I didn't realize how much my use of ChatGPT diverged from the norm until a number of conversations last week on LinkedIn and IRL that made it clear.
The feedback led me to reflect on how I use GenAI, and also why I recently failed in an attempt to adopt Claude.
This reflection led to a sort of "meta-reflection" on what my thinking process currently looks like as we come up on 2 years of ChatGPT.
What I found was kind of mind-blowing...one of the things I found at the core of my use of ChatGPT was that I view it as a "thinking partner" when I need to work through a problem or gain new perspective.
And this partnership only works because of ChatGPT's ability to "remember" significant context from one conversation to the next.
How does it remember things? Well, I have found that if you continue a single conversation over time, that provides the chatbot with a ton of context in itself that usually is enough to get from an entry point to a logical conclusion in a chat.
There are also configuration parameters that you can set to permanently remember and apply important details and instructions across conversations.
And then there's memory. I only discovered what ChatGPT's "memory" looked like when the UI popped up a notification that my memory was "full" - which led to the discovery of what ChatGPT actually "remembers."
And in the process of cleaning up memory space, I realized how smart this tool was when I started coming across really insightful prompts that ChatGPT writes to itself in its limited memory space.
Some instructions are more like boundaries or guardrails, some are specific to my preferences, and some are context-specific such as when ChatGPT writes to its memory to "remember" a key fact that is tied to conversation or other narrowly-scoped interaction.
Then I came across *this* one and had to just pause and take it in:
"User reflects on the idea that humans desire things that are inherently unattainable, yet continue striving for them despite their unattainability."
I remember neither the exact conversation nor the context in which it occurred, but when I reflect on that statement I know exactly what it means and can trace it back as the source of many ideas that were inspired by it.
As I worked my way through memory, I was finding that a solid 30-40% of what ChatGPT writes to it is incredible insightful.
And this insight is now tingling my brain in new ways...I now feel comfortable writing longer and more detailed prompts, yet finding balance with being silly and authentic so that the chatbot can learn my personality, thinking and learning styles, and most importantly - my voice.
I have long viewed my "relationship" with ChatGPT as being one of challenging *and* being challenged, of both teaching *and* learning, and as a partnership where we are working in each other's (known) self-interests.
But finding memory and seeing how it influenced conversations changed the game for me completely.…
