My years in big game studios taught me one thing: in mass markets, deep tech is just a cost line. Since 2024 I've been running the opposite experiment — a one-person studio making things where depth is the product.

I started in 2016 at NetEase, doing engine work — a terrain authoring tool, vegetation systems, streaming for an open-world action game. The fun kind of problem: how do you draw ten thousand blades of grass on a mobile GPU without melting it?

Then Ubisoft Shanghai, as a Technical Artist on AAA titles. Same industry, different universe: a thousand people shipping one game. My real education was watching how that machine coordinates — the pipeline is the organization, and most bottlenecks are people problems wearing technical hats. That's also where I learned what deep tech is worth in a mass market: nobody buys a game for its vegetation system — people only pay for what they can perceive. The depth was real; on the spreadsheet, it was overhead.

In 2021 I joined Taichi Graphics (later rebranded to Meshy.ai) as a Staff Engineer and built Taitopia Renderer from the ground up — an online rendering platform with an art asset pipeline and a front-end editor. React, Blender Python, the works. It was the first product I owned end to end, and it rewired how I think about software: your users aren't teammates anymore, they're strangers who leave when something breaks. The lessons that stuck were product lessons, not technical ones — start from a specific user with a real need, and validate before you optimize anything.

Since 2024 I've been independent — contract work, consulting, and collaborations. Mostly that's been Voxel Playground, a voxel sandbox on Meta Quest. In my own time I run experiments like Wind Tunnel Simulator, a real-time fluid dynamics playground. And I write this blog — sixty-odd articles on rendering, tooling, and game tech; it's also how people who care about this stuff find me.

This experiment only became runnable now. What blocks a one-person product was never depth — it's breadth: frontend, backend, storefront, all the work that used to force a deep specialist into an organization. AI has collapsed that breadth tax. Markets too small to feed a team can now feed one person. And as AI floods the world with sixty-point software, the relative price of depth only goes up. AI is a multiplier; depth is the base. So the answer I'm looking for: a market too small for any big company's cost structure to bother with, deep enough that the work stays interesting — and one product that can sustain a single person for years.