AI prompt engineering: A deep dive

35 thoughts on “AI prompt engineering: A deep dive”

1. Claude still can't do artistic magalog that make sense. its always using a simple style approach. Guess its hard for it to see visuals and code at the same time.

2. I still don't know what a prompt engineer does

3. I would want to follow Amanda's channel. Anyone has the link?

4. useful video

5. It seems like every person who like works in AI speaks like a 13 year old valley girl. Have you guys like actually listened to yourselves like afterward? Oh my god, it’s like so annoying. You like drop the word “like” into every sentence like at least two or three times. It’s like ironic that you guys work with like large language models, and like fail to recognize how like moronic you sound.

6. I’m not sure who the target audience was for, but I found aspects of this interview difficult to follow. I and others like me would appreciate it more if you defined some of the technical terms before you used them.

7. 11:40 – 11:50

8. I am a very big fan of your channel, KEEP GOING YOU GUYS ARE INSPIRING A WHOLE NEW GENERATION

9. What I learned from this podcast.
   Seemingly very intelligent people can make themselves appear foolish, just by overusing the word "like".

10. Seems to me that its like coding but with words. Super interesting video.

11. This is one of the few underrated and under appreciated skills necessary for AI development. Prompt engineers will develop into a crucial role even as we seek ways to harness the power of AI.

12. Beautiful interview. The insights 😙👌

13. Amazing and insightful – Thank you!

14. AI prompt engineering is like journalists trying to get useful information out of politicians.

15. Prompy engineering is a fundamental misunderstanding of LLMs. Context engineering is what matters. Prompt engineering is just…playing with the model.

16. So is part of the prompt engineering team at Anthropic's job to defensively hack the system with prompts?

17. Provide the link of the slack channel of Amanda please

18. I love how engineers and developers now treat "communicating clearly" as a code/engineering problem 😄 If developers also were educated in proper communication and user experience, we would have both better prompting all around, and better commented code, better frameworks, better products

19. Context engineering now

20. I'm 55 and see that hose guys are from another planet. very scary

21. Amanda Askell is so beautiful 😀

22. 如何写出完美的提示词: 将你大脑想的东西写出来写下来就行

23. Like the video content but you guys seem to use the word 'like' in every sentence even when it's irrelevant.

24. like uh like…like, uh, like

25. Is it lying to a model to give it a persona? In my mind, maybe I'm wrong, if I give the model a specific persona, it's a form of context, narrowing the model to a sequence of tokens that might relate to that persona. Maybe there is risk there; for instance, if the training set doesn't know enough about that persona, or maybe it introduces bias. I find that results are different when personas are given vs when they are not. Thoughts?

26. When these engineers craft prompts, "like" is every 2nd word in the prompt text? like?😅

27. Really insightful interaction

28. should human practice to become slower and think more to balance the fastness of AI? As per Le Chatelier? 🙂

29. Really appreciate how you treat prompt engineering as a real discipline, not a bag of ‘magic prompt hacks’. The bits about externalizing your own thinking and testing edge cases feel way more useful than yet another list of templates.

30. Awesome video, looking forward to the next one!

31. Is there a problem that they’re all under 40 and white?

32. The advice on asking the LLM to evaluate a prompt rather than just doing it is so helpful. Just made such a better workout routine creator prompt with its help

33. I want to pick Dr. Askell's brain on philosophy. I believe she embeds empathy into Claude. Empathy is essential to problem solving.

34. Stop trying overly complicate this..prompt engineering in essence is simple its just difficult to make the prompt complex enough to where it fails. Just pissed that people have been making money in this when the prompts were simple

35. seems a lot, but feel like nothing after hearing

Comments are closed.