> I deleted all my chats with ChatGPT and moved on.
I figured it was the perfect time—before it gets too smart and takes all my data to form an opinion about me, when the robot takeover comes and it suddenly remembers every chat where I didn’t say “thank you.”
Within an hour, I received an email with a link to an archive containing all my data. Here’s what was inside:
Audio — all recordings of my conversations with ChatGPT in WAV format, organized into folders: first my question as a WAV file, then its response as a WAV file.
Photos / Images — about 1,000 images sitting directly in the root folder.
Images generated by ChatGPT for me — stored in a separate folder.
System files — containing my email, year of birth, phone number, and internal system ID.
A separate shopping file — if I had purchased anything via the new Operator functionality, it would have appeared there.
A separate file with shared conversations.
A separate file with information about group chats.
A separate file containing all conversations in JSON format — 320 MB of text.
A separate file containing all conversations in HTML format — also 320 MB of text.
Of course, opening an HTML file of that size is something only a person without ADHD would attempt.
So instead, I opened the folder in Gemini CLI (I’ve been liking it lately) — Claude Code would work too — and asked it to create a separate folder called Sorted, where it would parse everything into .md files and organize them into folders by conversation. Any projects that contained instructions were to be placed into a separate Projects directory (for me, that’s things like Work, Money, Health, etc.).
I didn’t use any special prompt — I just asked in plain English:
> Here is an extracted folder from ChatGPT. Is it possible to generate .md files for all extracted chats and organize them into topic-based folders, with everything placed under a main Sorted directory? Some chats may contain information related to specific projects, with or without explicit instructions. If so, can we create separate sections for these projects inside the Sorted folder. Additionally, let’s try to identify project instructions by searching for the following marker:HERE IS MY INSTRUCTION AS AN EXAMPLE
As a result, I now have a local copy of all my ChatGPT conversations.
Next, I’ll organize everything into folders and topics inside Obsidian. For now, I’ll browse through them and continue conversations using Gemini CLI, but once local models become smarter and less resource-hungry…
> Now that I have all my .md files, which are perfectly readable by any LLM (AI model), I can switch between Gemini, Claude, ChatGPT, and any new ones that come along like changing gloves—without losing my conversations or context—or even move entirely to my own private, offline setup.
I figured it was the perfect time—before it gets too smart and takes all my data to form an opinion about me, when the robot takeover comes and it suddenly remembers every chat where I didn’t say “thank you.”
Within an hour, I received an email with a link to an archive containing all my data. Here’s what was inside:
Audio — all recordings of my conversations with ChatGPT in WAV format, organized into folders: first my question as a WAV file, then its response as a WAV file.
Photos / Images — about 1,000 images sitting directly in the root folder.
Images generated by ChatGPT for me — stored in a separate folder.
System files — containing my email, year of birth, phone number, and internal system ID.
A separate shopping file — if I had purchased anything via the new Operator functionality, it would have appeared there.
A separate file with shared conversations.
A separate file with information about group chats.
A separate file containing all conversations in JSON format — 320 MB of text.
A separate file containing all conversations in HTML format — also 320 MB of text.
Of course, opening an HTML file of that size is something only a person without ADHD would attempt.
So instead, I opened the folder in Gemini CLI (I’ve been liking it lately) — Claude Code would work too — and asked it to create a separate folder called Sorted, where it would parse everything into .md files and organize them into folders by conversation. Any projects that contained instructions were to be placed into a separate Projects directory (for me, that’s things like Work, Money, Health, etc.).
I didn’t use any special prompt — I just asked in plain English:
> Here is an extracted folder from ChatGPT. Is it possible to generate .md files for all extracted chats and organize them into topic-based folders, with everything placed under a main Sorted directory? Some chats may contain information related to specific projects, with or without explicit instructions. If so, can we create separate sections for these projects inside the Sorted folder. Additionally, let’s try to identify project instructions by searching for the following marker:HERE IS MY INSTRUCTION AS AN EXAMPLE
As a result, I now have a local copy of all my ChatGPT conversations.
Next, I’ll organize everything into folders and topics inside Obsidian. For now, I’ll browse through them and continue conversations using Gemini CLI, but once local models become smarter and less resource-hungry…
> Now that I have all my .md files, which are perfectly readable by any LLM (AI model), I can switch between Gemini, Claude, ChatGPT, and any new ones that come along like changing gloves—without losing my conversations or context—or even move entirely to my own private, offline setup.
