AI prompt engineering in 2025: What works and what doesn’t | Sander Schulhoff

In this episode of Lenny's Podcast, Lenny interviews Sander Schulhoff, an expert in prompt engineering, to discuss practical techniques for improving interactions with Large Language Models (LLMs). They cover basic methods like few-shot prompting, decomposition, self-criticism, and providing additional context, as well as more advanced techniques like ensembling. The conversation shifts to the critical topic of prompt injection and AI red teaming, where Schulhoff explains how AIs can be manipulated to produce harmful content or actions, and the challenges in securing AI agents against such vulnerabilities. He emphasizes that while AI safety is a serious concern, the benefits of AI development, particularly in healthcare, outweigh the risks, advocating for regulation rather than halting progress.