Sometimes I think we’ve been a little duped with AI.
Because of its speed, precision, and vast knowledge bank, its ability to make lightning-fast inferences, we start to treat it like it’s the gospel.
But here’s what we forget:
ChatGPT, Claude, and the rest are trained on the median of everything they can find: books, websites, articles; sometimes legally, sometimes questionably.
The LLMs don’t know the worst. They don’t know the best.
They know averages.
And while that can be good enough (it’s literally how inferences works) it’s also why the answers you get from LLMs can feel a little blah.
Also, a little too familiar like you’ve seem them before *** TRUTH – you probably have. ***
For example, you’ve opened a new chat window, asked a question, and feel like you’re getting an answer that you’ve seen somewhere, out there or just rearranged from something you’ve asked before.
At this point, most people assume they need better prompts, but in my experience, that’s not the case.
It’s how you’re using ChatGPT in the first place, that’s making everything seem blah.
The Shift From Using ChatGPT as a Tool to Using It as a System
A few months ago, I sat down with a cup of hot chocolate.
I was creating Amy — my AI agent for founders building offers — and brainstorming with ChatGPT on some key data points.
And I felt this strange kind of déjà vu.
Like ChatGPT had given me the same responses in a different chat window. Same topic, same answer, just dressed up differently.
Now, it would’ve been easy to get frustrated at that point and continue working in the same long-winded way, having the same conversations over and over again, like a goldfish circling a bowl.
But I stopped.
I remembered Sam Altman saying that younger people (age 25 and under) were starting to use ChatGPT more like an operating system, not just as a space to get answers.
Older folks, he said, weren’t using it well enough. Translation? The over 40 crowd used it like a Google search tool.
That line stayed with me.
ChatGPT Evolved
I decided right there that my age wasn’t going to define how I used ChatGPT.
And the solution I needed, as it turns out, was already sitting on Open AI’s platform, in a space called Projects right there in the Large Language Model.
ChatGPT had quietly evolved.
It was no longer just a place for random thoughts, and quick answers.
It had become something that could remember, organise, and grow alongside your work.
Yes! I know that sounds a little creepy. But honestly, a badly structured workflow is worse.
What ChatGPT Projects Actually Does
Projects are like little memory rooms inside ChatGPT.
Think about your desktop workspace, and how your work is organised into folders and subfolders.
Everything is easy to find, right? I hope you said yes because projects are like that, just with memory attached.
Each Project knows its role.
Each one holds context.
And each one builds on itself over time.
When I was building Amy, I had multiple Projects running at once.
Now that’s not as crazy as it sounds once you understand that each one had a different role.
- One kept track of my the build out strategy, the working parts of my automation flows.
- Another held notes for my system engineer: clean, separate, technical.
- And the last Project, because I can’t help myself, held the floating thoughts that jump around in my head. The big ideas that don’t always arrive neatly but I think (or hope) I can implement.
What I loved most about using Projects as a system wasn’t just the organisation. It was the continuity.
I didn’t have to reintroduce critical ideas all the time or remind ChatGPT what we were working on months ago.
There were less memory gaps. No continuous circling back.
I felt in flow and let me tell you, it was the first time I tasted how CHATGPT could feel like an operating system.
That’s when it clicked for me.
ChatGPT, and for that matter Claude and other LLMs aren’t just useful when you ask better questions.
They become super charged when you build a space that acts as your second brain. That’s what Projects was for me.
How I Built My Workflow Inside ChatGPT
If you’ve never used Projects before, setting one up is surprisingly simple.
You open ChatGPT, look to the left sidebar (at the time of writing this blog post), and you’ll see Projects sitting right under your chats.
You create a new Project, give it a name that makes sense : content, client work, a digital offer you’re building, menus you’re experimenting with.
Ok, stop for a second. I need you to do this. Before you hit create, you’ll see an option for memory.
When you toggle on project-only memory, you’re essentially telling ChatGPT:
“Everything that happens here stays here.”
That one decision matters more than most people realise because its what allows your work to build on itself and compound over time.
Now, what I’m about to explain next is the part most people either skip, or under estimate.
For a Project to really work, it needs Custom Instructions.
I’ve been working with Custom Instructions long before ChatGPT gave us a proper place to house them inside Projects.
No fancy-pants work. Just the same logic I’ve used for at least three years to get better output.
When you’re building a Project, you also need logic.
And by that I mean think carefully about what you’re building, the outcomes you want, the frameworks that will lead to the best outcomes and build accordingly.
ChatGPT Projects Simplified Step-By- Step
→ First, open ChatGPT.
→ On the left sidebar, you’ll see Projects (right under your chats).
→ Click New Project and give it a name that makes sense; maybe your content calendar, a client workspace, or a digital-offer hub.
→ Before you hit Create, look for More options → Memory.
→ If you see the toggle for Project-only memory, turn it on.
That tells ChatGPT, “Remember everything that happens in this space, but don’t mix it with my other work.”
→ Then click Create. That’s it.
Note: Make sure Memory is turned on in your Settings → Personalization → Memory before you start. If it’s off, ChatGPT won’t retain anything inside the Project.
Important: You can only set project-only memory when you first create the Project.
Once it’s created, you can’t switch. So think about how you want this Project to work before you hit ‘Create.’
The Four Parts You’ll Need To Set Up A ChatGPT Project
Let’s get nitty gritty for a second and set up a Project for copy writing or brain storming, anything that involves long form content.
This is especially useful when I want my writing in the AI to feel consistent and on brand.
I build my Project for this in 4 parts.
The idea is for them to work together to build one system inside a Project. First. I set the groundwork
I input personal information related to the specific hat I am wearing for the Project I’m working on.
Who I am, what I do, who I serve, what my community cares about.
This is no place to be vague because is ChatGPT does not understand the context everything will feel slightly off.
Next, I define how I want results to come back to me. Think tone, structure depth. This is where I define boundaries.
Remember how I talked about LLMs and how average they can be. Being average gets you nowhere. You need to get beyond the AI’s preference for pattern recognition and get it to focus on your POVs and preferences.
That only comes when you decide what to feed into Projects. Aka, the logic I referred earlier too.
Ok, the third thing I do is to get very honest about what I don’t want.
Words that feel generic. Phrases that make me cringe.
Styles that don’t match how I think or write: binaries, listicles, repetition, short sentences that all feel familiar because everyone is using them directly from AI.
I call this classic AI copy slop. Again, most people skip this part but I don’t.
Follow The Most Strategic User of ChatGPT Projects
The best users of LLMS know that moving friction is just as important as creating direction.
So consider yourself the orchestra director here within Projects
Anyway let’s talk about last part. I give it examples of what good looks like: my best work, pieces I admire and anything that reflects the standard I’m aiming for.
Again, inputing this is crucial.
Once those examples are inside the Project, ChatGPT starts to recognise your unique pattern without needing to be reminded.
People always ask how long Custom Instructions should be.
For me, it’s always about clarity rather than the number.
Still, don’t build your project frames with three sentences. That’s just way too vague.
If it’s 5,000 words, parts of it will get ignored.
Building your instructions with enough meat where everything is clear and intentional for the AI to grasp.
This is how ChatGPT Projects work best.
What Happens When ChatGPT Starts to Learn Your Work
What most people don’t realise is that building with ChatGPT Projects isn’t about instant mastery.
Like most things, it’s about building something that improves through use.
As AI moves to AGI and it gets better, faster, more superior things will improve but at the time of writing, it wont feel 100% percent perfect at the get go.
You’ll notice yourself correcting things. Adjusting tone. Tweaking structure.
If you’re doing this frequently, pay attention because whatever you’ve fixing and re-fixing belongs inside your instruction.
So go ahead and update those. With repeated tweaks something shifts.
You’ll start to notice your Projects getting sharper because they’ve learned your voice, preferences and way of thinking.
TBH, that’s how I feel when in flow because ChatGPT stops feeling like a place I visit to get answers and more like a brain that holds my operating system.
I’m still in charge.
I’m still the critical thinker but I have a second brain who I’ve trained.
That’s probably why I enjoy Projects so much. They feel less like folders.
And more like a digital studio that remembers what I’ve built.
A space that knows me, as much as I’ve taught it to do so.
And if you use it wisely and lock down your personal data (always be careful what you share) you start to realise something simple:
Your Projects are as smart as you make them to be.
It’s waiting for you to step in with instruction and intention.
And probably with a little more thought than what you’ve been giving your LLM.
Need a place to map your ideas, priorities, and next steps? My Business Reset Guide + Planner was designed to help you slow down, think well, plot your success path and execute. Click HERE or below 
