Add detail: problem, 3-axis comparison table, why-it-matters

index.html

@@ -93,6 +93,12 @@
   .badge:hover{border-color:var(--vkie);color:var(--vkie)}
   .badge b{color:var(--ink)}
 
+  .band{display:grid;grid-template-columns:repeat(auto-fit,minmax(250px,1fr));gap:12px;margin-top:8px}
+  .icard{background:var(--panel);border:1px solid var(--line);border-radius:14px;padding:18px}
+  .icard .ci{width:46px;height:46px;border-radius:50%;display:flex;align-items:center;justify-content:center;font-size:22px;margin-bottom:11px}
+  .icard h4{margin:0 0 4px;font-size:17px;font-family:var(--mono)}
+  .icard p{margin:0;font-size:13px;color:var(--muted)}
+
   .note{margin-top:22px;padding:15px 17px;border:1px dashed var(--line);border-radius:12px;font-size:12.5px;color:var(--muted);background:var(--panel)}
   .note b{color:var(--ink)}
   footer{margin-top:40px;padding-top:20px;border-top:1px solid var(--line);font-size:12px;color:var(--muted);font-family:var(--mono);display:flex;justify-content:space-between;flex-wrap:wrap;gap:8px}
@@ -105,6 +111,15 @@
   <h1 class="wordmark">VKIE <span class="nick">비키</span><small>VIDRAFT Kernel Inference Engine</small></h1>
   <p class="tagline"><b>VKAE</b> accelerates. <b>VKUE</b> saves. <b>VKIE</b> maximizes serving. One 34.7B model, from a datacenter GPU down to a free CPU — <b>every number measured</b>, every demo live.</p>
 
+  <!-- PROBLEM -->
+  <h2>The problem</h2>
+  <p class="h2sub">Speed, cost, and scale pull against each other — win one, usually lose two.</p>
+  <div class="band">
+    <div class="icard"><div class="ci" style="background:var(--vkae-soft);color:var(--vkae)">⚡</div><h4>Speed</h4><p>Slow inference is unusable — but going fast normally means an expensive datacenter GPU.</p></div>
+    <div class="icard"><div class="ci" style="background:var(--vkue-soft);color:var(--vkue)">💸</div><h4>Cost</h4><p>Serving a large model means a multi-GPU cluster. With GPU shortages, you often can't even buy them.</p></div>
+    <div class="icard"><div class="ci" style="background:var(--vkie-soft);color:var(--vkie)">📈</div><h4>Scale</h4><p>Model and hardware get locked in — no single engine spans datacenter down to edge / on-prem.</p></div>
+  </div>
+
   <!-- 3 AXES -->
   <div class="axes">
     <div class="axis vkae">
@@ -160,6 +175,20 @@
     </div>
   </div>
 
+  <!-- COMPARISON -->
+  <h2>Three axes at a glance</h2>
+  <p class="h2sub">Same 34.7B model, same principles — three optimization targets.</p>
+  <div class="tablecard"><div class="scroll"><table>
+    <thead><tr><th></th><th>🏎️ VKAE</th><th>🚗 VKUE</th><th>🚄 VKIE</th></tr></thead>
+    <tbody>
+      <tr><td>Focus</td><td>Speed</td><td>Savings</td><td>Unified · throughput</td></tr>
+      <tr><td>Best hardware</td><td>Datacenter GPU</td><td>CPU ~ small GPU</td><td>Full range</td></tr>
+      <tr><td>Strength</td><td>Top single-stream speed</td><td>Lowest cost · accessibility</td><td>Max tok/s · cost-efficiency</td></tr>
+      <tr><td>Measured before→after</td><td class="num">24→220 (9×)</td><td class="num">5.4→20 (3.7×)</td><td class="num">24→18,057 (750×)</td></tr>
+      <tr class="peak"><td>In a phrase</td><td>fastest</td><td>cheapest</td><td>most</td></tr>
+    </tbody>
+  </table></div></div>
+
   <!-- SPECTRUM -->
   <h2>One model, the whole spectrum · measured</h2>
   <p class="h2sub"><b>Ourbox-35B-JGOS</b> — 34.7B total / ~3B active MoE. Same weights, only the hardware changes.</p>
@@ -175,6 +204,15 @@
   </table></div></div>
   <p class="h2sub" style="margin-top:14px">Quality holds across every tier — <b>GPQA Diamond 86.4%</b> (Ourbox-35B, maj@8) up to <b>90.9%</b> (Darwin-398B). Multimodal: Janus-Pro-1B image generation on a low-cost T4 in <b>~28 s/image</b> (fp16, 2.1× our first cut).</p>
 
+  <!-- WHY IT MATTERS -->
+  <h2>Why it matters</h2>
+  <p class="h2sub">Breaking the rule that "big AI needs big money."</p>
+  <div class="band">
+    <div class="icard"><div class="ci" style="background:var(--vkie-soft);color:var(--vkie)">🏛️</div><h4>Sovereign AI</h4><p>Public sector, defense, healthcare, finance — data that can't touch the cloud. Frontier reasoning on an air-gapped on-prem CPU.</p></div>
+    <div class="icard"><div class="ci" style="background:var(--vkue-soft);color:var(--vkue)">💰</div><h4>Cost collapse</h4><p>From a multi-hundred-thousand-dollar GPU cluster to a ~$1,600 card — or a free CPU. Entry cost drops by orders of magnitude.</p></div>
+    <div class="icard"><div class="ci" style="background:var(--vkae-soft);color:var(--vkae)">🌍</div><h4>Accessibility</h4><p>Individuals, startups, SMBs, public bodies — anyone. Ready for the surge in on-device and edge demand.</p></div>
+  </div>
+
   <!-- LIVE TESTS (reuse existing spaces) -->
   <h2>Live tests · try it yourself</h2>
   <p class="h2sub">These run on their <b>real hardware</b> (not this page). Click a card to open, or load one on screen below.</p>