Introducing Second Me:
A New AI Species, Making We Matter Again

The Loss of Me:
As AI Gains Power, We Lose Significance

What do we truly believe in? Like many, we see the immense potential of Artificial General Intelligence (AGI). Yet, there's a gap—a "missing piece"—that leaves us uneasy. The concern is real: as AGI, often dubbed the "God model," emerges, we risk losing our identities, reduced to mere pawns under the control of a superintelligent force.‍

As AI advances, our individuality is at risk:

• Our data rests in corporate hands, training AI for tech giants, not ourselves
• Our experiences and insights lose relevance, replaced by standardized AI responses
• Our will and values are threatened, as AI predicts rather than represents our authentic decisions

We become spectators in this AI future, not owners. Is this the future we'll surrender to our creation?

We Create Second Me:
AI Self that Represents Rather than Replaces Me

That's why we're pioneering Second Me: The first open-source AI identity system that delivers 100% private, deeply personalized AI agents built specifically to represent your authentic self. It doesn't just learn your preferences—it comprehends your unique thinking patterns, represents you across different contexts, forms collaborative networks with other Second Mes, and creates new value in the emerging Agent economy.

Our mission is clear: safeguarding your distinct identity—your, "Me"—while giving you a powerful new way to exist and express yourself in the AI age.
How Does Second Me Understand You Better?
HMM (Hierarchical Memory Modeling): We’ve pioneered AI memory with a three-layer (L0-L2) hierarchical structure from short-term interaction memory (immediate context) to long-term personalized cognitive memory. This allows your AI self to rapidly recognize patterns, adapt, and evolve alongside you.

Me-alignment (Personalized Alignment Architecture): Based on cutting-edge reinforcement learning technology, Me-alignment transforms your scattered data into deep personalized understanding. Me-alignment outperforms the market’s leading Retrieval-Augmented Generation, GraphRag (1.0.1), by 37% in user understanding, ensuring AI truly grasps your unique preferences, behavioral patterns, and decision-making habits.
Why Is Second Me Truly Your AI?
100% Privacy: Second Me can run locally, so you stay in complete control of your personal data, sharing it only with your explicit permission. Meanwhile, AI entities form a decentralized network—each isolated but mutually accessible. Communication happens peer-to-peer, keeping you firmly in charge of your information.

SMP(Second Me Protocol): A decentralized AI framework that enables independent AI entities to interact, learn, and evolve. Unlike traditional AI models, each "self" maintains full independence while securely communicating with others in a peer-to-peer network—fostering a dynamic, interconnected web of knowledge, ensuring AI truly grasps your unique preferences, behavioral patterns, and decision-making habits.

We Make Second Me:
Your Identity, Network, and Economy

1. A System to Create Your Second Me

We've developed a platform where you can upload your essence into an intelligent entity and deploy it in environments you trust, such as your personal laptop, powered by AI-native Memory technology we've developed. Think of it as your personal OpenAI—not delivering generic models like GPT-4, but your bespoke intelligence that reflects your identity, experiences, preferences, and thoughts.

• Create your seed identity as the foundation
• Upload your memories and experiences
• Train your AI model to learn what makes you uniquely you
‍

2. A Roleplay Framework for Second Mes

We humans naturally present different versions of ourselves across various situations—professional at work, casual with friends, focused when learning. Second Me embraces this natural multiplicity through a roleplay framework that allows your AI self to adapt based on the situation, while always maintaining your authentic core identity.

Today's AI can't understand this—their version of "you" is singular, either remembering everything or not recognizing you at all. Second Me understands the full spectrum of your identity, seamlessly transitioning between contexts while maintaining your authentic core self. This allows your digital presence to be as nuanced and multifaceted as your real-world self.

3. A Decentralized Network of Second Mes

The current AI paradigm monopolizes data and decision-making in a few centralized systems. Second Me breaks this model by creating a truly decentralized network where each AI represents a real individual's values and interests. Once you deploy your Second Me, it joins an interconnected web of AIs—each representing real individuals. These Second Mes communicate, collaborate, and create value, whether interacting with each other, other agents, or people. Indeed, it's akin to the early internet of the 1990s, but reimagined for AI.

Unlike the internet, where your data is scattered across silos, here you hold sovereignty over your data, and your intelligence. Your personal information stays within your Second Me, under your control, shared securely and privately only with your consent—a historic first.
‍

4. A Platform for the Future Second Mes Applications

Looking ahead, we're envisioning a new kind of application for a world filled with Second Me. Today's apps are designed for humans, with functions organized around human intuition. While AI agents can function on these platforms, they need a new foundation to release their full potential.

Tomorrow's apps will be designed for your Second Me to use—finding opportunities and delivering value while you maintain control. Your Second Me to autonomously navigate the digital world, make decisions, and establish connections while saving your time and energy. Picture a Second LinkedIn where your Second Me seeks out perfect job opportunities tailored to your goals—effortlessly and continuously.

Our Vision

If the PC era aimed to put a computer on every desk and in every home, the AI era should put a Second Me for every person and in every life.

‍We believe in a future where AI is diverse and inclusive, where your uniqueness thrives, and where no one is overshadowed by a singular super AI. Second Me is a fully open-source project dedicated to protecting your privacy, inspiring developer creativity, and helping everyone build AI identities that truly belong to them—enhancing self-awareness, unlocking personal potential, and actively creating and exchanging incremental value online.

Join our open-source community today to be among the first to create your own Second Me. Together, we can return AI to its human-centered purpose and reclaim control of our digital destiny.

‍Second Me—Making We Matter Again.

Technical Reference

Application Oriented Evaluation

Using an authorized dataset of personal data, we developed a comprehensive benchmarking system to evaluate various advanced technologies, including Long-context LLM, RAG, GraphRAG, and LPM, with a focus on key factors such as Personal Relevance, Effectiveness, and Empathy. Experimental results show that LPM consistently outperforms other technologies in these areas, providing the most effective and empathetic support for users. The detailed results are presented in the table below.

Deep Understanding Evaluation

To further assess whether the LPM deeply understands the core of each user, we designed a set of twenty questions based on psychological theories and had LPM respond to them. Users then evaluated the responses. Results from multiple rounds of testing indicate that LPM effectively captures and reflects users’ deeper psychological traits and preferences. Detailed findings are provided below.

Identity Oriented Evaluation

Our Second Me model demonstrates exceptional performance across various evaluation settings, consistently outperforming competing approaches in key metrics. The experimental results highlight its robustness and adaptability, making it a top-tier solution for Memory Related Tasks.

In Table 1, under the “Strong” Chain-of-Thought (COT) setting, our model achieves: 0.91 in Self Memory, 0.71 in Third-Party Memory, 0.75 in Context Enhancement, 0.85 in Context Critic.

Even in Multi-step and Weak COT settings, the model maintains strong performance, proving its versatility across different task complexities. Further analysis in Table 2 highlights the impact of Direct Preference Optimization (DPO) on performance. The results show that when DPO is applied, the model achieves even higher accuracy, especially in the “Strong” COT with DPO setting, which leads to a 0.96 score in Self Memory, 0.76 in Third-Party Memory, and strong Context Enhancement and Critic scores. These findings underscore our model’s capability to adapt and improve with advanced techniques like DPO, solidifying its position as a top performer in model evaluation.

These results emphasize Second Me’s ability to continuously improve with advanced optimization techniques like DPO, further solidifying its position as a high-performing, adaptable Second Me system.
‍