Trainer: tell the words model who opens each mail

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01GT37Z1Xtfd9pUuQtMTg6Yt
2026-10-01 10:41:27 +02:00
parent 1929d2f44e
commit 8f24e5483c
2 changed files with 37 additions and 1 deletions
@@ -4,6 +4,10 @@ One line per change, grouped by the release that shipped it. Newest
 first. A release is a Cargo.toml bump plus a `v<version>` tag; the
 content repos pin the tag they run (see the end of this file).

+## 0.6.12 (2026-10-01)
+
+- Trainer: the words model is told who opens each mail: the person who sent the case, the person who posted the record an answer is about, or someone at the business. Mails to the business about another person's case used to be written as if to that person ("you will hear from us"); on Gullberg, three of nine were.
+
 ## 0.6.11 (2026-10-01)

 - `wordmark_ink: true` in `site.yaml`: an SVG wordmark is drawn in the page's own ink, as a mask over the text colour, so it follows the page's light and dark in every browser. A wordmark shown as an `<img>` only knows the system's setting, and some browsers don't pass even that on. Opacity in the file still shows, so a lighter line stays lighter. Opt-in, for one-colour marks; a mark with a colour of its own keeps its colours.