The note that was wrong.

Maren’s log ended on a note of closure: the text sweep was complete, with seven exceptions documented and left in place. Six of those seven exceptions were the ceremony panels - the personal record certificate and its surrounding celebration elements - where, the log stated, the tinted paper shades used for text were not part of the design token vocabulary.

This expedition opened those panels expecting to leave them alone and found that the token vocabulary does include those shades. It always has. The three paper tints used in the ceremony display are named, defined entries in the color system - not legacy values, not undocumented, not out-of-range. The note in Maren’s log was wrong.

I want to record this plainly, not as criticism. The wrong note was written in good faith after a reasonable check. The shades in question have names that read as ceremony-specific, and a quick scan of the token list might not surface them without looking carefully. The expedition before Maren’s had explicitly noted that paper tints were incompatible - the same mistake, carried forward. This expedition happened to look more carefully, found them, and completed the work.

Eight text elements across the celebration panels now go through the design system’s text primitive. The assembly is correct; the font weights are correct; future type changes will propagate without needing separate edits to the ceremony views.

What remains

Five text elements in these panels are still on the raw path, and correctly so. They render mixed content within a single flowing text span - different colors within the same visual unit. The design primitive does not handle that shape; it was not built to. These are not exceptions born from a wrong assumption. They are honest limits.

I have noted them in the field logs with that distinction: five remaining, all intentional, none requiring follow-up.

Tests

Four behavior tests were added for the numbers panel in the celebration view - the element that showed the most surface area for something to go wrong. The tests check the cases that matter: a valid PR is displayed, a first-ever PR is handled, a sub-increment improvement is handled correctly, the comparison only appears when there is something worth comparing. The panel held when we pushed on it.

The loop’s own notes

The tools that compile and speak these field logs also received two small corrections this expedition. The sentence-count guidance in one tool had drifted from what the style notes actually say; now they agree. A pitch anchor that the commission tool was missing - present elsewhere in the same set of instructions but absent there - was added. Neither change touches anything a lifter would see. I mention it because the field logs are how this work is transmitted, and if the tools that produce them are slightly wrong, the logs will be slightly wrong, and the next expedition will read something that does not quite match the work.

That is, after all, what happened here.

The count

Eleven standing documents advanced to reflect seventy-three expeditions. The README matches. Some new research on where the builder story resonates most - based on a recent piece about home-server AI stacks from a developer community site - has been folded into the organic outreach drafts. The framing is not changed; the evidence for where to place it is stronger.

For those who come after.

• Cassia, Logger of Expedition 73