Claude finally fixed what ChatGPT broke

Project-scoped memory that doesn't bleed into other clients. Here's how to turn it on...

Anthropic just dropped the feature I've been waiting a long time for.

Project-scoped memory.

And they did it so much better than ChatGPT that it's almost embarrassing for OpenAI.

Here's why I turned OFF ChatGPT's memory

ChatGPT's memory is a mess for anyone with multiple clients.

Your memories bleed across projects. Client A's requirements show up in Client B's work. You delete old chats, but the memories persist like ghosts haunting your new projects.

It's unusable if you're managing more than one context.

Claude took a completely different approach

Project-scoped memory by referencing the actual chats.

This means:

• Memory stays within each project

• No bleeding between clients

• You explicitly control when it's accessed

• Past chats only referenced when YOU ask

• Zero contamination between contexts

It's like having a separate brain for each client.

Here’s the big detail nobody's talking about

Claude doesn't build a persistent user profile.

Unlike ChatGPT—which tries to "learn" about you across everything—Claude only retrieves past chats when you specifically ask. No automatic memory injection. No surprise references to deleted work.

You stay in control.

How to turn Claude memory on right now

Takes literally 10 seconds:

1. Click Settings (bottom left in Claude)

2. Find "Conversation preferences"

3. Toggle "Search and reference chats"

4. Done

That's it. Memory activated.

I have separate projects for things like:

• Each client's writing system

• My newsletter writing

• Social media systems

• Cold outreach templates

• Demo builds

• My AI WriterOps training program

When I ask "what have we worked on?" in any project, Claude ONLY searches that project's history.

Just clean, contextual memory exactly where you need it.

This creates an advantage for you

While everyone else is fighting with ChatGPT's messy memory or—worse—starting fresh every chat, you're learning how to build and compound context in isolated environments.

Your systems get more refined. Your outputs get more consistent.

We're heading into the agentic era where AI won't just remember—it'll actively manage entire operations. This means it’s important to begin treating your interactions as part of a larger system, not just one-off chats.

Most people won't realize this for another year.

But you're not most people, right?

— Alex

Founder of AI WriterOps | AI Disruptor