Like a completions-based product, the advantage of a feed is that users don’t have to interact with a chatbot. The inputs to the model come from how the user interacts with the feed (likes, scrolling speed, time spent looking at an item, and so on). Users can experience the benefits of an LLM-generated feed (if any) without having to change their... See more

I think there’s almost certainly potential here for area-specific research agents (e.g. in medicine or law).

Agents are kind of like chatbots, in that users interact with them by typing natural language text. But they’re unlike chatbots in that you only have to do that once : the model takes your initial request and goes away to implement and test it all by itself.

The reason agents work and chatbots-with-tools don’t is the difference between asking an LLM... See more

The genius of this kind of product is that users never have to talk to the model . Like I said above, chat is a bad user interface. LLM-generated completions allow users to access the power of AI models without having to change any part of their current workflow: they simply see the kind of autocomplete suggestions their editor was already giving... See more