<span class="span-reading-time rt-reading-time" style="display: block;"><span class="rt-label rt-prefix">Reading Time: </span> <span class="rt-time"> 4</span> <span class="rt-label rt-postfix">minutes</span></span>
<figure class="wp-block-image size-full"><img loading="lazy" decoding="async" width="960" height="540" src="https://octet-gatsby.in2.cdn-alpha.com/wp-content/uploads/2025/06/Frame-1-2-1.webp" alt="The Illusion of Thinking" class="wp-image-23571" srcset="https://octet-gatsby.in2.cdn-alpha.com/wp-content/uploads/2025/06/Frame-1-2-1.webp 960w, https://octet-gatsby.in2.cdn-alpha.com/wp-content/uploads/2025/06/Frame-1-2-1-300x169.webp 300w, https://octet-gatsby.in2.cdn-alpha.com/wp-content/uploads/2025/06/Frame-1-2-1-768x432.webp 768w" sizes="auto, (max-width: 960px) 100vw, 960px" /></figure>



<p>Just days before Apple’s much-anticipated WWDC 2025, the tech giant quietly dropped a research paper that may have done more to shake up the AI world than anything announced on stage.</p>



<p>Titled “The Illusion of Thinking,” the <a href="https://ml-site.cdn-apple.com/papers/the-illusion-of-thinking.pdf" target="_blank" rel="noreferrer noopener nofollow">research </a>paper delivers a blunt and unsettling message: large reasoning models (LRMs)—including OpenAI&#8217;s o1 and o3, Claude 3.7 Sonnet Thinking, DeepSeek R1, and Google’s Gemini Flash Thinking—completely collapse when confronted with harder logic puzzles.</p>



<p>Yes, collapse. Not underperform. Not make a few mistakes. They give up.</p>



<p>And if you’re one of those riding the AGI hype train? This may be the cold bucket of reality you didn’t see coming.</p>



<h2 class="wp-block-heading">Apple’s shocking findings on AI eeasoning</h2>



<p>The research, conducted by Apple’s AI team, revisits classic logic puzzles:</p>



<ul class="wp-block-list">
<li><strong>Tower of Hanoi</strong>—Move stacked disks from one peg to another using strict rules.</li>



<li><strong>Jumping Checkers</strong>—Move pieces over others into empty spaces using logic.</li>



<li><strong>River Crossing Problem</strong>—Classic chicken-fox-grain problem where items must cross in a boat with constraints.</li>



<li><strong>Block Stacking</strong>—Arrange blocks in a specific sequence using spatial reasoning.</li>
</ul>



<p>They increased difficulty systematically—for instance, adding more disks to the Tower of Hanoi or more constraints in the river crossing puzzle.</p>



<p>&nbsp;You probably solved them in a school math class or stumbled across them in a logic game online.</p>



<p>But when Apple&#8217;s researchers used them to test the “reasoning” of top AI models, the results were damning:</p>



<p>“Accuracy progressively declines as problem complexity increases until reaching complete collapse (zero accuracy) beyond a model-specific complexity threshold,” Apple’s paper states.</p>



<p><strong>Apple found a three-stage behavior pattern across all LRMs:</strong></p>



<figure class="wp-block-table"><table class="has-fixed-layout"><tbody><tr><td><strong>Complexity</strong></td><td><strong>Behavior</strong></td><td><strong>Insight</strong></td></tr><tr><td><strong>Low</strong></td><td>Surprisingly, traditional LLMs outperformed LRMs</td><td>Possibly due to overthinking or misalignment in “reasoning” mechanisms</td></tr><tr><td><strong>Medium</strong></td><td>LRMs improved and outperformed LLMs</td><td>This is their “sweet spot”</td></tr><tr><td><strong>High</strong></td><td>All models collapsed in accuracy (approaching 0%)</td><td>Total reasoning breakdown</td></tr></tbody></table></figure>



<ul class="wp-block-list">
<li>Claude 3.7 Sonnet + Thinking and DeepSeek R1 began to fail when a fifth disc was added to the Tower of Hanoi puzzle.</li>



<li>Even with more computing power, the models still failed to solve more complex problems.</li>



<li>The models initially spend more tokens (i.e., “thinking effort”) as problems get harder — but after a certain point, they begin spending fewer tokens, essentially giving up.</li>
</ul>



<p>Let that sink in: as the puzzle gets tougher, the AI literally thinks less.</p>



<h2 class="wp-block-heading">AI can code and write, but can it really reason?</h2>



<p>You might be wondering, “Why is this a big deal? AI still performs well on math and code, right?”</p>



<p>Sure. But logic and reasoning are cornerstones of human intelligence—and more importantly, of Artificial General Intelligence (AGI). </p>



<p>For learners, pursuing a <a href="https://online.lifelonglearning.jhu.edu/jhu-certificate-program-applied-generative-ai" target="_blank" rel="noreferrer noopener">Gen AI certification</a> can help bridge the gap between today’s practical AI skills and understanding where the technology still falls short.</p>



<p>And if these LRMs—fine-tuned to do thinking tasks—fail on puzzles many humans can solve, it throws a wrench into the narrative that AGI is around the corner.</p>



<p>Even more alarming: when researchers fed the models the correct algorithm upfront, accuracy didn’t improve.</p>



<p>Yes, even with step-by-step instructions included in the prompt, the models still failed.</p>



<h2 class="wp-block-heading">Apple’s reluctance to join the AI arms race</h2>



<p>Until now, Apple has seemed like a reluctant participant in the generative AI boom. </p>



<p>While <a href="https://www.google.co.in/" target="_blank" rel="noreferrer noopener nofollow">Google</a> and Samsung have packed their devices with flashy AI features, Apple stayed relatively quiet—until WWDC 2025, where it finally launched “Apple Intelligence.”</p>



<p>But even then, reactions were lukewarm. &#8220;Underwhelming,&#8221; some said.</p>



<p>Now, this research may explain why. Apple may be signaling a radically different—and cautious—approach to AI: don’t fake thinking. Don’t pretend reasoning is solved. Build from first principles. Test it rigorously.</p>



<p>And to be fair, it’s not all bad news. Apple’s paper doesn’t say that AI doesn’t reason at all. It just points out that reasoning ability plateaus fast—and then collapses.</p>



<h2 class="wp-block-heading">The AGI skeptics are having a moment</h2>



<p>For AGI skeptics and critics of the AI hype cycle, this research is pure vindication.</p>



<figure class="wp-block-embed is-type-rich is-provider-twitter wp-block-embed-twitter"><div class="wp-block-embed__wrapper">
<blockquote class="twitter-tweet" data-width="500" data-dnt="true"><p lang="en" dir="ltr">Apple just GaryMarcus&#39;d LLM reasoning ability <a href="https://t.co/735UMGk4be">pic.twitter.com/735UMGk4be</a></p>&mdash; Josh Wolfe (@wolfejosh) <a href="https://twitter.com/wolfejosh/status/1931182279755178074?ref_src=twsrc%5Etfw">June 7, 2025</a></blockquote><script async src="https://platform.twitter.com/widgets.js" charset="utf-8"></script>
</div></figure>



<p>He’s not wrong. LLMs—no matter how powerful—aren’t magic. They’re statistical machines predicting the next token. Coding and summarizing? Sure. Deep reasoning? Not yet.</p>



<p>And here’s where it gets spicier: Apple’s researchers didn’t compare the models to human results. Why? Because many humans also fail at, say, Tower of Hanoi with 8 discs.</p>



<p>So yes, AI models aren’t “superintelligent.” But they might not be worse than average humans either—depending on the task.</p>



<h2 class="wp-block-heading">So, is AGI a mirage?</h2>



<p>Maybe. Or maybe we’re just not there yet.</p>



<p>This isn’t the death of AI. But it is a signal—a loud, inconvenient, and data-backed signal—that much of what’s being marketed as &#8220;thinking&#8221; by <a href="https://octet.design/journal/top-ai-tools/" target="_blank" rel="noreferrer noopener">AI tools</a> is really just a performance illusion.</p>



<p>It performs well… until it doesn’t.<br>It reasons well… until it breaks.<br>It tries hard… until it gives up entirely.</p>



<p>The real question is, are we ready to stop pretending?</p>



<h2 class="wp-block-heading">Final thoughts</h2>



<p>Let’s be clear. Apple didn’t release this paper to dunk on its competitors. They built these LRMs too. The real value of this research is the transparency and realism it brings to a field too often intoxicated by its own marketing.</p>



<p><strong>Here’s what the research teaches us:</strong></p>



<ul class="wp-block-list">
<li>Reasoning is fragile in current AI models.</li>



<li>More tokens ≠ better thinking.</li>



<li>AGI is not just a scale problem. Throwing more data or more parameters may not solve it.</li>
</ul>



<p>This paper is not an obituary for artificial intelligence. But it is a reality check. And we needed one.</p>
<footer class="author-bio-section" ><div class="author-flex"><div class="author-image"><img loading="lazy" decoding="async" width="90" height="90" src="https://octet-gatsby.in2.cdn-alpha.com/wp-content/uploads/2024/01/image-1-150x150.png" class="attachment-90x90 size-90x90" alt="Creative Director and Founder of Octet Design Studio- Aakash Jethwani" /></div><div class="author-info"><div class="author-name"><strong>Aakash Jethwani</strong></div><p class="author-bio">With over 12 years of experience and 300+ successful projects, Aakash Jethwani is a recognized design expert. As the founder and creative director of Octet Design Studio, he leads a team of 28+ designers and developers, delivering pixel-perfect designs that balance creativity and technology.

Aakash is known for crafting tailored design solutions that help businesses stand out in competitive markets. His commitment to innovative strategies and exceptional customer experiences drive sustainable growth for his clients, making him a trusted partner for business transformation.</p></div> </div><div class="authorbox-social-icons"> <a href="mailto:aakash@octet.design" target="_blank" rel="nofollow" title="E-mail" class="tooltip"><i class="fa fa-envelope-square fa-2x"></i> </a></div></footer>

Apple’s “The Illusion of Thinking” Is a Wake-Up Call for AI Industry

Learn how localized UI design adapts interfaces for different regions, shaping layouts, flows and user experiences to match local behaviours.

The Nuances Behind Localized UI Design Decisions

Creating a serious-looking website is not about minimalism alone or premium visuals in isolation. 

How to Create a Serious-Looking Website That Builds Credibility?

Color is not an afterthought or a finishing touch. It’s part of how ideas are structured, understood and remembered.

A Guide To Using Color In Presentation And Its Psychological Effect

Design systems for niche SaaS represent investments that don't pay off immediately.

Why Design Systems for Niche SaaS Require a Different Approach?

Discover 15+ best font pairing generator and tools to avoid typography mistakes. Explore free font pairing website for professional designs.

15+ Best Font Pairing Generator To Avoid Typography Mistakes

Find 20 website animation examples, from hero scroll effects, parallax to 3D rotations, micro-interactions & more. Boost user engagement with top examples!

Apple’s “The Illusion of Thinking” Is a Wake-Up Call for AI Industry

Aakash Jethwani

Inspire the next generation of designers

Read Next