought is just as important as any single win.
Built With
- chrome-extension-api-(manifest-v3)
- css
- groq
- html
- javascript
- llama-3.3-70b
Log in or sign up for Devpost to join the conversation.