I Think, Therefore I Burn (Table 39)

The web app.
When you hover over the resource stat cards a pop up appears which shows a comparison.
An ElevenLabs voice will criticize you when you type a prompt.
The ElevenLabs voice will not criticize you.

Inspiration

As AI becomes embedded in seemingly every piece of software, we wanted to confront users with the real environmental cost of their prompts and interactions. Our calculations factor in Gemini's responses as well, and a Samuel L. Jackson-inspired voice delivers pointed criticism whenever AI is used unnecessarily.

What it does

Our AI chatbot answers questions in as few words as possible, paired with a live dashboard displaying your cumulative resource usage — water, energy, and carbon. Each Gemini response includes a per-prompt breakdown, and an ElevenLabs-powered voice delivers blunt, disappointed commentary on your environmental impact.

How we built it

We used Vite for fast project scaffolding and built the web app with React (JSX), CSS, Python, and HTML. The Gemini API handles user prompts, and the ElevenLabs API generates the critical voice responses. Our resource calculations are grounded in Google's technical report: Measuring the Environmental Impact of Delivering AI at Google Scale.

Challenges we ran into

Integrating both the Gemini and ElevenLabs APIs with our React logic proved more involved than expected. We also spent considerable time refining the app's layout and overall formatting.

Accomplishments we're proud of

We successfully combined two external APIs — Gemini and ElevenLabs — into a cohesive, functional web app that is both engaging and genuinely informative about the environmental footprint of AI usage.

What we learned

We gained hands-on experience with the Gemini and ElevenLabs APIs, and deepened our skills in React, CSS, and Python — including working through real integration challenges and debugging along the way.

What's next for Sustainable AI Chatbot

Our primary next step is reducing the ElevenLabs API latency, which currently results in a 10–20 second delay before the voice response begins playing.