5 ChatGPT projects to build today (with project memory)

Before everyone figures this out

Jun 18, 2025

Over the next few weeks, I'm going to show you how to transform ChatGPT from a fancy chatbot into your operating system.

Today, we start with a 5-project framework you can adopt. A framework that will get more powerful as ChatGPT’s project memory improves. (Or a framework you can adapt for Claude if they ever release a similar feature.)

Most people waste their projects on vanity builds that sound cool but deliver nothing. They'll create a "Twitter Helper" and an "Email Assistant" and 17 other micro-projects that fragment their knowledge and destroy their productivity.

That's not a system. That's digital hoarding with extra steps.

Important: ChatGPT's memory isn't perfect yet—but it's only going to get better with time. The people building their systems now will have compound advantage when the technology improves. Start building today, refine as it evolves.

The memory supercharge nobody's talking about

Before we build anything, you need to understand the feature hiding in your settings: project-level memory.

ChatGPT can now remember every conversation within a project. Not just the current chat.

But only if you turn it on.

(Settings → Personalization → Reference saved memories ON, Reference chat history ON)

**A word of warning from my own experience:** I've noticed projects sometimes try to reference memories from your general ChatGPT chats instead of staying focused on project-specific knowledge. It's not ideal, but here's my current workaround: I only keep work-related conversations in ChatGPT and regularly audit and clear out old memories and chats that aren't relevant. Think of it as keeping a clean workspace—the cleaner your general ChatGPT environment, the better your projects perform. This isn't perfect, but it works until OpenAI (hopefully) fixes the memory isolation between projects.

This means your projects get smarter with every interaction. They build institutional knowledge. They become actual AI employees that remember their training.

Miss this step? You're building amnesia machines.

The 5-project framework

You could build 50 projects if you wanted. But that's exactly the trap—infinite options lead to infinite waste and confusion for your LLM.

Here's a potential sequence you can build if starting from zero today:

Project 1: The brain dump

Your overflow valve for everything

This is your AI scratch pad. Every system needs a junk drawer.

→ Purpose: Capture all the random thoughts, ideas, and information that doesn't fit elsewhere yet

→ Why first: You need somewhere to put things while you figure out your system. This project teaches ChatGPT how you think when you're unfiltered.

→ Files to upload: None initially. This is your collection point.

What makes it work: The lack of structure IS the structure. Your brain doesn't think in neat categories when you're creating.

Project 2: The daily driver

Where most your actual work happens

This isn't specialized. It's your default. Open ChatGPT? This is where you land.

→ Purpose: Your primary workspace for regular tasks, writing, analysis, and problem-solving

→ Why second: Once you know what you're collecting (Project 1), you can build your workhorse that handles the majority of your work

→ Files to upload:

Your voice DNA (how you write/speak)
Examples of your best work (or entire archive)
Key templates you use regularly

Memory maximizer: Every interaction here builds your AI's understanding of your work patterns

Project 3: The research engine

Your competitive intelligence machine

While others Google, you're building intelligence that remembers.

→ Purpose: Deep research, market analysis, competitor tracking, and trend identification

→ Why third: Information gathering feeds everything else you create. You need data before you can create insights.

→ Files to upload:

Industry reports
Competitor analyses
Market data
Your research templates

The compound effect: Each research session builds on the last.

Project 4: The creation studio

Content production at scale

One idea becomes 20 pieces of content. Not because you're recycling, but because you're systematically exploring every angle.

→ Purpose: Transform ideas into published content across all platforms

→ Why fourth: Now that you can think (1), work (2), and research (3), you can create with context

→ Files to upload:

Your best content examples (organized by platform)
Content frameworks and templates
Brand voice guidelines
Audience research

Scale secret: Upload your entire content archive. Your AI learns from your successes.

Project 5: The learning lab

Your skill development accelerator

Most people use ChatGPT to avoid learning. This project makes you dangerously good at getting better.

→ Purpose: Structured learning, skill development, and knowledge synthesis

→ Why fifth: Systems for growth come after systems for production

→ Files to upload:

Learning resources you're working through
Notes from courses/books
Your learning goals document
Progress tracking templates

Hidden benefit: This project identifies your learning patterns and optimizes future learning

The setup sequence

If you’re a beginner, there’s no need to build all five at once.

Here's a potential timeline:

Today (Hour 1):

Create all 5 projects with these exact names
Add instructions ONLY to Brain Dump and Daily Driver

Next 7 days:

Use only Brain Dump and Daily Driver
Let them learn your patterns
Don't touch the other three

Week 2:

Set up Research Engine when you need to research something
Use it for every research task that week

Week 3:

Configure Creation Studio with accumulated knowledge
Start feeding it your content archive

Week 4:

Build Learning Lab with insights from your usage patterns
Upload current learning materials

What not to build (yet)

Resist the urge to create:

Single-task projects ("Email Writer," "Tweet Helper")
Platform-specific projects ("LinkedIn Project," "YouTube Project")
Theoretical projects you might use "someday"
Separate projects for every client or topic

These are the vanity metrics of AI productivity. Build systems, not features.

The file limit

Each project can hold 20 files in its knowledge base. That's your real constraint.

Use it wisely:

Combine related documents
Delete outdated files regularly
Focus on quality references, not quantity

Twenty files of your best work beats 200 random documents every time.

The memory multiplier effect

Here's what can happen after 30 days with this setup:

Your Brain Dump becomes a pattern detector for your best ideas
Your Daily Driver knows your voice better than your editor
Your Research Engine builds a knowledge graph of your industry
Your Creation Studio can produce in minutes what used to take hours
Your Learning Lab has identified exactly how you learn fastest

☑️ Action steps to take right now

Check your memory settings: Settings → Personalization → turn on BOTH memory options (Reference saved memories ON, Reference chat history ON)
Create your first two projects:
- "Brain Dump"
- "Daily Driver"
Upload one document to Daily Driver—your best example of recent writing
Use ONLY these two projects for all your ChatGPT work today
Bookmark this newsletter to set up projects 3-5 next