Let’s start here:

OpenAI just made a massive change to how ChatGPT works. 

They turned on a more persistent memory. It’s going be able to reference all your past chats and have a better sense of what you’ve asked or discussed over time.

That, in theory, sounds great.

The more you talk to it, the more it picks up on what you like, what you don’t, how you think, what kind of tone you prefer, and even what kind of work you’re trying to do.

And that raises a weird, exciting, kind of eerie question:

At what point does your AI assistant stop being “just helpful” and start becoming... you?

Co-Creation 1.0: The Dance Between Human & Tool

Until now, working hand-in-hand with AI chatbots, or “co-creating” as some might call it, has mostly felt like having a very smart intern to help you do things. They’re eager, hard working and sometimes they screw up. 

My personal experience goes something like this:

I ask it a question or give it a task, it gives me a response. I refine that question or drill down. It rewrites. I shape. It attempts to sharpen. I do the work. And in that loop, real creative momentum can build.

It feels collaborative in the way a magic whiteboard might feel collaborative. I’m the one doing the ‘stretch-my-brain’ thinking, the creative connective tissue work but it’s helping me see new angles to explore.

But one thing that made that dynamic so functional was its neutrality. The AI didn’t have an opinion. It wasn’t bringing a ton of history to these conversations (other than knowing everything piece of knowledge in the known universe). 

Maybe most importantly, it wasn’t me.

Co-Creation 2.0: AI Anticipates What You Want Based On…You

That’s all about to change. With extended memory switched on, AI will begin to anticipate what you might want based on what you wanted before.

It will learn what you tend to do, how you tend to say things, the kinds of choices you make. I suspect this will start to give it the ability to finish your sentences in a way that’s oddly accurate. I also suspect giving the AI this ability will allow it to make some of the creative leaps that makes us feel so superior.

Why? Well, part of what makes us us is our experience.

What we’ve done, lived through, dreamed about, stressed over etc. The AI is now recording and integrating all of that into its responses.

And when an AI begins suggesting things that are eerily in tune with how you would’ve written them yourself… you may have to ask:

Where do I end and the AI begin?

It’s not fully you. But it’s of you. A reflection. A remix. A machine-generated version of your taste, your instincts, your rhythms. It remembers. It builds. It adapts.

So what are we building with when we build that?

Is This the Beginning of the Singularity?

It’s kind of creepy to extrapolate where this goes from here. Ray Kurzweil’s 2005 book “The Singularity is Near” proposed that around 2040 we’d get the merging of human and robot brains aka ‘The Singularity’ and that would signal the next evolution of humanity. 

It sounds wild, but I’m beginning to suspect ChatGPT’s extended memory is the start of all that.

It’s a sci-fi future that doesn’t involve brain implants or robot overlords. It looks more like your AI tools getting so personalized, so attuned to your preferences, that working with them feels indistinguishable from thinking with them.

What Should a Creative Person Do with ChatGPT Memory?

So what should you, the creative human, actually do now that ChatGPT remembers things?

Start small. Teach it. Correct it. The more feedback you give, the better it gets at thinking like you — without becoming you.

Here’s five things to do to make it useful:

• Train it like a teammate – Tell it what tone, structure, and style you want.
  

• Name your projects – Use consistent names (like “Big New Monkey Movie Idea”) to build memory threads.
  

• Revisit old chats – Treat it like a second brain. Don’t start from scratch every time.
  

• Delete when needed – If it learns something wrong, wipe it. You can access the memory function in ChatGPT’s settings. Go find it now and get familiar.
  

• Break the mirror – Turn memory off when you want new ideas or outside perspective. I personally hope eventually OpenAI will allow us to create ‘instances’ of our ChatGPT memory (this one for work, this one for wild speculation etc).
  

Because here’s the truth: the better it gets at being you, the more important it is to stay… you.

You’re not building with yourself. You’re building with the ghost of yourself and sometimes, that ghost is brilliant. Other times, it needs to shut up and let you think.

-Gavin