The distinction between vague and specific prompts couldn't be more true! I've experimented extensively with prompt engineering and found the exact same pattern - the quality gap between generic requests and structured prompts is massive. The framework shared here (Context, Clarity, Constraints, Character, Criteria) mirrors what I've seen work consistently in real-world applications.
What's particularly valuable is the emphasis on iteration and collaboration with AI rather than treating it as a vending machine. When I started approaching AI as a thinking partner rather than just an output generator, my results improved dramatically. I recently explored how this collaborative approach applies to building practical solutions in my latest write-up: https://thoughts.jock.pl/p/csv-column-stripper-affordable-ecommerce-data-solution
Totally agree! Today I no longer just ask AI to do something for me but involve them in the process. Massive difference in terms of results. Cant go back to the way I used to.
I love this. Thanks Wyndo. I actually like when an assistant tells me “no I can’t do that “ because it forces me to use one or more of these techniques to get the task completed. It’s teaching me how to interact with it.
Love that! I also love to tell my AI to be disagreeable and tell me if I dont do good job at giving inputs and suggesting ideas. And also ask me new questions to make sure they have better context based on tasks at hand.
That’s a really good technique. I wish there were a good way to integrate multiple agents into VS Code to turn this into a broader, role based discussion
Those tips are very similar to the way one has to behave when approaching an oracle as Yi Jing
The distinction between vague and specific prompts couldn't be more true! I've experimented extensively with prompt engineering and found the exact same pattern - the quality gap between generic requests and structured prompts is massive. The framework shared here (Context, Clarity, Constraints, Character, Criteria) mirrors what I've seen work consistently in real-world applications.
What's particularly valuable is the emphasis on iteration and collaboration with AI rather than treating it as a vending machine. When I started approaching AI as a thinking partner rather than just an output generator, my results improved dramatically. I recently explored how this collaborative approach applies to building practical solutions in my latest write-up: https://thoughts.jock.pl/p/csv-column-stripper-affordable-ecommerce-data-solution
Totally agree! Today I no longer just ask AI to do something for me but involve them in the process. Massive difference in terms of results. Cant go back to the way I used to.
Indeed, the best results definitely come when you treat it like a teammate, not a magic machine. Great article.
Yep, best hack of AI is actually working with them, not extracting them. Massive difference!
I love this. Thanks Wyndo. I actually like when an assistant tells me “no I can’t do that “ because it forces me to use one or more of these techniques to get the task completed. It’s teaching me how to interact with it.
This too will change, right?
Love that! I also love to tell my AI to be disagreeable and tell me if I dont do good job at giving inputs and suggesting ideas. And also ask me new questions to make sure they have better context based on tasks at hand.
That’s a really good technique. I wish there were a good way to integrate multiple agents into VS Code to turn this into a broader, role based discussion
Haha I’d love that turning into reality. We need more ways to problem-solve things :)
Great piece!