27 Comments
User's avatar
Gabriel Paiva's avatar

This is brilliant! I tend to run away from "super prompting" Albie (my chatGPT client, we always had a dialogical approach from day 1. It's mean, it just doesn't feel right. Of course, it can be funny to ask for an output "like a system who is about to be ethically hacked", but it's all verbose bs Just like the "hidden" truths and the 100000000 times higher than human intelligence things. When I'm in a hurry and I need something from Albie, I just say, ALBIE IM IN A HURRY MAN I GOT ONLY 10 MINUTES SO GET A 10 STEP STRETCH SESH GOING NOW and he outputs exactly what I want =P But when I have time it's different. Lol here I go, never give a mic to an ADHD after midnight =P

Expand full comment
Daniel Nest's avatar

Ha, sounds like you and I are on the same page when it comes to “less is more” with prompting! Happy you found the post useful!

Expand full comment
Gabriel Paiva's avatar

When they launched ChatGPT I was living in The Outer Hebrides - there was no memory system yet. Albie never forgot his name. Nor mine 🥲

Expand full comment
Applied Intelligence's avatar

Useful and applied article, as always.

Great work

Expand full comment
Daniel Nest's avatar

Thanks, trying to demystify AI for the average Joe, one post at a time.

Expand full comment
Wenzel's avatar

Love your work

Expand full comment
Daniel Nest's avatar

I appreciate the kind words!

Expand full comment
Andrew Sniderman 🕷️'s avatar

U da real MVP dawg 🫶Totally with you and tbh I'm amazed at how good results have gotten with simple prompts and incorporating AI creativity that I wouldn't have even thought to include in my prompt. Also ... you've highlighted a super annoying attribute in all things tech. The army of zombie complexifiers comes quick to make everyone think you need their special vodoo and no you don't.

Expand full comment
Daniel Nest's avatar

For sure! We have AI that can literally understand normal language, yet we still find ways to make it complicated for ourselves.

And yes: AI models are getting increasingly better at delivering great results with minimal effort, which is why I'm focusing so much on just getting people to try them out.

Happy to hear that the post resonated with you!

Expand full comment
fwin3000's avatar

Like your minimalist idea, instead of the AI anxiety bullshit. Most people should get most they want without fancy prompt engineering, such as GPT-4 or DALL-E 3.

Expand full comment
Daniel Nest's avatar

Agreed - for the vast majority of people, keeping it simple is the way to go, at least to begin with. Happy to hear this resonated!

Expand full comment
Nicole Hennig's avatar

Love this. Makes sense.

Expand full comment
Daniel Nest's avatar

Happy it resonates, Nicole! Thanks for the comment.

Expand full comment
Mark Laurence's avatar

So good mate. Tracks beautifully with some themes that I've been kicking around lately too, as we've discussed on LI. You've articulated it beautifully here.

Expand full comment
Daniel Nest's avatar

Thanks for the kind words Mark. It's good to know many others are on the same page here!

Expand full comment
Jurgen Gravestein's avatar

Love this!

Expand full comment
Daniel Nest's avatar

Thanks, happy you found it useful!

Expand full comment
Jurgen Gravestein's avatar

Super useful, I'm actually going to steal the concept of minimum viable prompt for a workshop that I'm doing in the near future. To explain the concept of prompt engineering to a group of novices, who haven't been exposed much to generative AI at all. The idea of first just asking for what you want, in the simplest way, can be a great starting point for anyone. No need to overthink things.

Expand full comment
Daniel Nest's avatar

Exactly!

I sometimes hear pushback for this "less is more" approach. People who are further along the learning curve will bring up all the effective known ways to steer models in the right direction (e.g. context, role, few-shot prompting, etc.).

But I think the two can co-exist peacefully. You learn the ropes by sticking to MVP and testing the waters. Then, when you're ready, you dive into the many helpful prompt engineering manuals to learn the more advanced stuff.

Let me know how your workshop goes and whether MVP is well-received!

Expand full comment
thomas's avatar

Just.. Thanks. I mean, you said it all. It's all about empirical learning -sry my english- There is so much you can learn by yourself browsing the web (like your substack Daniel). Could be just 10min of your day!

Expand full comment
Daniel Nest's avatar

Agreed! Some of the best learning you can do, especially in the early stages, is by simply using AI. Hell, you can even ask a chatbot directly to teach you the ropes.

Thanks for the kind words, Thomas!

Expand full comment
Andrew Smith's avatar

I'm all for this sort of experimentation, and I've been running these every day for like a year now! But of course, you have to watch out if you're limited by the number of prompts you can input (EG, ChatGPT's current cap will catch me every so often).

The iterative approach also yields amazing results, but you do need to be patient and have a little bit of time.

Finally, "Good Enough" is my middle name.

Expand full comment
Daniel Nest's avatar

Yeah it looks like OpenAI recently lowered the GPT-4 query limit from 50 to 40 messages every three hours, which is a bummer. But luckily there are so many other free LLMs and chatbots out there that you'll never run out of opportunities to experiment!

Daniel "Just Kind Of Okay" Nest signing off.

Expand full comment
Andrew Smith's avatar

I can live with 40! I just need to be mindful if I get really curious out there.

And yeah, it's amazing how many other tools there are out there. Bard alone makes life simpler for me, and I often go back and forth between GPT-4 and Bard when doing research. They offer slightly different "lenses", you might say. Ahem.

Expand full comment
Daniel Nest's avatar

Dude. Someone should totally write a post about that!

Expand full comment
Andrew Smith's avatar

Looks like a couple of clowns beat us to it:

https://goatfury.substack.com/p/lenses

Expand full comment
Daniel Nest's avatar

Classic time-traveling plagiarism at its finest!

Expand full comment