Field Notes measured from China

About

One person in China, five sections, and a rule about numbers.

I test AI tooling for a living: building the task books, acceptance scripts and reporting pipelines that decide whether an AI coding agent is actually usable, or only looks usable in the demo. Field Notes is where the reusable part of that work goes — the method rather than the announcement.

The same habit runs through the rest of the site. I live in China, and a lot of what is written about this place from outside is either a press release or a mood. The part I can add is the ordinary version: what a factory quotation actually contains, what freight costs this month, what a market stall charges, and where the number in a headline came from.

Two streams come out of this. The long pieces — the ones with the tables and the sources — land in Posts, roughly every two weeks. Between them Notes carries something shorter: a link, the number inside it, and why it was worth my time. The sites that keep me honest are listed on Reading.

The five sections

  • AI and engineering — the main line, and the one the newsletter carries: evaluation method, pipelines, and the numbers worth keeping, from model prices to page weight.
  • Supply chain — sourcing, factory quotes, freight rates, month by month.
  • Life in China — everyday life here, sorted into what has actually been measured and what is only habit.
  • Data and the world — open datasets, charted: global flows, and what they look like up close in one place.
  • 海外华人 — the one Chinese-language section, written for readers already outside: cost of living, cross-border price checks, paperwork.

What you will and will not find

Published: evaluation methodology, failure taxonomies, pipeline notes, public datasets, and the occasional write-up of something that broke in a way worth remembering.

Not published: anything under an NDA, anything about a specific employer's internals, anything about family or anyone else who did not choose to be written about, and anything I have not been able to reproduce myself.

Why there is no name on this

The site is written under a pen name, and that is a consequence of the second half of what I write about: factories, suppliers and prices are conversations with people who did not agree to be in a blog. A real name attached to a day job makes that harder for everyone involved. So the rule is public data, my own measurements and methods — no private conversations, and no identifying details about the places and people behind them.

Contact

There is no public address here yet. For now the fastest way to reach me is a reply on whichever platform you found the post — those I read myself.