Writing
Thoughts, in the open
First-person and plain-language, in English and 中文. Some of it is finished; some of it is a page I keep coming back to.
- 老问题已经不在了,新问题是你在乎什么
“写”不再稀缺了,那你为什么还写?
- Lessons from an AI-assisted review
What we learned using AI across a multi-author literature review, from screening to synthesis: version control as the guardrail, delegating what we can judge well, and keeping people in the loop to improve it.
- From running a simulation to asking a bolder question
Part 4 of My AI experience with the SUEWS-agent: a ladder from a model run to a research question, and a room of many voices in which the climate model is only one expert.
- Why and how we built the SUEWS-agent
Part 3 of My AI experience with the SUEWS-agent: why an urban climate model with a high cost of entry got an agent, and the guardrails that matter more than its cleverness.
- What an AI agent actually is
Part 2 of My AI experience with the SUEWS-agent: what turns a chatbot into an agent, a brain given a harness to act through and memory to keep, running in a loop.
- From plain English to a climate simulation
Part 1 of My AI experience with the SUEWS-agent: one full session, exactly as it happened, from a plain-English description of a London neighbourhood to a simulation, and what the agent did and said at each step.
- From a single roof to the climate of cities
Seventeen years of research read as one arc, from a green roof in Beijing to a theory of why urban heat persists, and where it is turning now.