The appeal is undeniable: never forget anything important ever again. The method is practical: using the phone you carry with you everywhere you go. And the results are promising, if not a full delivery on the promise. It’s called TwinMind, and it’s an AI-powered “second brain” to help our aging, tired, and forgetful first brains capture and remember everything important.
“We think of TwinMind like Jarvis from Iron Man or a second brain,” CEO Daniel George told me on the TechFirst podcast.”The idea is to have an AI that just understands your whole life and can help you in real time proactively before you even ask.”
I’ve been testing the TwinMind app in pre-release beta for several weeks now. Turn it on when you are in a meeting or having important conversations, and TwinMind will record and transcribe everything that’s going on. Check it later, and it’ll provide a clean and efficient summary of the conversation, along with suggested action item for follow-up. You can check the full summary for all the details, or you can even query the recorded knowledge with detailed questions.
Just one example: I turned it on during a meeting with a client. Checking the record in TwinMind today, I find a pre-suggested question: “How does competitor analysis vary in Korea and China compared to elsewhere?”
Tapping that suggested query, I get a comprehensive summary of the insights from the hour-long meeting on specifically that topic, including an overview of the limitations of the insights provided based on the amount of data shared in the meeting.
Impressive, to say the least.
Having a powerful AI friend walk around with you everywhere you go is not a new idea. It’s recently been recently tried with the Rabbit R1, and the AI Pin, recently sold off for parts to HP. But using a separate piece of hardware for the purpose means carrying around yet another thing, charging it, sometimes paying for a cellular subscription for it, and learning a new user interface. The markets have essentially spoken at this point and said that’s a non-starter.
So TwinMind runs as an app on your phone. George says it’s so efficient, running on Apple Silicon on iOS, that it doesn’t drain your batter excessively, even if you use it all day. The iOS app optimized can over 12 hours a day without draining the battery, TwinMind says. (I haven’t personally tested it for that long.)
Because TwinMind runs on-device, again using Apple Silicon’s AI cores, it doesn’t send recordings or data to the cloud by default. It uses several local LLMs for transcription and intelligence, switching between various flavors for different tasks. If you wish, you can encrypt your data with Apple’s security protocols, and if you want a backup, you can choose to store backups in iCloud: again relying on Apple’s high-quality privacy and security protocols. A year’s worth of transcribe data would only occupy 100 megabytes, TwinMind says.
As a paid extra, you can enlist high-level AI engines for tougher tasks, George says, like ChatGPT, Gemini, or Claude. A paid version will be $15/month.
There’s multiple reasons for keeping most things running on-device, and cost is one of them.
“That’s how we were able to make this actually financially feasible because the costs are 20 times lower if you do most things on device, using AI models that run on the edge and then only send the tough queries that it’s hard to do with a tiny model to the cloud,” says George.
The added benefit to this edge AI set-up: if you don’t have internet access, TwinMind still works.
Privacy is another benefit to keeping things on-device:
“Unless you choose to back it up, everything can stay locally on your phone,” George says. “It can transcribe locally, and even you ask a question to an LLM on the cloud, it doesn’t get retained on the cloud, and your memory just stays on your phone.”
90% of our memories are lost within a week, TwinMind says.
While that’s probably good in some ways–wetware garbage collection so we don’t remember every mundane detail–we also lose important things, like dates and tasks and context for big decisions.
This is one way to keep them organized and retained.
Over the long term, our personal AIs could grow to help us manage our finances, home responsibilities: a personal assistant that remembers everything. And, importantly, is under our own control: not something that is engineered to make us engage more or pay more.
