“The next frontier for content recommendations is going to be AI merging content rather than users flipping between articles. Imagine an app that is a continuous article and writes itself as you read based on its understanding of your interests & your interaction with the content.”

It’s important to identify the low-hanging fruit in these types of situations. Where are the trade-offs the least severe? This is known as a wedge, and it can be accomplished by narrowing the initial product and/or the initial market.

Tesla started at the high-end, where people were willing to pay so much money for an electric sports car that economies of scale didn’t matter.

YouTube started with personal videos people wanted to embed on their MySpace pages, rather than professionally created content (and all the concomitant legal headaches).

Uber started in San Francisco thanks to its hills, poor taxi or public transit availability, lack of parking, and lots of wealthy tech workers.

One of the easiest tasks for current LLMs is summarizing news articles.

Google search is the closest thing we have today to the infinite article — especially now that they’re pulling information out of articles and displaying it directly on the search results page.

when it comes to true personalization, nobody knows more about our prior information consumption than Google, so they’ve got a head start in that department.

They trained the latest model of GPT-3 with information up through June 2021, which means it doesn't know anything that happened after that.