<aside> 💡

LLMs by their nature are static captures of the state of the world on their training cutoff date. For example, ChatGPT’s latest model (o1-preview) is working with world data it absorbed up until Oct. 2023. Anything newer it is factually blind to. Web search using Retrieval Augmented Generation (RAG) helps solve for this challenge.

Reaching out for web pages not only can bring more current, but also more targeted domain specific information into a chat context window sharpening the results.

</aside>

Example: exploring new product concept

FifthRow CEO Jan Beranek shared this video to LinkedIn. If you pause it at 50 seconds in, you’ll see how they are indeed using open web research and web scraping to source key documents from recent SEC Filings. Their custom LLM-powered app then composes a new product Proof of Concept based on that fresh research.

Jan Beránek on LinkedIn: ~80-90% data in the world is unstructured. Pre-2022, you’d either have…

Example: Pulling live reports from web

Using market trend data to identify emerging opportunities or niches for client campaigns.

• Data Source: Market Research Reports (e.g., Statista, Nielsen)
• Transformation: Accessing real-time market data helps agencies stay ahead of industry trends and monitor competitors' activities. This data can inform strategic decisions and creative direction.

Commercial App: Perplexity.ai

A current challenger to Google’s search dominance position. Perplexity uses live web search and RAG to generate results that are cited and sourced. Note, you can enable AI Automations with Perplexity and make.com to perform real time search to perform complex operations.

***last updated, November 21st, 2024***