SecondSight

Inspired by Diogenes’ search for honesty, acts as a real-time cognitive firewall. It analyzes speech, video, and data to reveal manipulation, contradictions, and misleading narratives instantly.

Comment 1

## Inspiration

I’m from Argentina, where public figures and media outlets can spread false or misleading data live on air with little to no consequences. Watching misinformation happen in real time, and only being corrected hours or days later, was deeply frustrating.

Inspired by Diogenes of Sinope, who carried a lantern in daylight searching for honesty, I wanted to build a modern version of that lantern.

What it does

Augmented Critical Thinking. A real-time AI companion that brings objectivity back to public discourse.

SecondSight is a second-screen app that listens to and watches media in real time. It detects potential misinformation, logical fallacies, contradictions, and visual manipulation as they happen, providing contextual alerts instead of post-event fact checks.

🔮 The Vision: A 'VAR' for Democracy

I believe this project is a proof of concept for a new standard of civic integrity.

I imagine a future where political debates feature a real-time "Trust Score" broadcast alongside the candidates. Just as sports use technology to enforce rules, AI can help us enforce logic and honesty.

  • If a politician knows their score drops when they lie, the incentive changes.
  • We could move from a system that rewards persuasion at all costs to one that rewards accuracy and integrity.

SecondSight is my first step toward that future.

⚙️ How It Works (The Dual-Engine)

To mimic human judgment, I designed a hybrid architecture: Intuition + Research.

  1. The Observer (Gemini Live API):
    • The "Fast" Brain. It watches the video feed with me in real-time (via WebSockets) to spot tone shifts, ad hominem attacks, and manipulation tactics instantly.
  2. The Researcher (Gemini Flash + Google Search):
    • The "Slow" Brain. When a specific fact is claimed, the system pauses to verify it against the web, updating the interface with a ✅ Verified or ❌ Debunked verdict.

✨ Key Features

  • Real-Time Clarity: Insights appear instantly as the conversation flows.
  • Multimodal Empathy: It understands that how something is said (tone/body language) matters as much as what is said.
  • Live Fact-Checking: Grounds the conversation in reality using Google Search.
  • Privacy First: All processing happens anonymously in the session. It observes, it helps, and it forgets.

What I learned

I learned that combating misinformation isn’t about replacing human judgment, but augmenting it. When context is delivered at the right moment, people can think more critically and that timing matters as much as accuracy.

Note: Since I don´t have access to Gemi 3 Live API, I used 2.5 Live for the multi modal reasoning and 3 flash for the fact check.

Share this project:

Updates