Three former Google X scientists purpose to offer you a second mind nearly — not within the sci-fi or chip-in-your-head sense — however by an AI-powered app that good points context by listening to all the pieces you say within the background. Their startup, TwinMind, has raised $5.7 million in seed funding and launched an Android model, together with a brand new AI speech mannequin. It additionally has an iPhone model.
Co-founded in March 2024 by Daniel George (CEO) and his former Google X colleagues Sunny Tang and Mahi Karim (each CTOs), TwinMind runs within the background, capturing ambient speech (with consumer permission) to construct a private information graph.
By turning spoken ideas, conferences, lectures and conversations into structured reminiscence, the app can generate AI-powered notes, to-dos, and solutions. It really works offline, processes audio in real-time to transcribe on-device, and may seize audio constantly for 16–17 hours with out draining the gadget’s battery, the founders say. The app may also again up consumer knowledge so conversations will be recovered if the gadget is misplaced, although customers can opt-out of that. It additionally helps real-time translation in over 100 languages.
TwinMind differentiates itself from AI assembly note-takers like Otter, Granola, and Fireflies by capturing audio passively within the background all day. To make this potential, the staff constructed a low-level service in pure Swift that runs natively on the iPhone. In distinction, many opponents use React Native and depend on cloud-based processing, which Apple restricts from operating within the background for prolonged intervals, George mentioned in an unique interview.
“We spent about six to seven months final yr simply perfecting this audio seize constantly and getting there to search out a variety of hacks round Apple’s walled backyard,” he instructed TechCrunch.
George left Google X in 2020 and bought the concept for TwinMind in 2023 when was working at JPMorgan as Vice President and Utilized AI Lead, attending back-to-back conferences every day. To avoid wasting time, he constructed a script that captured audio, transcribed it on his iPad, and fed it into ChatGPT — which started to know his initiatives and even generate usable code. Impressed by the outcomes, he shared it with associates and posted about it on Blind, the place others confirmed curiosity however didn’t need one thing operating on their work laptops. That led him to construct an app that might run on a private cellphone, quietly listening throughout conferences to collect helpful context.
Along with the mobile app, TwinMind affords a Chrome extension that gathers extra context by browser exercise. Utilizing imaginative and prescient AI, it could possibly visually scan open tabs and interpret content material from numerous platforms, together with electronic mail, Slack, and Notion.
Techcrunch occasion
San Francisco
|
October 27-29, 2025
The startup even used the extension itself to shortlist interns from over 850 functions they acquired this summer season.
“We opened all of the LinkedIn profiles and CVs of the 854 candidates in browser tabs, then requested the Chrome extension to rank one of the best candidates,” George mentioned. “It did a implausible job — that’s how we employed our ultimate 4 interns.”
He famous that present AI chatbots — together with OpenAI’s ChatGPT and Anthropic’s Claude — can not simply course of a whole lot of paperwork or parse sign-ups from instruments like LinkedIn or Gmail to collect contextual info. Equally, AI-powered browsers resembling these from Perplexity and The Browser Firm lack the power to construct information out of your offline conversations and in-person conferences.
The startup at present has over 30,000 customers, with about 15,000 of them energetic every month. As a lot as 20–30% of TwinMind customers additionally use the Chrome extension, George mentioned.
Whereas the U.S. is the most important base for TwinMind to date, the startup can also be seeing traction from India, Brazil, the Philippines, Ethiopia, Kenya, and Europe.
TwinMind targets the final viewers, though 50–60% of its customers are at present professionals, about 25% are college students, and the remaining 20–25% are people utilizing it for private functions.
George instructed TechCrunch that his father is among the many people utilizing TwinMind to put in writing their autobiography.
One among AI’s vital drawbacks is its potential to compromise consumer privateness. However George asserted that TwinMind doesn’t prepare its fashions on consumer knowledge and is designed to work with out sending recordings to the cloud. In contrast to many different AI note-taking apps, TwinMind doesn’t let customers entry audio recordings later — the audio is deleted on the fly — whereas solely the transcribed textual content is saved regionally within the app, he famous.
Google X expertise helped pace issues up
The TwinMind co-founders spent just a few years engaged on numerous initiatives at Google X. George instructed TechCrunch that he labored on six initiatives alone, together with iyO — the staff behind AI-powered earbuds, which not too long ago made headlines for suing OpenAI and Jony Ive. That have helped the TwinMind staff transfer shortly from idea to product.
“Google X was really the proper place to organize for beginning your individual firm,” mentioned George. “There are round 30 to 40 startup-like initiatives taking place at any given time. No one else will get to work at six early-stage startups over two or three years earlier than launching their very own — a minimum of not in such a brief span.”
Earlier than becoming a member of Google, George labored on making use of deep studying to gravitational wave astrophysics as a part of the Nobel Prize–successful LIGO group on the College of Illinois’ Nationwide Heart for Supercomputing Purposes. He had accomplished his PhD in AI for astrophysics in only one yr — on the age of 24 — a feat that led him to affix Stephen Wolfram’s analysis lab in 2017 as a deep studying and AI researcher.
That early reference to Wolfram got here full circle years later — he ended up writing the primary test for TwinMind, marking his first-ever funding in a startup. The latest seed spherical was led by Streamlined Ventures, with participation from Sequoia Capital and different buyers, together with Wolfram. The spherical values TwinMind at $60 million post-money.
TwinMind Ear-3 mannequin
Along with its apps and browser extension, TwinMind has additionally launched the TwinMind Ear-3 mannequin, a successor to its current Ear-2, which helps over 140 languages worldwide and has a phrase error charge of 5.26%, the startup mentioned. The brand new mannequin may also acknowledge totally different audio system in a dialog and has a speaker diarization error charge of three.8%.
The brand new AI mannequin is a fine-tuned mix of a number of open-source fashions, skilled on a curated set of human-annotated web knowledge — together with podcasts, movies, and films.
“We discovered that the extra languages you assist, the higher the mannequin will get at understanding accents and regional dialects as a result of it’s coaching on a broader vary of audio system,” George mentioned.
The mannequin prices $0.23/ hour and will be available through an API to builders and enterprises over the subsequent few weeks.
The Ear-3, not like the Ear-2, doesn’t assist a whole offline expertise, as it’s bigger in measurement and runs on the cloud. Nonetheless, the app mechanically switches to Ear-2 if the web goes away after which strikes again to Ear-3 when it’s again, George mentioned.
With the introduction of the Ear-3, TwinMind now affords a Professional subscription at $15/month, with a bigger context window of as much as 2 million tokens and electronic mail assist inside 24 hours. Nonetheless, the free model nonetheless exists with all the prevailing options together with limitless hours of transcriptions and on-device speech recognition.
The startup at present has a staff of 11 members. It plans to rent just a few designers to boost its consumer expertise and arrange a enterprise improvement staff to promote its API. Moreover, there are plans to spend some cash on buying new customers.