Stats: base-pair FNS likelihood 1.88 vs 1.79; bacteria recovery 35.3%

app.py

@@ -355,9 +355,9 @@ STATS_HTML = _wrap(
     "<div class='l'>cheaper per query<br>(one 74M specialist active)</div></div>"
     "</div>"
     "<table><tr><th>metric</th><th>DaisyChain</th><th>Carbon-500M</th></tr>"
-    "<tr><td>Likelihood — bits/base (↓ better)</td><td class='n'>1.79</td><td class='n'>1.75</td></tr>"
+    "<tr><td>Likelihood — bits/base, base-pair FNS (↓ better)</td><td class='n'>1.88</td><td class='n'>1.79</td></tr>"
     "<tr><td>Seq-recovery, eukaryote (↑ better)</td><td class='n'>31.0%</td><td class='n'>42.2%</td></tr>"
-    "<tr><td>Seq-recovery, bacteria (↑ better)</td><td class='n'>36.5%</td><td class='n'>49.5%</td></tr>"
+    "<tr><td>Seq-recovery, bacteria (↑ better)</td><td class='n'>35.3%</td><td class='n'>49.5%</td></tr>"
     "</table>"
     "<div class='sub'>Four ~74M specialists (≈295M total, <b>under Carbon-500M</b>); only one runs per "
     "query, so it's ~7× cheaper per token. Behind the 500M / 1T-token monolith but within striking "