Enhance "Tiny Recursive Model" project page with a new embedded YouTube video for better engagement. Update styling for links and improve image handling in the Markdown renderer to accommodate architecture diagrams. Revise subtitle for clarity in the content file.

Author: vukrosic

app/blog/tiny-recursive-model/page.tsx

@@ -186,7 +186,7 @@ export default function TinyRecursiveModelProject() {
               )}
 
               {/* Links to Paper and GitHub */}
-              <div className="flex flex-wrap items-center justify-center gap-4 mb-8">
+              <div className="relative z-10 flex flex-wrap items-center justify-center gap-4 mb-8">
                 <a 
                   href="https://arxiv.org/pdf/2510.04871v1"
                   target="_blank"
@@ -218,7 +218,7 @@ export default function TinyRecursiveModelProject() {
               </div>
 
               {/* Glow effect for the title */}
-              <div className="absolute inset-0 text-4xl md:text-5xl lg:text-6xl font-medium leading-tight blur-sm">
+              <div className="absolute inset-0 text-4xl md:text-5xl lg:text-6xl font-medium leading-tight blur-sm pointer-events-none">
                 <span className="bg-gradient-to-r from-blue-400/20 via-purple-400/20 to-cyan-400/20 bg-clip-text text-transparent">
                   {heroData?.title || 'Tiny Recursive Model'}
                 </span>
@@ -280,6 +280,18 @@ export default function TinyRecursiveModelProject() {
 
               {/* Article Body */}
               <div className="px-8 sm:px-12 pb-20">
+                <div className="mb-8">
+                  <div className="relative" style={{ paddingTop: '56.25%' }}>
+                    <iframe
+                      className="absolute top-0 left-0 w-full h-full rounded-lg shadow-2xl"
+                      src="https://www.youtube.com/embed/51uVH-sEvk0"
+                      title="YouTube video player"
+                      frameBorder="0"
+                      allow="accelerometer; autoplay; clipboard-write; encrypted-media; gyroscope; picture-in-picture"
+                      allowFullScreen
+                    ></iframe>
+                  </div>
+                </div>
                 <div className="prose prose-lg prose-invert max-w-none">
                   <MarkdownRenderer content={markdownContent} />
                 </div>

components/markdown-renderer.tsx

@@ -106,6 +106,10 @@ export function MarkdownRenderer({ content }: MarkdownRendererProps) {
           img: ({ src, alt }) => {
             if (!src) return null;
 
+            // Check if this is the architecture diagram that should be smaller
+            const isArchitectureDiagram = alt?.includes('Architecture') || src?.includes('architecture');
+            const imageClassName = isArchitectureDiagram ? "w-1/2 h-auto mx-auto" : "w-full h-auto";
+            
             // Handle external images
             if (typeof src === 'string' && src.startsWith('http')) {
               return (
@@ -115,7 +119,7 @@ export function MarkdownRenderer({ content }: MarkdownRendererProps) {
                     alt={alt || ''}
                     width={800}
                     height={400}
-                    className="w-full h-auto"
+                    className={imageClassName}
                     loading="lazy"
                   />
                   {alt && (
@@ -135,7 +139,7 @@ export function MarkdownRenderer({ content }: MarkdownRendererProps) {
                   alt={alt || ''}
                   width={800}
                   height={400}
-                  className="w-full h-auto"
+                  className={imageClassName}
                   priority={false}
                 />
                 {alt && (

public/content/tiny-recursive-model/tiny-recursive-model-content.md

@@ -9,7 +9,7 @@ hero:
 
 ## The New AI Reasoning Architecture 
 
-#### How a 7M Model Beats 1T Models at Sudoku, Mazes, ARC-AGI
+How a 7M Model Beats 1T Models at Sudoku, Mazes, ARC-AGI
 
 **The Tiny Reasoning Model (TRM)** uses a 2-layer transformer (7M parameters) that reuses the same layers hundreds of times to reason about problems recursively.