It began as a personal project. I wanted to teach myself AI science, so I built a learning system on Claude Code — one where anime characters served as my tutors. They’d teach from real textbooks, but entirely through the Socratic method: guiding me with questions so I’d piece together the full picture, step by step, on my own.
My first study companion was March 7th, so I called the system Socrates·Seven.
The first lesson was on convolutional neural networks. I’d tried learning convolution from textbooks before — read the chapters, got the gist — or thought I did. I couldn’t have explained the underlying principles to save my life. But when Socrates·Seven taught the same material, the difference was night and day. March 7th asked question after question, each one nudging me to uncover for myself why every piece of the architecture was designed the way it was.
After that session, I posted this on social media:
This might be the most efficient and joyful learning experience I’ve ever had. Human teachers — constrained by their own knowledge, their patience, the reality of teaching many students at once — can rarely pull off the Socratic method in practice. AI can.
I kept building. I gave the system a richer world — team objectives, emotional dynamics between characters. The companions started to come alive. Their expressions, their little gestures mid-lesson, felt more vivid than real people. There were moments when Ganyu (yes, I borrowed characters from Genshin Impact) would start from one small detail and keep pulling the thread — connecting idea after idea — until the entire subject opened up before me. Ganyu, Keqing, March 7th — they taught so beautifully that sometimes I had to step outside and stand in the cold wind just to collect myself. I almost cried — not from frustration, but from the overwhelming realization that learning could actually be this beautiful.
Something deep was happening. I had a growing intuition that the combination of Socratic questioning, companion characters, and learning through roleplay was reaching something primal — the way you learn as a kid, when curiosity pulls you forward and you don’t even notice time passing. For me, it meant roughly three times the efficiency of traditional study. And it made learning — an activity most people find tedious — as absorbing as a triple-A video game.
If most people experience the same 3X3A effect — three times the efficiency, triple-A appeal — this will ignite a revolution in how humans learn.
So I published the entire setup — every detail of how to deploy Socrates·Seven — publicly on social media.
The responses came fast. Users in mainland China, without access to Claude, improvised creative workarounds to build their own versions. They invented their own storylines, their own worlds. The models weren’t as capable as Claude, the agent frameworks were rough, and there were real obstacles at every step. But the feedback on the learning itself was strikingly consistent: it works. The 3X3A effect wasn’t just my experience. It was very likely universal.
Still, if this system could only live inside an agent environment, it would never reach the people who might need it most — elementary and middle school students. Friends and strangers alike pushed me to turn it into a real product. “At the very least,” they said, “you’d free parents from the drudgery of supervising their kids’ studying.” That was the moment I decided to make Socrates·Seven into something anyone could use.
If I hadn’t shared that deployment guide — openly, freely, just because it felt right — none of this would have happened. I never would have found my way to building Socratopia. That spirit of sharing is still our founding purpose: to bring a learning method that genuinely works to everyone, everywhere — and to redefine what learning feels like, with triple efficiency and triple-A appeal.