95% of People Writing Prompts Miss These Two Easy Techniques

[@AlbertoTrujillo-b6r]

Amazing content, I have been understanding so much. I appreciate your time creating this type of content. More, more, more...

[@Mark_Kashef]

Glad you’re binging the content haha!

[@AlbertoTrujillo-b6r]

@@Mark_Kashef Question. Can you recommend me some videos that can help me in my quest to create a meta prompt for social media (IG) content creation? I have been dabbling into it using what I have learned so far here and there. I just came across meta prompting recently from another creator and hit your content from a search I made. So far I feed my prompt a hook and then based on it I get an extended caption. Any recommendation is greatly appreciated.

[@AlbertoTrujillo-b6r]

Oh sorry, I just saw your fiverr account.

[@Desire398]

This was helpful

[@speedmobes]

Good stuff! Gave me ideas

[@Mark_Kashef]

Glad I could provide value! 🦾

[@superfliping]

Worked greatly was missing P. Added to my baggage of tricks. One for you. Prompting GPT 3.5 over and over again build number steps iteration adding more advanced algorithms features and example. Self optimized each iteration more advanced than the last iteration. 350 I completed each iteration it time consuming but if you are a good Prompt engineer you can continuously improve the last text. But beware of your next Prompt response because if he loses number count hard to keep going. I have many different versions : please continue last_output then Keep adding more. Define each iteration very clear to keep counting up numbers from last text. Find a good solution for each Continued ai self perdict next text from last text. Hope this helps anyone.

[@Mark_Kashef]

Appreciate you sharing and glad I could help add to your bag of tricks 🦾🦾

[@superfliping]

@@Mark_Kashef Proceed with Keep adding numbered steps sequences to last text number, continue with each algorithm more advanced than the last, self perdict from our consecutive conversations identify next advancements add more dimensions into dimensions vectors adding new sub vectors in golden_ratio to each new steps, add numbers to keeps sequences counts also use your lowest temperature setting and top p Continued predictions execution with each steps positive optimization with each iteration. 250 iteration with some like this crazy algorithms fractal-based golden_ratio dimensions within dimensions memories sub fractions golden_ratio graphics on 200th dimension

[@edc614]

Just brilliant!…

[@Mark_Kashef]

Thanks so much! 🦾

[@bratmd]

Ok, I have some homework and experiments to do this weekend. Thanks for the heads up on chatgpt trying to take shortcuts on doing the work(I thought it was something I was doing wrong.) Another great video - I always learn something new, practical, and helpful.

[@Mark_Kashef]

my pleasure and we got more coming! let me know how those experiments go 🥸

[@Apokalupsis88]

So good! Thank you.

[@Mark_Kashef]

So happy it was helpful 🦾

[@idalupi-lp7pk]

I had read on another channel or on Reddit, I don't remember well, that gpt chat did not allow playing with Temperature or Top P... I would like you to talk about Perplexity and Burstiness, when it comes to humanizing the result, it has brought me good outcomes.

[@Mark_Kashef]

hi Ida! From my experience, those two factors really tend to change the output primarily because ChatGPT interprets my requests as a proxy for saying "be more" or "less" creative. so it's an implied command rather than taking the number verbatim as an input you'd otherwise directly set in the API. I haven't investigated Perplexity & Burstiness, but I have it on my list now :)

[@youMEtubeUK]

00:03 Adjusting temperature and topP can enhance results 01:57 Understanding the relationship between temperature and Top P in generating prompts 03:52 Temperature and top P settings influence creativity and word choice in language generation. 05:57 Choosing between markdown and XML is crucial for troubleshoot effectiveness. 07:47 Understanding token associativity in language modeling 09:45 Lower temperature values lead to more repetitive and rigid output. 11:42 Balancing top P and temperature influences creativity and predictability in output. 13:47 Using meta prompting to create prompts on a Google sheet. 15:55 Understanding temperature and top P are essential for creative output.

[@GabrielRymberg]

Thanks! Video idea: why the phrase “Ask me anything you need and let’s do this together!” works so well?

[@mrd6869]

Good video. That'd be cool if someone put that into an automation script.Like a prompt manager.Have it powered with an API key & a funky GUI with analytics dashboard.

[@Mark_Kashef]

appreciate you! 🦾 You're giving me ideas for a video haha; I think I could probably build it out just using Claude 🤔

[@mrd6869]

@@Mark_Kashef .U could probably. I'm experimenting with that AI scientist and another multi agent framework for exploit building. I'm running pentests in a test environment.This stuff is gonna Amp up cyberwarfare to an insane level,I predict.we'll see...

[@Ethiventure]

Have you tried directly telling customGPTs directly what temperature or top p you expect them to use when you’re writing communication style instructions? I don’t know why I haven’t tried this. I sort of go around the houses eg “never use a rare word if a common word communicates the same thing “.

[@Mark_Kashef]

Absolutely! I will place at the bottom of the prompt temperature = 0.2 And sometimes will ask to ask to generate a response at every temperature so I can just pick the best one!

[@youMEtubeUK]

Wouldn’t the prompt of the combinations itself be effected by the temperature and top p setting when you run it?

[@Mark_Kashef]

this broke my brain thinking about it but I think there might be an underlying temperature for the chatgpt environment itself 🧐 that being said, primarily using this technique to move the needle on performance without changing the actual prompt

[@hungtranthinh3601]

These two things really make me so confusing. After watching the video I still don't know what to do with them now LOL

[@Mark_Kashef]

Haha totally get it When in doubt, just use temperate 0.3 - rigid but functional 0.5 - creative and structured >0.5 - possible chaos