About
I'm an indie developer. Been writing code for over a decade.
Over the years I've shipped tools, office systems, collaboration software, and process automation for companies of various sizes — small shops to larger orgs. The work has spanned streaming, media, AI image generation, gaming, e-commerce, online finance, and payment infrastructure — all consumer-facing. You see a lot of patterns that way: how engineering meshes with the business side, why products die, what users actually don't care about. Every round of work makes those clearer.
For the past few years, most of my time has gone into wrestling with AI agents and multi-agent workflows — keeping an LLM running for more than half an hour without crashing, keeping multiple agents from stepping on each other, pushing every experiment result back into the codebase reproducibly. From AI as a side assistant to AI running long stretches of autonomous development, I've been along for the whole arc. The notes pile up. The memory pile is even bigger. I'll be unpacking it here, gradually.
This site isn't a strategy. It's what was left after years of trying things and cutting them all down — the one thing I still wanted to do.
What goes here
Notes on AI and agents — not news roundups. Stuff I've run, stuff I've broken, what I think after reading the paper. When industry terms get muddled together, I'll pull them apart.
Indie engineering perspective — how one person builds a product, makes decisions, and trips over the same things repeatedly. Small scale doesn't mean small problems.
The occasional essay — not always engineering-related. I write when I have something to write, not to feed an algorithm.
On voice
Chinese and English run in parallel — same topic, written twice. Chinese tends toward the conversational; English tends toward the rigorous.
No publishing schedule. When there's something, I write it. When there isn't, I don't.