Move model notes above tests, restyle pink and tighter

index.html

@@ -18,7 +18,7 @@
     --canvas:#0a0a0b; --bg-panel:#1c1a22; --bg-higher:#26232d; --line:#3f3f46; --hard:#000;
     --ink:#fafaf9; --ink-soft:#e4e4e7; --ink-muted:#a1a1aa; --ink-dim:#71717a;
     --brand:#e11d48; --brand-soft:#fb7185; --brand-deep:#9f1239;
-    --accent:#8b5cf6; --accent-soft:#a78bfa; --gold:#eab308; --gold-soft:#fcd34d; --amber:#f59e0b;
+    --accent:#8b5cf6; --accent-soft:#a78bfa; --gold:#eab308; --gold-soft:#fcd34d; --amber:#f59e0b; --pink:#ec4899; --pink-soft:#f9a8d4;
     --font-display:'Cinzel',Georgia,serif; --font-serif:'Cormorant Garamond',Garamond,Georgia,serif;
     --font-body:'Inter',system-ui,-apple-system,'Segoe UI',sans-serif; --font-mono:'JetBrains Mono',ui-monospace,monospace;
     --border:3px solid var(--hard); --shadow:6px 6px 0 0 var(--hard); --shadow-sm:3px 3px 0 0 var(--hard);
@@ -77,8 +77,9 @@
   blockquote{margin:6px 0 26px; padding:0; text-align:center}
   blockquote p{margin:0 auto; max-width:760px; font-family:var(--font-display); font-weight:700; font-size:clamp(17px,2.1vw,23px); letter-spacing:.04em; line-height:1.35; color:var(--ink-soft); text-shadow:2px 2px 0 var(--hard)}
 
-  .note{font-size:14px; color:var(--ink-muted); border-left:4px solid var(--accent); background:rgba(139,92,246,.07); padding:.7em 1.1em; margin:2px 0 22px}
-  .note strong{color:var(--ink)} .note a{color:var(--accent-soft)}
+  .note{font-size:13.5px; color:var(--ink-muted); border-left:3px solid var(--pink); background:rgba(236,72,153,.06); padding:.55em 1em; margin:0 0 10px}
+  .note:last-of-type{margin-bottom:22px}
+  .note strong{color:var(--pink-soft)} .note a{color:var(--pink-soft)}
 
   ul{margin:0 0 16px; padding-left:1.3em} li{margin:.32em 0} li::marker{color:var(--brand)}
 
@@ -127,6 +128,8 @@
 
   <h2>The <span class="em">PlotPoints</span> Benchmark</h2>
   <p>A leaderboard for how well models actually <em>roleplay</em>: staying in character, holding a scene across many turns, prose quality, and not hijacking your agency. Not a knowledge or willingness test, and not one vibe score. SFW and NSFW.</p>
+  <p class="note"><strong>Requesting a model:</strong> we'll run just about anything within reason, as long as it's on <a href="https://openrouter.ai">OpenRouter</a>. We already add the popular ones we come across, and we take requests, so if it's on OR it can go on the board.</p>
+  <p class="note"><strong>Local models and finetunes:</strong> we actively want to bench them, and we're hunting for an OpenRouter-style aggregator that hosts them in real abundance. The only blocker is fairness: every local model is set up differently and runs differently on every machine, and mismatched provider setups can quietly poison results. Point us at one consistent aggregator with a deep library and we will run them regularly and wildly. Know one? Tell us in the Community tab.</p>
   <p>The difference from the boards you already know: <strong>nothing here is private.</strong> The test harness, the judge prompts, the scoring rubric, and the raw vote dumps are all open-source (CC-BY 4.0), so every number traces back to a file you can read and rerun. Six tests, each measuring a different facet:</p>
   <ul>
     <li><strong>Arena ELO.</strong> Single-turn pairwise blind voting. Humans pick the better reply, blind.</li>
@@ -137,8 +140,6 @@
     <li><strong>Calibration.</strong> Quality control on the votes themselves; roughly 1 in 8 ballots is a trap to catch random clickers.</li>
   </ul>
   <p>Every model runs through OpenRouter to keep providers comparable. Human votes and machine judging both count, and where they disagree is half the fun.</p>
-  <p class="note"><strong>Requesting a model:</strong> we'll run just about anything within reason, as long as it's on <a href="https://openrouter.ai">OpenRouter</a>. We already add the popular ones we come across, and we take requests, so if it's listed on OR it can go on the board.</p>
-  <p class="note"><strong>Local models and finetunes:</strong> we actively want to bench them, and we're actively hunting for an OpenRouter-style aggregator that hosts them in real abundance. The only thing holding us back is fairness: every local model is set up differently and runs differently on every machine, and mismatched provider setups can quietly poison results. The day the community points us at one consistent aggregator with a deep library, we will run them regularly and wildly. Know one? Tell us in the Community tab.</p>
   <div class="stats">
     <div class="stat"><div class="n">3,000+</div><div class="l">Human Votes</div></div>
     <div class="stat"><div class="n">20+</div><div class="l">Models</div></div>