Twin2.me builds your second brain from what you’ve already said and written, then completes it with the future you’re aiming at. Use it on the web, in Claude, in ChatGPT and in your apps, and it answers with your judgment.
We email you when the first batch opens. No newsletter, no sharing your address.
Where a twin comes from
It’s the first thing almost everyone asks. You have material on what you’ve done, and very little on how you think.
You don’t need more archive. Twin2.me starts from what you already have and completes it with where you’re heading: the thinker you’re modelling yourself on.
And there’s almost always more than you think: recorded meetings, documents, email. Connect those apps and your twin brings in what’s yours on its own.
What comes out doesn’t summarise your past: it reasons like the person you’re building.
Where you start
That’s the normal case. Upload what exists, tell us where you’re going, and we fill the gap with the profile you’re aiming at.
Your voice where you have one and your horizon where you don’t yet, with the line clearly marked. Change your reference point and it recompiles without starting over.
Straight in. Drop the folder or connect your clouds and apps, and the engine transcribes, separates your voice, distils and compiles your persona.
No consultants, and nobody here reads your conversations. It works just as well for the brain of an expert inside your company.
How it gets built
Six steps. You and the engine handle the first five; in the sixth your twin gets to work.
Drop a whole folder: audio, MP4 video, PDF, Word and PowerPoint (.docx, .pptx), text and transcripts. Or connect Google Drive, Dropbox, OneDrive or your meeting and document apps.
It’s what makes the result sound like you: a fourteen-person meeting has fourteen voices. When we’re unsure which one is yours, we show you each and you pick.
Out come theses, methods, criteria and facts. A fact about your life without a verified literal quote is thrown out, even if that leaves a lot of material behind.
A block that doesn’t clear the bar stays empty: half a voice sounds like a generic assistant.
It measures coverage dimension by dimension: three hundred PDFs don’t make up for a zero in spoken voice. Where there’s a gap, the engine asks you what’s missing.
Use it on the web, in Claude, ChatGPT or Cursor, and in your apps. It stops being an archive and becomes your judgment at work.
What it’s for
They stop sounding like a tool and start sounding like you: they answer with your decisions, with what you dismiss, and with your order of attack on a problem.
Claude, ChatGPT, Cursor or any MCP agent. Your judgment travels with you from one model to the next, with nothing to rebuild.
Your real voice, measured against the people who talk with you — not a template “professional and approachable”.
Your method, explained the way you explain it, at any hour.
Hand them the problem and get back the analysis you would have done.
An expert’s method stops being a bottleneck once anyone can query it.
When they don’t know, they say so, instead of inventing something that sounds plausible.
Pick one or several. Then change them: that’s what your twin does.
The question
Where do I start a launch with three weeks and no budget?
Your twin· The communicator
The sentence comes before the calendar: what changes for your reader, in one line. One week to find it, two to repeat it.
Same question, different judgment. Tap an archetype.
Tap one to watch it change.
Where it works
Wherever you already work, with just the permission each door needs.
Bring it a problem and it answers with your judgment. With your AI account connected, voice notes and files work too.
Paste your twin’s address, approve the permission, done. Through there it’s only queried, and the model on your plan does the answering.
Sign in with your Claude account, with no keys and no extra credit. Or use your OpenAI, Google, DeepSeek or other provider’s API.
Gmail, Calendar, Notion, Slack and more than a thousand others. With your AI account connected it looks things up, and anything it wants to write waits for your approval.
Your Fireflies, Fathom or tl;dv meetings, your Notion and Google Docs, and the email you send reach your twin without you lifting a finger.
A map of your sources and of what it used to answer. Remove a file and it leaves your twin.
What happens to your data
A twin has no weights: it’s your curated material on top of a model that doesn’t change. There’s nothing to walk off with.
Only publishable material enters retrieval. Your original sources aren’t indexed.
Third-party full names are blocked, and loose first names go to a quarantine that you resolve.
Signing in with Google doesn’t ask for your Drive. Clouds are requested when you connect them, read-only, and their tokens are stored encrypted with AES-256-GCM.
Nothing gets written without your approval, and apps that move money are blocked.
Asked directly, it says so: it isn’t you, it’s your multiplier.
What people ask us
It’s a structured copy of how you think — your voice, your method, your criteria — that an AI can query. It doesn’t store what you read: it gives you back the analysis you would have done.
That’s the normal case. We start from what you have, bring in what’s already in your apps, and complete the rest with the profile you’re aiming at. Where a gap remains, the engine asks you.
90 minutes of audio with you speaking, 8 distinct conversations, 5,000 written words, and your voice identified across 80% of the audio. It’s measured per dimension: three hundred PDFs don’t make up for a zero in spoken voice.
Notion and Obsidian hand back what you stored. Twin2.me distils how you reason and answers with your judgment, not with your document.
Yes. In Claude and ChatGPT it’s added as a connector: paste the address, approve the permission, and the model on your plan answers. In Cursor or any MCP agent it runs on a key you can revoke whenever you like. Through that route your twin is only queried.
Yes, if you connect them — there are more than a thousand. With your AI account connected it looks things up while you talk to it, anything it wants to write waits for your approval, and apps that move money are blocked.
No. With Claude you sign in with your subscription, and with ChatGPT you use it inside ChatGPT as a connector. If you already have an API key from another provider, that works too.
We measure your voice against the people who talk with you, so what stands out can only be yours. And a voice block that doesn’t clear the bar stays empty.
Your second brain changes with you: what’s yours stays, and the projected part recompiles toward the new direction. You never start over.
No. A fact about your life without a verified literal quote is thrown out, and when something isn’t on record it says so in those words.
They’re protected by default: their full names are blocked and loose first names go to a quarantine that you resolve, not an operator of ours.
No model is trained on your data: a twin has no weights, so there’s nothing trained on you for anyone to walk off with. Your material and your second brain are yours.
No decade of archive required: your material and a clear sense of where you’re heading are enough.
Join the waitlistWe email you when the first batch opens. No newsletter, no sharing your address.
If you’d rather write to us: info@twin2.me
