How to switch from ChatGPT to Claude without losing context
You switch AI models for one reason. Another one got better at something you need. Better at code. Better at long documents. Better at your language.
Then you open the new one and it asks, "How can I help you today?" It knows nothing about the project you have been building for three weeks. So you start explaining. Again.
The problem is not the new model. The problem is that your project only ever lived inside the old conversation.
Fix that, and switching takes ten seconds.
Put the project in a folder, not a chat
A PREP folder is three things in your cloud drive:
PREP.md, what the project is, where everything lives, the rules.LOG.md, what happened, in order.memory/, snapshots of past sessions.
That is the whole idea. The permanent stuff lives in files. The chat stays disposable.
The actual switch
Here is the move, start to finish.
In ChatGPT, when you have done good work, you save it with
PREP Save. You paste the conversation's summary,
and PREP Save writes a clean snapshot straight into your project's
memory/ folder in your own Google Drive: the decisions, the current
state, what comes next. Nothing is stored on our side.
Now you open Claude and tell it where to look:
In my Google Drive, open the Sunday Sourdough folder inside PREP and read PREP.md.
Claude reads PREP.md, skims the log, opens the latest snapshot. It now
knows what you knew. You keep going from the exact sentence you stopped on.
No 40-minute re-explanation. No pasting a giant transcript. No forgotten decision.
Why pasting the old chat does not work
A three-week conversation is mostly noise. Jokes, dead ends, things you tried and dropped. The new model has to pan for a few grams of gold in all that gravel, and it will miss some. A snapshot is the gold, already separated. That is the difference.
The shift that makes it click
Most people try to preserve conversations. Stop. Preserve the project instead.
Once the important things live in the folder, the chat is free to be temporary. Delete it. Start fresh every morning. Nothing is lost, because nothing important was in the chat to begin with.
This is not AI memory
ChatGPT and Claude both have built-in memory now. It is useful, and it solves a different problem. Built-in memory remembers you. Your tone, your preferences. It stays with that one vendor.
PREP remembers the project, and it belongs to you. You can hand it to another model, another person, or your future self. Your writing style belongs in memory. Your roadmap belongs in the folder.
Try it once
Take one real project. Work in ChatGPT today, and save it with
PREP Save when you stop. Open Claude tomorrow and
tell it to open that folder in your Google Drive and read PREP.md.
The first time a different AI picks up mid-thought, you stop worrying about which model to marry. You just use whichever one is best this week.
Conversations come and go. Models come and go. Your project should not.
Save your first project with PREP Save →, or read the spec to see how the folder works underneath.