EX-Google X Trio wants to be artificial intelligence is your second brain-and they only raised $ 6 million to achieve this

Three former Google X scientists aim to give you almost a second brain-not in scientific treatment or in your sense-but through an artificial intelligence-backed application that acquires context by listening to everything you say in the background. Startup, twinIt raised 5.7 million dollars of seed financing and issued Android version, along with a new letter for Amnesty International. It also has an iPhone version.

Participated in its founding in March 2024 by Daniel George (CEO) and former Google X Sunny Tang and MAHI KARIM (both CTOS), whose Twinmind is running in the background, where they pick up the surrounding speech (with the user’s permission) to build a personal identifier graph.

By converting ideas, meetings, lectures and spoken conversations into organized memory, the application can create notes operating with the same, task and answers. It works in a non-connection mode, treats the sound in the actual time to copy the device, and can capture the sound continuously for 16-17 hours without depleting the device’s battery, the founders say. The application can also backup user data so that conversations can be recovered if the device is lost, although users can cancel the subscription to it. It also supports translation in actual time in more than 100 languages.

TWINMIND is distinguished from facing Ai Meeting Notes such as OTTER, Granola and Fireflies by picked up the sound negatively in the background throughout the day. To make this possible, the team built a low -level service in Swift Pure originally working on iPhone. On the contrary, many of the original RACT competitors are used and depend on the treatment -based processing process, which Apple restricts to run the background for long periods.

“We spent about six to seven months of last year only to master this sound and get there to find a lot of breakthroughs about the walled Apple park,” he told Techcrunch.

George X left in 2020 and got an idea for Twinmin in 2023 when he worked in JPMorgan as Vice President and Amnesty International, with successive meetings every day. To save time, he built a text program that got the sound, copied it on his iPad, and fed it in ChatGPT – which began to understand his projects and even create a usable code. He was impressed by the results, share it with friends and posted it on Blind, where others showed interest but they did not want to turn on anything on their laptops. This prompted him to create an application that can work on a personal phone, and to listen calmly during meetings to collect a useful context.

In addition to Mobile applicationTWINMIND a Extension of chrome This combines an additional context through the browser activity. Using Vision AI, it can visually wipe the open tab and interpret the content from different platforms, including e -mail, stagnation and idea.

Even the startup company used the same extension for the brief list of more than 850 applications it received this summer.

“We have opened all LinkedIn and CVS profiles of 854 of the browser tabs, then they asked Chrome to extend the best candidates.” “She did a great job – so we rented the four final trainees.”

He noted that the current AI Chatbots-including Chatgpt from Openai and Clauds’s Claude-can process hundreds of documents or recordings from tools such as LinkedIn or Gmail to collect contextual information. Likewise, browsers on behalf of artificial intelligence lack such a result of confusion and the browser company to the ability to build knowledge from your internet conversations and personal meetings.

Upon starting currently more than 30,000 users, with about 15,000 of them are active every month. George said that up to 20-30 % of TwinMind users also use chrome extension.

While the United States is the largest base for Twinmind so far, starting start also sees a jar from India, Brazil, the Philippines, Ethiopia, Kenya and Europe.

Twinmins target the general public, although 50-60 % of its users are currently professional, about 25 % of students, and staying 20-25 % are the individuals who use it for personal purposes.

George told TECHRUNCH that his father is among the individuals who use TwinMind to write their CV.

One of the important defects of artificial intelligence is the ability to waive the user privacy. But George confirmed that Twinmin does not train his models on user data and is designed to work without sending records to the cloud. Unlike many other notes blogging applications, Twinmind does not allow users to access audio recordings later-the sound is deleted during flying-while the text that was copied only locally in the application is stored, he indicated.

Google X has helped accelerate things

The participating founders have spent a few years working on many projects in Google X. Io The team is behind the ear powers of intelligence, which Newspaper headlines have been released recently to prosecute Openai and Jony IVE. This experience helped TwinMind to move quickly from concept to product.

“Google X was really the perfect place to prepare for your private company starting,” George said. “There are about 30 to 40 projects that are similar to startups that happen at any time. No one else works in six emerging companies at an early stage over two or three years before its launch-at least not in such a short period.”

Before joining Google, George worked on the application of deep learning on astronomical physics for the gravitational wave as part of the Nobel Prize -winning LIGO group at the National Center for the University of Illinois for supercover applications. He had completed his doctorate in artificial intelligence of astronomical physics in only one year – at the age of 24 – an achievement that led to joining the research laboratory Stephen Wolfram in 2017 as a researcher in deep learning and AI.

This early association with Wolfram came after years of the department-ended up writing the first check for TwinMind, which represents his first investment ever in an emerging company. The last seed tour led simplified projects, with the participation of Sikoya Capital and other investors, including Wolm. TWINMIND Round values ​​at $ 60 million after money.

TwinMind Ear-3 Model

In addition to its applications and expansion of the browser, Twins also provided TwinMind Ear-3, a successor to the current Ear-2, which supports more than 140 languages ​​around the world and has a 5.26 % error rate, Startup said. The new model can also recognize different speakers in a conversation and has an error rate in a 3.8 % march note.

The new artificial intelligence model is a delicate mixture of many open source models, trained on a group of human data made of human data-including podcasts, videos and movies.

“We have found that the more languages ​​you support, the better the model in understanding regional dialects and dialects because it is training on a wide range of speakers,” George said.

The form costs $ 0.23/ h and It will be available through the application programming interface For developers and institutions during the next few weeks.

Ear-3, unlike EAR-2, does not support a complete experience of non-connection, because it is larger in size and operates it on the cloud. However, the app automatically turns to Ear-2 if the Internet goes and then moves to Ear-3 when he returns, George said.

With Ear-3 insert, TwinMind now provides a supportive subscription at $ 15 a month, with a larger context window up to 2 million icons and email support within 24 hours. However, the free version is still present with all the current features, including unlimited hours of copying and identifying the speech on the device.

Upon starting, a team of 11 members is currently starting. It plans to employ a few designers to enhance its user experience and prepare the business development team to sell its application programming interface. Moreover, there are plans to spend some money on obtaining new users.