Former Google X Scientists Develop AI App to Function as a ‘Second Brain’
TwinMind, a startup founded in March 2024 by ex-Google X researchers Daniel George, Sunny Tang, and Mahi Karim, aims to revolutionize personal productivity by creating an AI-powered app that continuously listens to ambient speech and organizes it into a structured personal knowledge graph. The company recently secured $5.7 million in seed funding led by Streamlined Ventures, with contributions from Sequoia Capital and AI researcher Stephen Wolfram.
AI-Powered Continuous Speech Capture and Contextual Memory
Unlike traditional AI note-taking tools that activate only during meetings, TwinMind operates passively in the background, capturing spoken thoughts, meetings, lectures, and conversations with user consent. The app transcribes this audio locally in real-time, enabling offline functionality and preserving user privacy by deleting audio immediately after transcription. It can record continuously for up to 16–17 hours without significant battery drain, leveraging a native Swift service on iPhones to circumvent Apple’s restrictions on background app activity.
Users benefit from AI-generated notes, to-do lists, and contextual answers derived from their personal audio data. Additionally, TwinMind supports real-time translation in over 100 languages and offers a Chrome extension that collects contextual information from browser activity, scanning content from platforms like email, Slack, and Notion using vision AI.
Unique Differentiators and Privacy Considerations
TwinMind distinguishes itself from competitors such as Otter and Fireflies by its continuous, passive audio capture and on-device processing, minimizing reliance on cloud services. CEO Daniel George emphasized that the company does not train its AI models on user data, and audio recordings are not stored, addressing key privacy concerns.
Founders’ Background and Funding
The founders’ experience at Google X, where they worked on multiple early-stage projects including AI-powered earbuds, accelerated the development of TwinMind. Daniel George’s prior work in AI for astrophysics and his connection with Stephen Wolfram led to Wolfram becoming the startup’s first investor. The seed round values TwinMind at $60 million post-money.
Introducing the Ear-3 AI Speech Model
TwinMind has launched the Ear-3 model, enhancing language support to over 140 languages and improving speaker diarization accuracy to 3.8%. This cloud-based model supplements the existing offline Ear-2 model by providing more robust recognition capabilities, automatically switching between models depending on connectivity. The Ear-3 will be accessible via API for developers and enterprises at a rate of $0.23 per hour.
User Base and Future Plans
Currently, TwinMind serves more than 30,000 users globally, with approximately 15,000 active monthly users. Its user demographic includes professionals (50–60%), students (25%), and general consumers (20–25%). The company offers a freemium model with unlimited transcription hours and a Pro subscription at $15 per month, providing expanded features such as a larger context window and prioritized support.
TwinMind plans to expand its team, focusing on design improvements and business development to commercialize its API. Marketing efforts will target user acquisition across diverse international markets including the U.S., India, Brazil, and Europe.
FinOracleAI — Market View
TwinMind’s innovative approach to continuous, privacy-focused ambient speech capture positions it well in the growing AI productivity tools sector. The seed funding and strategic backing from industry veterans provide strong validation. However, challenges include navigating privacy regulations and user adoption at scale. Market watchers should monitor TwinMind’s ability to convert its early user base into paying customers and the uptake of its API among developers.
Impact: positive
