Submitted to the Google Cloud NEXT '26 Writing Challenge
The hook
Google just shipped one-click memory import from ChatGPT into Gemini at Cloud Next '26.
I've been trying to vibe-code my way to this exact workflow for months. Exports, parsers, custom ZIP handlers, half-broken browser extensions. Thousands of tokens burned on prompt gymnastics to stitch my own history together.
And then Google just… did it. Clean, native, one click.
Thank you. Seriously. This is the feature a lot of us have been quietly hoping for, and it just dropped.
But importing your history is step one. The real leverage is what you do with it once Gemini has it. So this post is two things: a thank-you to the Gemini team, and a practical guide — five workflows I've been wanting for months that now just work.
Why this matters more than it looks
Your ChatGPT history isn't chat logs. It's a record of how you think, what you obsess over, how you phrase things, and what you've already solved. Most people treat it as disposable. It's actually the closest thing to a portable brain snapshot that exists.
Until last week, that snapshot was locked inside one product. Now Gemini can read it. That changes what an AI assistant can actually be — not a tool you re-introduce yourself to every session, but one that already knows your voice.
Five ways to profit from it once you've imported
After importing, open a fresh Gemini chat and try these. I call them bootloader prompts — they spin Gemini up into a specific useful mode using the memory it just gained.
1. The voice profile
Read across my imported ChatGPT history and extract my voice profile.
Tone, sentence length, words I overuse, words I never use, how I open
and close ideas. Return it as a reusable style guide I can paste into
future prompts.
Now every piece of writing you generate with Gemini — emails, posts, drafts — can be pinned to a style guide built from the real you, not a generic "professional tone."
2. The unfinished-ideas miner
Scan my imported history for ideas I started but never finished.
Half-built product concepts, essay drafts, business ideas, technical
designs. Rank them by how many times I came back to them. Return the
top 10 with a one-paragraph summary each.
You'll be shocked. Mine surfaced three ideas I'd forgotten I'd had, one of which turned into a product I'm shipping right now.
3. The pattern recognizer
Based on my imported history, what topics do I keep circling? What
problems do I solve over and over in slightly different ways? What
blind spots show up — subjects I avoid, skills I never ask about?
This one is humbling. It's a mirror. It tells you what you actually care about versus what you say you care about.
4. The personal SOP writer
Look at my imported history. Any time I asked for help with [task type:
e.g. debugging, cold emails, PR reviews], extract the pattern and write
me a standard operating procedure. Include prompts I've already proven
work for me.
Your own best prompts, promoted to reusable templates. This is how you stop re-inventing the wheel every session.
5. The decision archaeologist
Find every major decision I talked through with ChatGPT in the last
year — product direction, career moves, relationships, finances. For
each, summarize what I was considering, what I chose, and what my
reasoning was at the time.
A year of decisions, written down by the AI that helped you make them. Useful for reflection, accountability, and noticing your own patterns.
Why the Gemini team nailed this
Three design choices I want to call out because they're easy to miss:
ZIP-based import instead of live OAuth. This was the right move. It puts the user in control of what gets shared. You can review the file before upload. It also sidesteps every single privacy and platform-lock concern that would've killed an OAuth bridge.
Memory, not raw logs. Gemini is ingesting your history as retrievable memory, not just dumping it into context. That means it scales — your whole history isn't eating your token budget on every turn.
Shipping it at NEXT, not quietly. Putting this on the main stage signals that Google actually believes AI memory portability is a user right, not a moat. That's a cultural win for the whole ecosystem.
A small personal note
I built a little browser extension in three days with Codex that does roughly the same export step on the client side — it was my hacky answer to not wanting to wait 21 days for OpenAI's native export. I'm going to keep building it, because I think there's room for tools that let people do even more with these archives locally.
But honestly? Google's version is cleaner for most people. If you just want your history in Gemini, use the official import. It's a better experience than anything I or anyone else was going to ship this year.
Lowkey shoutout, elgoog.
The prompts again, copy-paste ready
- Voice profile: Extract my tone, vocabulary, and style from my imported history.
- Unfinished ideas: Find half-baked ideas I kept coming back to.
- Pattern recognizer: Show me what I circle on and what I avoid.
- Personal SOPs: Turn my proven prompts into reusable templates.
- Decision archaeologist: Summarize my year of big decisions with reasoning.
Import once. Run these five. Thank me later. Actually, thank the Gemini team.
If you've got your own bootloader prompts for the new import feature, drop them in the comments. The whole point of these unlocks is compounding them together.
Top comments (0)