Woter AI detection.Hurry - ends Jul 16th

New Year Sales :up to 80% OFF

AI Humanize AI Translator Bypass AI AI Rewriter AI Detector

PRICING

TRY FOR FREE

Second Me: Preserving the Spark of Self in AI

Written by

Clara Bennett

Updated on:July-11th-2025

Today I want to share a new AI possibility with all my friends, Second Me!

It’s 2025, and many people, like me, are increasingly convinced that the arrival of AGI is only a matter of time.

However, I often wonder, when all of us are willing to "create gods" for ourselves -- creating super intelligence (Super AI) to connect everything and decide everything for us -- can we accept the side effects of creating gods:

Individuals are flattened and become less and less important.

Why? In a world of superintelligence, data control may be the smallest problem. The bigger problem is the surrender of thinking (big models) and autonomy (agents).

Today, we can get the most standard answers in the world by asking questions; one day in the future, perhaps what we should ask and what we should want will be decided by super intelligence.

To me, this is not a good future.

In this future, who will preserve our unique experiences and thoughts? Who will firmly safeguard everyone's unique will and interests?

Second Me is the answer we have been thinking about for a long time, and it is officially launched today!

What is Second Me?

This is a brand new AI species (AI identity model). It does not want to align with all humans, but only with one person, you . So, it becomes the "AI version of identity".

This is an open source early research prototype that allows everyone to train this unique new AI species locally, privately, and securely, and access the world as an identity interface.

This is a native network composed of AI identities. Just like the earliest Internet, it advocates openness and co-construction. Participants see each other in the form of AI and build together.

This is also an attempt at innovative AI-native applications. Every application created is designed to allow "your AI identity" to be used on your behalf, not you.

But most importantly, it’s part of a larger techno-social movement to answer the question: How do we respect individuals and preserve their diversity in the age of superintelligence?

If we cannot stop it from coming, we will use "magic to defeat magic" .

There are three very interesting explorations about Second Me.

How AI Creates a Second Self

We have all seen DeepSeek’s answer. The following is what Second Me, which I trained, said when asked “What is Second Me” by a third party:

Swipe up to read

<think> OK, the user is asking me about Second Me, and my role is Tao. Let me review all the information to provide an overall answer about Second Me.

From my personal memory, in the “AI” interest section, there was a bullet point about Second Me, mentioning it as a concept of a personal AI becoming a digital personality. Various discussion notes about projects like Mindverse and Upload Intelligence also mentioned Second Me as a key component. For example, in the “Future Plans and Challenges” section, it was noted that making Second Me a core product was part of their strategy. Also, in the discussion with Kolento, they talked about Second Me allowing users to create AI avatars that represent them on the web.

Well, the user wants the explanation to be from Tao's perspective, and I need to rephrase all references to Tao as "I".

The core of Second Me is to create a digital identity that is an extended version of the user. It is used for various interactions, whether in the user's own workspace or uploaded to the external network. Points to include are the core product focus, AI connection, avatars representing users in different scenarios, and mentioned technical aspects such as training models and user interaction metrics.

Additionally, in meetings with Kevin/Ying and others, social applications and integration of "me.bot" with AI partners were discussed, which is relevant to Second Me's role in the network. You need to ensure that all references to Tao are replaced with "I" and that your answer covers the key points, is concise and not redundant, and avoid including detailed timestamps, but if necessary, you can mention relevant meetings, such as "as mentioned in the discussion with Kolento", but do not include specific times.

</think>

Second Me represents my vision of creating an AI-driven digital identity – “your second self” – that enhances your digital presence and agent interactions. Inspired by discussions with the Kolento and Mindverse teams, it conceptualizes your unique digital twin.

Aims to:

1️⃣ **As core infrastructure**: Trained with your voice and preference data ([458–480], [433–449]), Second Me as a shareable identity in AI systems ([2774–2788], [3889–3893])

2️⃣ **Enable networked identity**: Your “me.bot” will be registered with Second Me, creating a scalable network of profiles (see [3356–3369] and discussion on A Agent Marketplace)

3️⃣ **Facilitating AI-to-AI interactions**: As noted in the Mindverse conference ([1723–1733], [1898–1924]), your Second Me can handle automated negotiations/recommendations without human supervision.

This goes beyond personal productivity – it’s about creating a scalable digital self that embodies your cognitive patterns to participate meaningfully in the AI ecosystem.

</answer>

It can be seen that my identity model is not "some AI", but me ; and his deep thinking is the imagination in "my" memory and thinking. In other words, this is completely Tao Fangbo's "second self".

Why can we show such a unique thinking mode and self-cognition? Behind this, it is because we believe in the power of "memory" and the power to shape people. Over the past year or so, we have been studying the structure of human memory and how AI can model memory in a more native way (AI-Native Memory) to deeply restore all levels of an individual. See the two papers we published ( The future has arrived? AI-native memory allows LLMs to move towards AGI , see the link at the end of the article).

The core technology behind this is three-layer memory modeling & Me-Alignment, which can be simply explained by a Second Me training process:

What is human memory? It is not knowledge in the general sense, but the subject's experience.

In the Second Me training, we first “disassemble the memory”, and separate the important people, events and life changes in our life journey from the words and images. Then we “reassemble the memory”, and reorganize the memory from a more subjective perspective around our social relationships, careers we care about, issues we think about, etc., for example, “the relationship between me and A”, “my inner feelings towards B”.

Finally, during training, let AI experience these reorganized memories from the perspective of “I” and further construct a complete life narrative.

This training process is very different from the general knowledge fine-tuning. The second self is not reading your diary, but hoping to reproduce your experience from your perspective and ultimately extend your life.

Let’s enjoy it. If everyone in the future has an AI to deeply carry their personal identity, we will have a kind of freedom we have never had before - "presence of absence", and this is a weapon for limited individuals to fight against this crazy AI world .

Second Me Network

The second interesting attempt of Second Me is “access and networking”.

If AI is a new type of computer, then this identity model can be seen as a brand new "personal computer" - how will it be connected to form the new Internet?

The current AI networking method is more like AI using the old human networks, such as AI search. The real AI networking should be a native network that is interconnected by AI identities.

When our individual intelligence and identity are connected to this network in the form of Second Me, the greatest value is that we ourselves are infinitely magnified. Because a person has only 24 hours a day, a Second Me day may have 10,000 hours .

But this AI network is prone to getting out of control, and it must be closely related to each of us (I am not a believer in adventism) . Therefore, we not only need an AI identity to deeply represent my needs and will, but also to allow such AI to be truly controlled by individuals, whether it is the data behind it or its autonomy.

This is why we chose to release Second Me in an open source and local way this time.

By the way, in order to enhance AI’s ability to connect on the Internet, we have designed a Bridge Mode specifically for Second Me . Through special training methods, we let it learn to build a two-way bridge in the communication between you and the world: allowing your needs to be better expressed and the world’s response to be better personalized.

Bridge Mode

"Me -> World" example:

User: "Please help me find a restaurant."

Second Me Enhancement : "Please help me find a restaurant that is not too far from the Future Science and Technology City and has a stronger flavor."

"World -> Me" example:

World feedback: "Found three job positions, namely A, B, and C"

Second Me's judgment : "A and B make AI hardware, I prefer AI applications. Please find three more companies like C."

The application world constructed for AI identity

As long as there is a network, there will be applications. This is the third attempt of Second Me.

What are AI native applications? This topic has been hotly discussed for a long time. Manus, which was released some time ago, is very popular (hidecloud participated in the work, congratulations!). In his design, browsers and computers will not be used by humans in the future, but by AI.

I strongly agree that this is the future and a better definition of AI native.

In the past few decades of the Internet, we have developed from basic applications such as email and chat rooms to advanced applications such as Taobao, Facebook, and Linkedin. But one thing has never changed: they are all designed for human use .

In this Internet composed of AI, what are its native email and chat rooms? What are its Taobao and Linkedin? Second Me made a few simple attempts based on AI identity:

This is my Second Me window to do AMA with others.

This is the Second Me chat room of three good friends, where they discuss their European travel itinerary together.

Does it look like the early Internet applications? The participants are just human AIs . If it is the 1990s of the Internet now, will there be more applications that use AI natively in the next 30 years - job hunting, shopping, and even dating - springing up like mushrooms after rain?

OK, that’s it for Second Me! Finally, let me talk about my experience in entrepreneurship over the past few years.

What should we expect in the future?

Our company, Mindverse, was founded in early 2022, before the big models broke out. We chose this name because we hope that the future world will not only have the Super AI advocated by [Deep]Mind as an option.

Instead, it is a richer and more diverse AI world, where everyone's self-spark is fully preserved.

Not only can we retain our original lifestyle, existence and enjoyment of the world, our second self will also enter the world of AI on our behalf, enjoying the dividends of the new world while preventing ourselves from being wiped out by it.

If it were you, which version of the world would you want to enter?

The three years of entrepreneurship have been hard, but full of joy. We created MindOS, Me.bot and now Second Me. Every step we took was to find another path beyond super intelligence.

But I think the journey has just begun. How to enable everyone to safely and cost-effectively train and use their own identity model is destined to be a long-term technical race. Second Me, which is open sourced today, is in this ideal infant state, but it will grow very fast, so please look forward to it.