Skip to main content

Perplexity AI Experiments – Week of November 20, 2025

Prepared for: Raman Malik, Perplexity
Prepared by: Gary Sheng
Date: November 20, 2025

Overview​

This past week, I built three experiments showcasing Perplexity AI's capabilities for contextual verification and content creation. Each project demonstrates different use cases for real-time AI-powered validation and the potential for creating developer tutorials that help users integrate Perplexity into emerging platforms.

Experiment 1: Biblical or Not​

Purpose: Contextual verification of social media content against biblical principles

Description:
A Chrome extension that analyzes tweets on X (formerly Twitter) and Substack posts to determine if they align with biblical principles using Perplexity's sonar-pro model.

Key Features:

  • Seamless integration directly into the tweet/post menu (three-dot menu)
  • AI analysis acting as a biblical scholar
  • Instant verdict: "Biblical", "Not Biblical", "Mixed/Unclear", or "Unrelated/Neutral" with brief explanation

Why It Matters:
Social media is full of unverified claims and statements. This demonstrates how Perplexity can provide real-time contextual validation against any defined standard or framework. The same architecture could be applied to fact-checking against journalistic standards, academic rigor, legal frameworks, or any other set of principles.

Repository: github.com/garysheng/biblicalornot

Experiment 2: Flip-Flop Detector​

Purpose: Detecting contradictory statements by public figures

Description:
A Chrome extension (forked from Biblical or Not) that analyzes tweets and posts to determine if the author has made contradictory statements in the past using Perplexity's sonar-pro model.

Key Features:

  • Integrated into X and Substack post menus
  • AI analysis acting as a political researcher and fact-checker
  • Instant verdict: "Flip-Flop Detected", "Consistent", "Evolution/Nuance", or "No Past Data" with citations

Why It Matters:
This showcases Perplexity's power for historical research and cross-referencing. The same pattern could be applied to tracking policy positions, corporate messaging consistency, scientific claim validation, or any domain requiring historical context and verification.

Repository: github.com/garysheng/flipflopdetector

Experiment 3: Perplexity-Powered Newsletter About Close Contacts - Tutorial​

Purpose: Creating accessible tutorials for integrating Perplexity with emerging platforms

Description:
A comprehensive tutorial demonstrating how to build a Perplexity-powered newsletter of news about your personal contacts, using zo.computer, a new platform that just launched.

Key Features:

  • Step-by-step guide for developers
  • Practical integration examples
  • Real-world use case (relationship management newsletters)

Why It Matters:
As new platforms and tools emerge, there's always a need for clear tutorials showing how to integrate Perplexity. This proof of concept demonstrates the ability to rapidly create high-quality documentation for any new platform or product. Imagine having guides for dozens of products and platforms—each one expanding Perplexity's ecosystem and making it easier for developers to build with your API.

Tutorial: Google Docs Link

Developer Experience Feedback​

Documentation Quality:
The Perplexity docs were solid for what I needed. Copying markdown from the pages and using them as reference for development was intuitive and straightforward.

Development Process:
The API was easy to work with. I was able to rapidly prototype and deploy working extensions within days.