Prompts are the interfaces for humans to interact with large language models, and they are also the "new API" for applications to set context, establish guardrails, inject data from external sources, and more.

So what should builders know about using prompts to create great applications?

We're going to talk about that today with Shreya Rajpal of Guardrails, Grant Slatton of RowZero, Laurel Orr of Numbers Station, and Diego Oppenheimer of The Factory.

I'm Jon Turow, and this is the AI for Builders Podcast.

This is Episode 1 of AI for Builders.

There's a major limitation of foundation models. They only know the facts they were trained on. And almost any serious application needs to incorporate fresher and broader data -- think about a travel app which needs flight inventories, or an analytics tool that needs access to a data warehouse.

Today we're going to talk about the power and opportunity of foundation models to reason about external data sources.

We'll do that with some of the founders who have thought more about it than almost anyone else.

Here's Jon Turow, and today I'm with Matt Busigin the founder of Cognosis and Jerry Liu the founder of GPTIndex. We'll also be joined a bit later by Diego Oppenheimer, founder of Algorithmia and a Partner at the Factory.

"Is this thing on?"

Our first episode of AI for Builders, with Jon Turow, Moin Nadeem (Stanford, MosaicML), Bill Lennon (Ai2), and special guests Tristan Zajonc (Continual) and Matt Busigin (Cognosis). We spoke about (1) paths to let foundation models reason about external data more efficiently, and (2) durability of the need for FM dev frameworks like Langchain or Dust.