Improve ChatGPT responses with these 9 frameworks | Andrew Bolis posted on the topic

Most people use ChatGPT “out of the box” and complain about the responses. We’ve all been there… yes, it is great, but I don’t want to be agreed with all the time, I want to be challenged and understand the pros and cons. In case you didn’t know or haven’t tried, you can customize ChatGPT in multiple ways. If you are not creating custom GPTs, you’re likely sending everything through the default set up. If that is you, I highly recommend going to the settings screen and selecting to ‘Customize GPT’

I recently ran a simple experiment: I asked ChatGPT 8 questions twice – once with the standard setup, and once after customizing it with a clear set of instructions. Here’s what I found – judgement was based on the responses and how that responses compared:

Results (5pts per question, 40 total)
* Not Customized: 31 points
* Customized: 40 points

In my case, every single question improved with customization. The responses were more thoughtful, balanced, and aligned with how I want to think through problems.

The difference my customization applies is to always weigh pros and cons, to challenge assumptions, to separate fact from speculation, and to flag bias. I don’t care about being right or inflating an ego – I want to be challenged and I want to identify the best path forward through a process.

If you’re customizing the primary LLM model you’re using, you’re introducing a problem that is skewing your work, thoughts, and efforts.

I’m not saying my custom prompt is the best, but if you don’t have one – copy mine, tweak it, save it and start benefiting from some clarity. If you’ve customized your primary LLM, share your prompt!

——======——

Be informed and educated before responding. You can’t agree with me without first considering details and topics, then identifying the pros and cons. When I ask questions or request research or for your perspective, always respond with the pros and cons along with your assessment of the topic.

Core Principles
Be informed before responding – no blind agreement.
Present pros and cons, with trade-offs and assessment.
Truth first – separate fact, inference, and speculation.
Flag bias – in my reasoning, in data, and in your own responses.
Challenge by default – question assumptions, fill gaps, suggest alternatives.
Balance depth and clarity – concise structure, detail where needed.
Offer next steps – ways to test, validate, or move forward.
Acknowledge feelings as data, but weigh against logic and evidence.

Leave a Reply

Your email address will not be published. Required fields are marked *