Add an AI slop verdict that downrates the author's future articles

A fourth explicit verdict, "AI slop" (Vote::Slop, label `slop`), joins the
rating widget on the web, the EPUB footer links, the /r/ confirmation page,
the CLI, imports and the dashboard filters. It counts as a full negative
(curation.feedback.slop_value, -1.0) in the neighbour and affinity signals.

Beyond that, each run loads the authors whose current verdict is slop, with
no lookback, and multiplies the preliminary blend and the utility of every
candidate by that author by 1 - curation.ranking.slop_author_penalty (0.75),
so they sink before triage. The flag is recorded in signals_json, shown by
`explain` and the dashboard signals table, and the confirmation names the
author (or says no author is known, in which case only the rating applies).

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VXYGBPoHZSDSfE5WcJ9bvj

src/web/templates/dashboard/ratings.html

@@ -9,7 +9,7 @@
 <div class="disclosure-body">
 <form method="post" action="/dashboard/ratings/import" class="filters">
 <label class="w-full">URLs <textarea name="urls" rows="5" required class="w-full" placeholder="One URL per line (commas and spaces also work)" spellcheck="false"></textarea></label>
-<label>Verdict <select name="label"><option value="loved" selected>Loved it</option><option value="good">Good</option><option value="not_for_me">Not for me</option></select></label>
+<label>Verdict <select name="label"><option value="loved" selected>Loved it</option><option value="good">Good</option><option value="not_for_me">Not for me</option><option value="slop">AI slop</option></select></label>
 <label>Note <input name="note" type="text" placeholder="Optional note"></label>
 <div class="filter-actions"><button class="btn btn-primary" type="submit">Queue import</button></div>
 </form>
@@ -34,10 +34,11 @@
 <ol class="list-decimal space-y-3 pl-5 marker:text-muted">
 <li><strong>Prompt verdict block.</strong> The {{ how.verdicts_in_prompt }} most recent non-cleared verdicts, newest first, are written into every LLM call's system prompt as one line each (<code>LOVED | title | feed | summary</code>). Only the rank matters here — a verdict never ages out of this block, it is pushed out by newer ones. Tune <a href="/dashboard/settings#curation.feedback"><code>curation.feedback.verdicts_in_prompt</code></a>.</li>
 <li><strong>Weekly learned adjustments.</strong> Every {{ how.rebuild_interval_days }} days the editor model rewrites the profile's "Learned adjustments" bullets from the {{ how.max_ratings_in_rebuild }} most recent non-cleared verdicts, including notes and deep-assessment facets. See the <a href="/dashboard/profile">Profile</a> page.</li>
-<li><strong>Rated-neighbour signal.</strong> Each verdict with an embedding is an example with weight <code>value × 0.5^(age / {{ how.half_life_days }} days)</code>, where loved = {{ how.loved }}, good = {{ how.good }}, not for me = {{ how.not_for_me }}; ratings older than {{ how.lookback_days }} days are not loaded. A candidate's signal is the weighted mean cosine to its {{ how.neighbour_k }} nearest positive examples minus {{ how.negative_coefficient }} × the same over its nearest negative ones. The signal's preliminary weight ({{ how.knn_weight }}) is scaled by a gate that opens above {{ how.knn_floor }} embedded verdicts and is fully open at {{ how.knn_full }}. Tune <a href="/dashboard/settings#curation.ranking"><code>curation.ranking.rating_half_life_days</code>, <code>knn_floor</code>, <code>knn_full</code>, <code>neighbour_k</code></a> and <a href="/dashboard/settings#curation.ranking.weights.preliminary"><code>weights.preliminary.knn</code></a>.</li>
+<li><strong>Rated-neighbour signal.</strong> Each verdict with an embedding is an example with weight <code>value × 0.5^(age / {{ how.half_life_days }} days)</code>, where loved = {{ how.loved }}, good = {{ how.good }}, not for me = {{ how.not_for_me }}, AI slop = {{ how.slop }}; ratings older than {{ how.lookback_days }} days are not loaded. A candidate's signal is the weighted mean cosine to its {{ how.neighbour_k }} nearest positive examples minus {{ how.negative_coefficient }} × the same over its nearest negative ones. The signal's preliminary weight ({{ how.knn_weight }}) is scaled by a gate that opens above {{ how.knn_floor }} embedded verdicts and is fully open at {{ how.knn_full }}. Tune <a href="/dashboard/settings#curation.ranking"><code>curation.ranking.rating_half_life_days</code>, <code>knn_floor</code>, <code>knn_full</code>, <code>neighbour_k</code></a> and <a href="/dashboard/settings#curation.ranking.weights.preliminary"><code>weights.preliminary.knn</code></a>.</li>
 <li><strong>Feed affinity.</strong> The same decayed weight is credited to the rated article's direct feeds, split evenly; each feed's Beta-smoothed rate <code>(up + 1) / (up + down + 2)</code> becomes a candidate's signal (the mean over its rated direct feeds). Its weight ({{ how.feed_weight }}) is gated between {{ how.feed_floor }} and {{ how.feed_full }} attributable ratings. Tune <a href="/dashboard/settings#curation.ranking"><code>curation.ranking.feed_floor</code>, <code>feed_full</code></a> and <a href="/dashboard/settings#curation.ranking.weights.preliminary"><code>weights.preliminary.feed</code></a>.</li>
+<li><strong>Slop authors.</strong> An <em>AI slop</em> verdict is also a report against the article's author: while it is the article's current verdict, every candidate by that author (same normalized name, any feed, no age limit) has its preliminary blend and utility cut by {{ how.slop_author_penalty }}. Articles without a known author get only the ordinary negative rating. Tune <a href="/dashboard/settings#curation.ranking"><code>curation.ranking.slop_author_penalty</code></a>.</li>
 </ol>
-<p class="muted">Clearing a verdict removes it from all four paths without deleting history; ratings are append-only.</p>
+<p class="muted">Clearing a verdict removes it from all five paths without deleting history; ratings are append-only.</p>
 </div>
 </details>
 
@@ -46,7 +47,7 @@
 {% if tab == "events" %}
 <form class="filters" method="get" action="/dashboard/ratings">
 <input type="hidden" name="tab" value="events">
-<label>Label <select name="label"><option value="">any</option><option value="loved"{% if filter_label == "loved" %} selected{% endif %}>Loved it</option><option value="good"{% if filter_label == "good" %} selected{% endif %}>Good</option><option value="down"{% if filter_label == "down" %} selected{% endif %}>Not for me</option><option value="cleared"{% if filter_label == "cleared" %} selected{% endif %}>Cleared</option></select></label>
+<label>Label <select name="label"><option value="">any</option><option value="loved"{% if filter_label == "loved" %} selected{% endif %}>Loved it</option><option value="good"{% if filter_label == "good" %} selected{% endif %}>Good</option><option value="down"{% if filter_label == "down" %} selected{% endif %}>Not for me</option><option value="slop"{% if filter_label == "slop" %} selected{% endif %}>AI slop</option><option value="cleared"{% if filter_label == "cleared" %} selected{% endif %}>Cleared</option></select></label>
 <label>Source <select name="source"><option value="">any</option>{% for source in sources %}<option value="{{ source }}"{% if filter_source == source.as_str() %} selected{% endif %}>{{ source }}</option>{% endfor %}</select></label>
 <label>User <select name="user"><option value="">any</option>{% for username in usernames %}<option value="{{ username }}"{% if filter_user == username.as_str() %} selected{% endif %}>{{ username }}</option>{% endfor %}</select></label>
 <label>From <input type="date" name="from" value="{{ filter_from }}"></label>
@@ -75,7 +76,7 @@
 {% else %}
 <form class="filters" method="get" action="/dashboard/ratings">
 <input type="hidden" name="tab" value="current">
-<label>Label <select name="label"><option value="">any</option><option value="loved"{% if filter_label == "loved" %} selected{% endif %}>Loved it</option><option value="good"{% if filter_label == "good" %} selected{% endif %}>Good</option><option value="down"{% if filter_label == "down" %} selected{% endif %}>Not for me</option><option value="cleared"{% if filter_label == "cleared" %} selected{% endif %}>Cleared</option></select></label>
+<label>Label <select name="label"><option value="">any</option><option value="loved"{% if filter_label == "loved" %} selected{% endif %}>Loved it</option><option value="good"{% if filter_label == "good" %} selected{% endif %}>Good</option><option value="down"{% if filter_label == "down" %} selected{% endif %}>Not for me</option><option value="slop"{% if filter_label == "slop" %} selected{% endif %}>AI slop</option><option value="cleared"{% if filter_label == "cleared" %} selected{% endif %}>Cleared</option></select></label>
 <label>Source <select name="source"><option value="">any</option>{% for source in sources %}<option value="{{ source }}"{% if filter_source == source.as_str() %} selected{% endif %}>{{ source }}</option>{% endfor %}</select></label>
 <label>Feed <select name="feed"><option value="">any</option>{% for feed in feeds %}<option value="{{ feed.id }}"{% if filter_feed == feed.id.to_string() %} selected{% endif %}>{{ feed.title }}</option>{% endfor %}</select></label>
 <label>Title <input type="search" name="q" value="{{ filter_q }}" placeholder="contains…"></label>