I stumbled on this long-form convo with the team behind Anthropic’s prompt engineering — and honestly, it’s the most grounded, practical breakdown I’ve seen so far.
Amanda Askell, Alex Albert, Zack Witten, and David Hershey go deep into how they actually think about prompting — not just tricks, but stuff like using metaphors, designing personas, understanding model reasoning, and adapting prompts for research vs enterprise use.
There’s also a bunch of underrated gems — like how they refine prompts over time, what “good prompting” really looks like, and how they deal with jailbreaking and edge cases.
It’s a goldmine if you’re building anything with AI — especially if you’ve been feeling like prompt engineering is still mostly guesswork.
Kush Prajapati