Abstract: Wrapping up my second year of exploration from February 2025 to January 2026

When my first startup experience ended in early February 2025, the safe path was clear: return to big tech or join an AI lab or some hot startup, take a breather, reset. Instead, I doubled down. This second year of exploration taught me that finding your direction isn’t about discovering what you’re capable of—it’s about discovering what you can’t walk away from.

February - April: Testing the Limits

I started the year building in the open. I established PathOnAI.org as a 501(c)(3) nonprofit, creating a collaborative space for AI agent research. Over three months, I published two papers at top-tier venues (NAACL, ECML-PKDD) and shipped multiple AI-native projects spanning agent reasoning, Roblox AI NPCs, and full-stack infrastructure.

In April, after 11 years trapped in the USA by visa restrictions, I finally traveled internationally again. Three weeks across Shanghai, Hong Kong, Macau, Tokyo, and Hakone. The trip reminded me what I’d been missing—not just places, but the perspectives that come from moving through different worlds.

But the real question wasn’t whether I could build across research, engineering, infrastructure, and community. I could. The question was: what should I build?

May - June: When Capability Meets Purpose

By mid-May, I faced a different kind of challenge. Empowered with AI tools, I could do almost anything: full-stack development, infrastructure design, model training, hardware prototyping, research, community building. The problem wasn’t ability. It was direction.

Capability isn’t the same as purpose.

I needed something I wouldn’t be able to walk away from—something that would keep me excited for at least a decade, not because it was difficult, but because it mattered deeply. The answer came from a belief I’ve carried for years: knowledge should never be gatekept. Too often, experts build walls around complexity and treat access as currency. I’ve always found more meaning in the opposite—breaking down barriers so more people can build.

That conviction led me to open source robotics. In June, as team leader, I won the SF Bay Area LeRobot Hackathon. The direction was set.

July - August: The Cost of Commitment

What followed was two months of controlled chaos. I experimented with collaborators, pushed on technical accessibility, and raised a funding round. Some weeks tested everything I thought I knew—about people, about money, about my own limits. I learned hard lessons. I did things that made me uncomfortable. But I was beginning to understand that discomfort is the tax you pay for commitment.

September - January: Building From Zero

In September, my venture officially launched. After more than a decade in the USA, I returned to Shanghai—this time to build a team. I took a builder-first approach: before hiring anyone, I worked through every layer of our technical stack myself, from AI models to infrastructure to frontend. Not to prove I could do everything, but to understand our needs well enough to recognize the right people when I found them.

Over the next 1.5 months, I moved between Shanghai and Shenzhen, culminating in IROS 2025 in Hangzhou. At the conference, I injured my feet but kept walking the exhibition floor anyway—there were people to meet, companies to connect with, conversations that couldn’t wait. By the day before my flight, I was in the hospital. The journey home—Hangzhou to Hong Kong to SFO in a wheelchair, with help from airport service personnel—took nearly 24 hours. Every connection I made was worth it.

Back in the US, November and December became a rhythm of recharge and execution. I traveled to Vancouver for Thanksgiving and Las Vegas before Christmas, having become unexpectedly good at credit card rewards optimization. Between trips, I was deep in execution: building alongside early team members and actively hiring—designing take-home projects, evaluating candidates, searching for people who shared not just the skills but the vision.

January shifted into pure momentum. Building product, building team, creating content, building in public. The pieces were finally coming together.

What I Learned

Year two taught me that exploration isn’t about trying everything. It’s about finding what you can’t abandon, even when the cost is real.

I went from publishing papers to building a robotics company. From working solo to assembling a team across continents. From capability to clarity. The year left its marks—injured feet, sleepless weeks, the constant tension between building and hiring. But it also gave me something more valuable: conviction.

I know what I’m building and why. I know the technical stack inside out. I know the kind of people I need beside me. Year three begins with a team, a vision, and momentum.

The foundation is set. Now we build.