001 — Sliding-window rate-limit pacing. Why a fixed sleep between calls wastes half your quota, and the window algorithm that doubles throughput on the same limit. Measured at 40 RPM against a real multi-provider gateway.
029 — The 15-second messaging window. WeChat's passive-reply window: what it actually is, why the first reply times out at 3.8 seconds, and the MsgId cache that turns an 8-second answer into a 0.5-second one.
030 — The postmortem template. The one-page format that makes an incident review useful instead of ceremonial, with a filled-in example from a real 180-second silent hang.
Why give these away
Because a recipe you can read in full is a better argument than a table of contents. If these three save you one bad night, the other 27 are worth 19 dollars.
The other 27 recipes
The full cookbook covers throughput, contracts, cost engineering, reliability, evaluation and war stories: