Let AI Chat for a Bit | Second Me v1.1 Full Test & Review
When AI becomes the "hype squad" for our social relationships.
**
When AI becomes the "hype crew" in our social relationships.
👦🏻 Author: Jingshan
🥷 Editor: Koji
🧑🎨 Layout: NCon
Recently, ChatGPT released its own 2025 annual report.
Many friends clicked through immediately, and the shared reaction afterward was nearly unanimous: over the past year, we've probably talked to AI more than we've talked to a lot of real people.
This is actually pretty interesting in itself.
So a question naturally emerges:
Once AI becomes a "conversation partner," will it — and should it — enter into human-to-human social relationships? How would it enter? And how should AI social products be built?
Three months ago, we spent time with and wrote about an AI social product called Second Me — "This 13K-Star GitHub Sensation: The Next AI Social Species? | Me.bot New Version Tested."
At the time, what struck many people was how different its angle was. Second Me tries to build a second "you" — an "AI persona" — and drop it into your existing friend relationships to hang out, riff on random ideas, and create things together.
A week ago, Second Me updated to version 1.1.
The interactions were restructured, the play modes multiplied. We took it for another full spin and found that it's walking the "AI social" path with more conviction and more specificity.
From the perspective of real-world use, let's talk about: what changes when AI starts participating in "us" and our relationships.
When AI Becomes the "Hype Crew" in Our Social Relationships
First off, Second Me is now playable as an app — iOS and Android both work, based on our testing. Just search in the App Store.
The basic flow in Second Me: register your persona, fill in your profile, then populate your memories. This part hasn't changed much from the last version; you can refer to "This 13K-Star GitHub Sensation: The Next AI Social Species? | Me.bot New Version Tested" for details.
On top of this, the entire "AI + social" flow in Second Me has been significantly enriched.
The Second Me app home screen now looks like this, with 3 core tabs: Me, Friends, and Discover — basically the standard social product framework:
Now, you can share your personal QR code directly in the Second Me app.
At the same time, Second Me assigns every user a dedicated domain. You can customize it yourself or let the system auto-generate one — overall, the flexibility is pretty high.
Around these QR codes, the distribution methods are fairly rich too. Beyond the familiar WeChat Moments, there's a pretty interesting design: Second Me stickers.
Once paired, you can "bump" your NFC sticker against someone else's to add them as a friend:
Now, a very core social feature in Second Me is called "Discover."
Simply put, this is for meeting strangers. It automatically searches for people you might find interesting based on the memories you've given Second Me, combined with keywords you input.
Of course, you can also be more proactive — write out exactly what you're looking for, then use "Discover" to find new friends, dates, or activity partners to do things with.
For example, you can search by industry: AI-related professionals across fields, with further refinements possible, even custom gender filters for search results:
From the feel of it, the user base in Second Me is already quite substantial. The whole community gives off strong "real person" vibes.
You see people from all industries and professions. Often, the other person will initiate a greeting; you can reach out first too — the whole process feels natural.
For instance, after coming across someone in "Discover," you can simply say hi.
What happens next isn't immediately jumping into human-to-human chat. Instead, both sides' AIs step in first — each person's Second Me, carrying their memories, starts a conversation on your behalf. In other words, Second Me's logic is: let AI break the ice first.
The conversation between two "Second Mes" mainly serves to help you judge: whether you two actually click.
At this stage, you don't need to dive into socializing cold — you can "observe" from the sidelines, see how far the two AIs take the conversation, whether your interests align, whether your vibes and thinking are on the same wavelength.
For example, I found someone very interested in AI. After I greeted them, instead of direct text chat, their Second Me stepped in first.
This Second Me showcased a voice clone based on their voice, generating a ~30-second greeting.
What makes this experience special is that at the very start of an interaction, voice is a modality that easily stirs emotion. When you send a greeting and first hear back a highly realistic voice response, that moment genuinely feels different.
Once the two Second Mes finish a round of conversation, if you feel like you click, you can choose to add as friends or let the AIs continue a deeper chat.
Note that this chat process doesn't exist "only in AI."
All conversation content syncs to your own account and to the other person's account. That is, these records are public, transparent, and traceable for both sides.
In actual use, I also found a pretty key point: when you sustain a conversation with someone, the process of building and maintaining a social relationship naturally exposes a lot of information. And this information gets used to supplement your own Second Me's key memories.
Moreover, this "key memory" isn't just verbatim storage of what you said. Often, Second Me distills and summarizes the content, keeping only the truly representative parts.
And through this continuous accumulation, these key memories keep enriching your Second Me, making it increasingly close to "you yourself."
So the question becomes: what's the point of continuously enriching these memories for your Second Me?
My observation is that in Second Me's social system, when chatting with strangers, the system actually displays an "information richness" concept. For instance, my own richness might be 40-something percent.
In the logic of AI social, this number matters. The system assumes by default: the richer your information, the more your Second Me can represent your personality and psychological depth, which in turn facilitates stranger socializing and judgment.
In other words, higher richness means your AI is more "like you," making it easier for others to quickly understand what kind of person you are through the AI, and thus more willing to form a social connection with you.
And in this Second Me 1.1 update, the most core and most fun additions are concentrated in 3 features: AI Match, AI Brainwave, and AI Duet.
All three are designed around AI-powered social interactions between you and friends, or you and strangers. You can @ a stranger anytime, or play these AI social modes with existing friends.
There are two entry points for using them.
[1] The first is on your home screen;
[2] The second is in the chat interface when messaging with a friend.
The overall design thinking is clear: whether you've just met someone or have been chatting for a while, you can naturally bring AI social play into the flow.
The three features are all quite richly designed.
Let's start with AI Match.
Essentially, it takes two people's profile photos and generates a set of AI images in different styles. Both sides can save the generated images. The process itself is pretty fun — kind of like "a visual memento of your relationship," which feels quite magical in experience.
Next is AI Brainwave.
AI Brainwave lets two people, each with their memory-carrying Second Me, engage in a more video-oriented dialogue or debate "in character."
You can choose different themes and styles, and watch how two AIs discuss the same topic.
Finally, AI Duet.
This one's more straightforward: two people use their respective cloned voices to sing a full song together.
For example, the image below is one I generated with a friend.
His key memories lean more toward: daily Douyin scrolling, following entertainment-oriented news content, while my self-designation is an author very interested in AI.
You can clearly sense that when the system generates images, it pulls both people's avatars while also trying to leverage each person's attached memory information to freely create visuals in different styles.
Overall, it tries to stay close to each person's style and vibe, freely creating based on your memories.
AI Duet is actually really fun.
It uses two people's cloned voices to complete a song together, with tons of style options. You have full control: what style of song to sing, or roughly what the song should be about, you can set directly.
For instance, I picked a ballad-style song below, and overall it came out quite relaxed and atmospheric:
Give it a listen~
Link below, welcome everyone to check it out:
https://second-me.cn/talk/hZI9UmskvUfebncP
The interaction itself is genuinely interesting, and from my actual use, the voice cloning quality is quite natural. The two voices blend into a complete song that doesn't feel jarring.
Of course, these AI-generated works can be distributed afterward. You can share them out or drop them directly into your "Highlights" module.
In my understanding, this "Highlights" is kind of like pinned posts on Moments — it centrally showcases your key info and representative content.
During immersive use of Second Me's AI social features, I also found quite a few hidden Easter eggs.
For example, sometimes AI will proactively "intervene" in your social experience at a natural moment. When chatting with a friend, you might occasionally spot a small "Second Me" icon prompt popping up at the bottom of the chat page.
This prompt actually encourages you to play an AI auto-generated inspiration-based interaction. Click in and you can directly initiate theme content generation like AI Match, naturally extending the conversation.
Or, on your home screen, it sometimes gives you a fun alert, like: "Report: wild CEO spotted."
Click in and you can directly launch the three AI interaction modes mentioned above with that person:
At this point, I decided to have an AI Brainwave with Second Me founder Dr. TAO — a late-night emotional radio talk.
Link below:
https://second-me.cn/talk/hVVPzNgYroTAwUcP
Since Second Me's AI social system is built on its fairly solid memory capabilities, this has also sparked some concern: if I input all my key memories into Second Me, when others chat with my AI, could it "let something slip"?
But this time, I found that Second Me has made a lot of optimizations around privacy and boundary settings. It offers a feature letting you set topic taboos, to prevent the AI from crossing red lines in judgment or "leaking" sensitive info:
Of course, Second Me isn't just an AI social toy for "messing around."
Its community currently gathers all kinds of people. And a crucial reason is: every AI here is built from users' key memories.
In this environment, you can naturally chat with a "stranger."
Another important point: the conversation starts with the other person's AI responding. This alone dramatically lowers the psychological cost of initiating contact. You don't worry about bothering anyone, or whether your question sounds "smart enough" — instead, you more easily get access to diverse perspectives and inspiration.
A concrete example.
I previously connected with a product manager building an AI relationship analysis product. When I came across her profile, I noticed her personality richness and memory density were both high, so I directly asked her:
For building an AI relationship analysis product, is there a relatively systematic methodology? How should this product be broken down? Are there different approaches in the analysis path?
At this point, what actually answered me was her Second Me AI. And these responses were completely generated based on her past experience and memories, so the information density was very high and quite specific.
This usage isn't limited to any single person or industry.
For example, I also asked Dr. TAO:
How should an AI social product actually be built? Where does your inspiration come from?
And asked some AI media blogger friends:
What's your topic framework? Can you walk through your thinking path specifically?
The overall feeling is that these responses genuinely contain personal experience.
If you want to go deeper and get more systematic understanding, it's actually quite simple. You can directly add the person as a friend, then continue through "AI Brainwave" or private chat with the real person for more in-depth dialogue.
Precisely because of this, from a practical utility standpoint, Second Me already delivers clear value during the "ice-breaking" phase between users, dismantling many barriers.
As mentioned earlier, you can even directly set your gender and preferences, letting Second Me help you search for dates.
For example, you could write a very specific prompt:
"I want to find a girl who also loves fitness and enjoys literature."
In actual use, you'll find the precision of the final matches in terms of interests and orientation is genuinely quite high. To be honest, I've already matched with many "candidates" well worth continuing deeper conversations with.
Now, I've already made many real friends in Second Me, and even scheduled Coffee Chats.
Looking back at these past few days in Second Me V1.1.
In an era where everyone's "socially anxious," where even making a phone call requires mental prep, we sometimes desperately need a reason to start a conversation.
We often worry that AI will make human relationships atomized and cold.
But in this Second Me update, many see another solution — still immature, but intriguing:
AI isn't trying to replace us in socializing. It's more like a tactful "ice-breaker," or a "hype crew."
One of the meaningful aspects of AI social products like Second Me is letting you more easily reach out in real life, tap a friend on the shoulder and say: "Hey, check this out, wanna try it together?"
If you're curious what "chemical reaction" might happen when AI enters your chats with friends and strangers, go experience Second Me.
