How-to · Switching assistants without losing context

How do you move your AI memory from one assistant to another?

You do not. Platform memory does not leave the platform that made it. What moves is a record you own. A Pupul profile compiles into one block that any assistant capable of reading text can use, so switching models means handing over one record, not rebuilding months of context.

Start your profile free See the five steps

A personal profile is free to start. You choose who reads it, and you can take it back from any reader at any time, free, forever.

The problem

Platform memory is a house
you can never move out of.

Every major assistant now builds a memory of you. It is genuinely useful, right up to the day you want to use a different assistant. Then you discover the plain fact underneath the feature: platform memory does not leave the platform that made it.

01 / The lock

The memory belongs to the vendor's product

What ChatGPT remembers about you lives inside ChatGPT. What Gemini learns stays in Gemini. None of it is designed to be handed to a competitor, because keeping it is the point.

02 / The restart

Switching means starting over

A better model ships somewhere else and the real cost of trying it is not the subscription. It is re-explaining your work, your family, your projects, and your preferences from a blank chat box.

03 / The fix

Keep the memory outside all of them

A Pupul profile is a record you own, held outside every assistant. It compiles into a block any assistant that reads text can use. The models compete. Your context does not reset.

The steps

Five steps, one record, every model.

A Pupul profile is a dated, append-only record of you that lives in your hands. Here is how it becomes portable memory.

1

Start a profile you own

Open the personal room and create your profile. A personal profile is free to start, and it starts honestly small, because it is made of time that actually passed. There is no shortcut to a long record, which is exactly what makes a long one worth something.

2

Add entries as your life happens

Write entries as they happen. Each one is dated the day it lands. You can bring old writing into your profile and keep it there, so nothing you have already written is wasted, but imported history never counts toward how long the profile has really existed. The record's age is real or it is nothing.

3

Compile the profile into one block

When you want an assistant caught up, the profile compiles into a plain block of text. No vendor format, no export wizard, no API key. If a model can read text, it can read you.

4

Hand the block to whichever assistant you are using today

The same profile reads in Claude, ChatGPT, Gemini, Copilot, and Perplexity. Paste it at the start of a conversation and the model opens already knowing what months of chats would otherwise have to teach it. Try a new model on Tuesday, go back on Wednesday. The record does not care which one is winning this quarter.

5

Take it back whenever you want

You choose when to hand the profile over, and you can take it back from any reader at any time, free, forever. Portability without revocation is just leakage with better marketing. This is both.

Why this works

The memory that moves is the one nobody's product owns.

Every platform memory feature answers the question "how do we know you better than our competitor does?" A Pupul profile answers a different question: "what is true about you, on the record, over time?" That answer is useful to any model, which is why it can travel between all of them.

The definition worth keeping: a portable AI memory is a record of you that you own, that any assistant can read, and that you can take back. Platform memory fails the first and last test by design. A document on your laptop passes the first test but nobody vouches for when it was written. A Pupul profile passes all three, because entries are dated the moment they land and can never be edited or reordered, by anyone, including us.

This also changes what switching costs. Today the real switching cost between assistants is not price, it is accumulated context. When the context lives in a record you carry, the cost of trying the next model drops to the length of one paste. That keeps every vendor honest, which is the whole idea.

FAQ

The questions people actually ask.

Can I export ChatGPT's memory and import it into Claude or Gemini?
Not in any way the platforms support as a feature. Platform memory does not leave the platform that made it. The workable pattern is to keep the memory outside all of them: a Pupul profile compiles into a block that reads in Claude, ChatGPT, Gemini, Copilot, and Perplexity alike.
What does a Pupul profile cost?
A personal profile is free to start. Taking your profile back from any reader is free, forever, and so is deleting it.
Can I import my old journals and chat history?
Yes. You can bring old writing into your profile and keep it there. But imported history never counts toward how long the profile has really existed. Age is measured in time that actually passed, which is what keeps every profile's age honest.
What happens when I want an assistant to stop reading my profile?
You take it back. You choose when to hand the profile over, and you can revoke any reader at any time, at no cost, forever.
Why does my profile start small?
Because it is made of time that actually passed. Entries are dated the day they land and cannot be backdated, so a new profile is honestly new. That is the property that makes an old one meaningful.
Start today

Own the memory. Rent the models.

Start your profile now and it starts aging now. Free to start, yours forever, readable by every assistant you will ever try.

Start your profile free