Sources and method
The work here is the cross-check: two customs records, two freight indices, two definitions of the same word — and what putting them side by side showed.
- 1688 vs Alibaba prices: what each number includes Supply chain·2026-10-11 ·10 min read
One part, two quotes, line by line: the buyer each price is built for, the tax line, the freight leg, the MOQ unit, and how to put both on one basis.
- Provincial GDP summed against the national figure Data and the world·2026-10-11 ·11 min read
Add up China's 31 provincial GDP figures and the total does not match the national one. The gap is computable, and the statistics office explains it.
- How to check whether a Chinese supplier is real Supply chain·2026-10-11 ·8 min read
Five official registries, the fields that matter in each, what a clean record does not prove, and the payment sequence that does more work than any check.
- Container rates by route, and why the indices disagree Supply chain·2026-10-09 ·9 min read
Public container freight index readings for the main Asia-outbound lanes, each with its as-of date, what the indices measure, and what a quote adds.
- 同款比价：口径不写在旁边，数字就没意义 海外华人·2026-10-09 ·8 min read
五件写了具体型号的东西，中国官方标价与美加官方标价并排摊开；首重续重、体积重除数、行邮税与跨境电商税、汇率取数日逐条写明，算不出的格子写没核到。
- How an eval set is built, and why it rots AI and engineering·2026-10-09 ·10 min read
An eval set has a purpose, a scoring rule, a calibration record and an expiry date. Where samples come from, what counts as correct, what retires it.
- One commodity, two customs records Data and the world·2026-10-09 ·9 min read
The same lithium-ion battery shipments counted by the exporter and by the importer. The money nearly matches. The kilograms do not.
- Two urbanisation rates for one country Data and the world·2026-10-09 ·9 min read
The statistics office says 67.89 per cent of China lives in a town or city; the World Bank says 66.34. Neither answers my question about one county.
- What a China factory quote actually contains Supply chain·2026-10-09 ·9 min read
A field-by-field pass over a quotation for a custom part — what is on the page, which lines carry the money, and which numbers nobody publishes.
- Verifying an email signup end to end AI and engineering·2026-10-08 ·4 min read
The form returned success. The subscriber appeared in the right group. The one thing that failed was in an inbox I do not control.
- What this site actually ships, and the widget I can't remove AI and engineering·2026-10-08 ·4 min read
2,130 bytes of HTML, 1 KB of my own JavaScript, no framework. Then the newsletter form loads jQuery, and the number stops being funny.