Inspiration The intellectual brilliance of AI comes with a silent, heavy price: environmental destruction. Every interaction with a Large Language Model triggers high-performance GPUs that require massive amounts of electricity and water for cooling. I realized that while we embrace AI's efficiency, we are unknowingly accumulating an immense "Carbon Debt". Inspired by the duality of Dr. Jekyll and Mr. Hyde, I wanted to create a tool that transforms resource-heavy, "Hyde-like" prompts into lean, sustainable "Jekyll" alternatives.
How I Built It I developed Jekyll’s Eco-Prompt as a Chrome extension to provide a seamless, real-time interface for sustainable AI usage.
Real-Time Diagnosis: I injected a "Diagnosis Bar" directly into the ChatGPT composer using a MutationObserver to track text changes.
The Optimizer: I built a regex-based engine that identifies and strips redundant fillers and over-polite phrases, reducing token waste by up to 60%.
Impact Dashboard: The dashboard calculates environmental metrics using specific coefficients to convert tokens into tangible resource costs.
Environmental Debt Logic: Instead of traditional "savings" trackers, I implemented a "Carbon Debt" system that visualizes the user's footprint as a liability to be repaid through reforestation.
The Challenges I Faced The primary challenge was the dynamic nature of the ChatGPT DOM. Since the composer surface lacks a static ID, I had to implement robust attribute-based selectors and a ResizeObserver to ensure the Diagnosis Bar remained perfectly flush with the input box during window resizing. Additionally, creating a "Success Toast" that felt integrated rather than intrusive required fine-tuning CSS animations to prevent overlapping with native UI elements.
What I Learned Through this project, I gained deep insights into the resource-to-token ratio in generative AI. I learned to model environmental impact through mathematical projections:
E total=∑(T optimized ×C token )
where T represents tokens and C represents the carbon coefficient per token. This experience taught me that small changes in prompt engineering can lead to massive cumulative reductions in global energy consumption.
Log in or sign up for Devpost to join the conversation.