At first, prompting seemed to be a temporary workaround for getting the most out of large language models. But over time, it's become critical to the way we interact with AI. On the Lightcone Podcast, Garry, Harj, Diana, and Jared break down what they've learned from working with hundreds... of founders building with LLMs: why prompting still matters, where it breaks down, and how teams are making it more reliable in production. They share real examples of prompts that failed, how companies are testing for quality, and what the best teams are doing to make LLM outputs useful and predictable. 0:58 - Parahelp’s prompt example 4:59 - Different types of prompts 6:51 - Metaprompting 7:58 - Using examples 12:10 - Some tricks for longer prompts 14:18 - Findings on evals 17:25 - Every founder has become a forward-deployed engineer (FDE) 23:18 - Vertical AI agents are closing big deals with the FDE model 26:13 - The personalities of the different LLMs 27:26 - Lessons from rubrics 29:47 - Kaizen and the art of communicationshow more

Tune in:

If you're not learning AI in 2025, you're falling behind. Join 1,000,000+ early adopters reading and learn AI in just 5 minutes a day (for free).

@LightconePod Prompt engineering will persist when you consider that everything from emails to legal documents can be thought of as prompts. LLMs need guidance - with language - to achieve a goal. Not all guidance is created equal.

@LightconePod Can't prompt forever, autonomy ftw 🤖

@LightconePod 👀

@LightconePod Prompting is now an art form, not just a hack. We’re witnessing the evolution of how we collaborate with AI. Count us intrigued.

@LightconePod If you're into this episode, check out @clairevo's show "How I AI", or this recent YouTube video (and new channel) for yours truly. I'm open sourcing and sharing code as much as I can as I go.

@LightconePod good podcast

Prompt engineering has clear limitations. It's not an exact science, often inefficient, and challenging to maintain consistency amid frequent frontier model updates. Even established techniques like Few-Shot prompting and Chain-of-Thought reasoning remain error-prone. For custom tasks absent from model training corpora, prompt engineering typically delivers lower accuracy than custom model training or traditional ML methods that offer greater efficiency.

@LightconePod If you're prompting a lot and need a better tool than a simple chat interface, please check out @AIFlowChat

@LightconePod on the topic of llm personality, i recently gave the top models the big 5 personality test. it really helped me understand when to go to which llm and why