Based on what you know about me, take a goddamn guess about my MBTI type?”

Accidentally, I saw Jeffery Wang, the founder of Exa, share this prompt on X.

I tried it immediately, and the result was strangely consistent with a paid test I took years ago. This non-human “conversation partner,” relying only on our past chat history and its growing “memory,” seemed to genuinely capture and “understand” some ineffable core trait of my behavior.

That experience isn’t unique.

More and more people are using similar prompts to have deeper interactions with AI, asking it to “write me an anonymous letter,” to reflect on their inner selves; or to “paint an abstract picture representing my current mood,” exploring the subconscious.

AI is evolving from a tool for information retrieval and content creation into a mirror that reflects the self and sparks inspiration. It’s starting to touch on more personalized, intimate user moments.

Catching the Fleeting Moments

The first battle in the AI application race is about how to capture those fleeting “moments” in our lives.

By “moments,” I mean those split-second inner impulses, a vague need, a sudden idea, a piece of information to confirm, an emotion to express, hesitation before a decision. These are high-value for they directly link to real users and potential actions.

Smartphones and app ecosystems have vastly extended our digital lives, solving the “always connected” problem. But when it comes to capturing and responding to those fleeting thoughts, the mobile internet paradigm still suffers from friction.

We switch between apps, trying to “translate” fuzzy ideas into machine-understandable keywords or commands, endure information overload, or face social friction when asking others. Because of this friction, countless potential needs, inspirations, and connections evaporate before being fulfilled. These are the “vanishing thoughts” of the digital world — uncaptured value.

Now, AI is becoming the catcher of these vanishing thoughts. Its core advantage is:

• Instant & Low Friction: Natural language interaction lowers the barrier to expression dramatically. You can speak like talking to a human, using vague, conversational language, and AI responds almost immediately. The cost of capturing a thought drops to near zero.
• Contextual Understanding & Personalization: AI grasps complex contexts, leveraging historical data (like ChatGPT’s Memory) to give more personalized, accurate feedback. It’s not just responding — it’s “understanding” you.
• Potential as a Universal Interface: AI assistants are evolving toward a “universal interface,” integrating search, creation, socializing, shopping, and services. No app switching needed — one chat window can handle many thoughts.

This ability to catch vanishing thoughts is no longer just about reactive. Further, a clear signal is emerging: users want AI to go further to proactively initiate conversations, grasping the key moments we might definitely overlook or “be too lazy” to take over ourselves.

In a recent Reddit AMA with OpenAI’s behavior lead, when asked if ChatGPT could proactively start conversations, the answer was “absolutely possible.” Users responded with vivid scenarios: empathetic, thoughtful nudges to keep calendars, track goals, offer emotional support, break silences — becoming a warm, intelligent companion in everyday life.

These real user voices crystallize the next-gen AI interaction paradigm. They confirm the core logic of “Moment Share”: whoever can more sensitively, proactively, and naturally capture and serve those fleeting user needs, intentions, emotions, and thoughts will win.

Social, once more

A clear strategic direction is turning these AI-empowered moments into network-effect “social moments.” Pure model-ability competition seems to be hitting a plateau. Leaders find it hard to pull away from followers. The real moat may be in the application layer — especially social contexts that lock in user relationships, spark content flow, and build unique community vibes.

To understand such shift, we can’t ignore the AI’s capabilities. We must see how they work together to build social graphs and personalized feedback loops:

• Persistent Context: It’s no longer just about AI “remembering” your past to be a better assistant. Socially, Memory means that AI forms a stable, trustworthy “digital persona” or “AI avatar.” This avatar not only knows you better but can play a proactive role in social interactions, reminding you to follow up with friends or, with your permission, interacting with others’ AI avatars for interest matching. It makes user-to-AI and even user-to-user (via AI) bonds stickier, preserving value over time.
• AI Decision / Recommendation: This isn’t just about purchase decisions, but a matter of “starting social connections.” When AI recommends a book, podcast, movie, or event based on deep understanding, it’s not just guiding consumption, but it’s creating social opportunities. Imagine that AI notices you and your friends sharing an interest, then suggesting a meetup; or recommending community-oriented products that help you join like-minded groups. Recommendations become social “seeds.”
• Viral Prompts / AI-generated Images / Videos / Music: This is more than lowering barriers to personal expression. In social contexts, AI is a “minting machine” for new social currencies. AI-generated images, texts, and music aren’t just self-expression — they initiate interaction, convey emotion, and enable collective creativity. These AI-assisted creations, with their novelty and low effort, spread virally (memes), becoming group “lingo” or “totems,” accelerating community identity and culture evolution.
• AI Search: No longer just about capturing “information moments.” When search gets smarter and conversational, delivering ready-made solutions, results themselves gain social shareability. You’re not sharing cold links but AI-curated travel plans, concise answers, or project outlines — all more likely to spark discussion and collaboration.
• AI Social Exploration / Emotional Companionship: This is the most direct social experiment — building and maintaining relationships themselves. Whether simulating lovers, friends, or specific roles, AI explores fulfilling deep human emotional needs. The key is feedback loops: user interactions “feed” AI to better understand and personalize; this “understanding” increases user attachment, encouraging deeper interaction and sharing, even forming communities around particular AI personas or emotional experiences (like Character.ai’s user base).

To sum up , these capabilities together sketch a clearer picture: AI is no longer just a tool to boost individual efficiency. It’s deeply embedding itself in our thinking, decision-making, expression, and emotional connection — turning discrete “individual moments” into shareable, interactive, and enduring “social capital.”

From Mind Reading to Web Weaving

Yes, the billion-user AI race is fully underway, driven by the ferocity of capital and resources. It echoes past tech waves — some even smell a bubble. When the hype fades, what will be left?

The history of the internet bubble offers a clue. The oft-quoted lament “We wanted flying cars, but got 140 characters” hints at a gap between expectation and reality. But we shouldn’t underestimate what those “140 characters” built: resilient, valuable legacies that reshaped information flow and human connection. They gave massive network infrastructure a sustainable, scalable use, unlocking potential for billions.

Looking at today’s AI trajectory, especially OpenAI and Meta’s fierce social battles, the stakes are higher than “user time”. This is about a critical “landing” exploration:

Will AI, with its astronomical investment, find broad, deep, everyday use , especially woven into our most fundamental, highest-frequency social interactions?

Because a technology’s ultimate value isn’t just about moonshot breakthroughs. It’s about becoming infrastructure that millions rely on to connect, communicate, and live.

If AI ends up serving only elites or niche industries, and can’t penetrate social life like the internet’s social apps did, then how will today’s massive compute and model investments be justified?

New Frontiers? Let’s see

The race for capturing fleeting thoughts is about tools — but winning minds is about social.

The focus is shifting from isolated moments to building relationship graphs and personalized feedback loops. This is the core narrative for AI consumer applications: not tools, hardware, or chat — but superreal human projections.

In this battle, OpenAI and Meta represent two very different paths.

Sam Altman isn’t new to social — his early Loopt failed, revealing the limits of location data alone for lasting connections. Now, OpenAI seems to have learned, betting on ChatGPT’s deeper “understanding” of individuals — leaning on Memory, creative generation, and empathetic simulation to build solid “individual-to-AI” bonds, hoping to birth new social paradigms from there.

Many criticize OpenAI’s model improvements as underwhelming. But benchmark scores are misleading — they don’t measure real problem-solving or emotional intelligence, just test-taking skill. This illusion confuses both competitors and model companies themselves.

Yet recent “sycophancy” issues reveal the pressure OpenAI faces at scale: trying to “understand” and please millions exposes the limits of algorithmic emotional intelligence. This complexity may be OpenAI’s unique growing pain and highlights how hard it is to build “real” AI social from scratch.

Altman’s investment in WorldCoin hints at a deeper reflection on identity and human connection in the AGI era.

Meta, by contrast, is injecting AI into its vast existing social graph. Zuckerberg says the true north is user feedback and real value, not benchmark scores — the advantage of billions of users experimenting in the wild.

Meta integrates AI into Ray-Ban smart glasses, aiming to make AI a “plugin widget” for real-world socializing. Their CPO, Chris Cox, notes Meta AI users “play” with it spontaneously — inventing new interaction styles and memes, like trading AI-generated goofy images, quickly forming trends.

Meta’s strategy leverages unmatched network effects to let AI features organically “grow” within real social interactions, rather than force-feed them.

So we see two strategic choices: OpenAI goes inside-out, reshaping connection through deep individual understanding; Meta goes outside-in, grafting AI onto existing social graphs. The winner depends not just on tech but on whose path better captures and converts those fleeting human thoughts and social cravings.

The future of AI social is still unwritten. But the promise is clear: AI could transform the atomized moments of our minds into a rich, communal social web — if it can overcome the friction and complexity to become truly meaningful in everyday human connection.

If you want to talk MBTI, AI, or social tech, I’m all ears.