About

Hey you, Lan here. Lately I have been focused on agent systems and agent runtime / eval infrastructure. I care about how LLMs can reliably use tools, manage context, coordinate multi-agent work, and leave reproducible, evaluable evidence across real environments and long-horizon tasks. I am building and maintaining Neo Coworker, a file-system based agent runtime / desktop app that includes core agent system infrastructure such as telemetry / observability, eval harness, and context management. I am also moving further toward the model side, with the goal of connecting system implementation, experimental evaluation, and product experience across engineering and research. I follow frontier AI community discussions and keep distilling high-quality ideas, translations, and practical notes on my personal site and social media. Ooh, and if your party needs a DJ—hit me up anytime. XD