The Courage to Build What’s Real
What Apple’s new paper reveals about AI, and why it’s a call to founders to go deeper, not just faster
“It’s not what you look at that matters, it’s what you see.”
— Henry David Thoreau
Recently, Apple quietly published a research paper called The Illusion of Thinking, which might be one of the most honest pieces of AI research we’ve seen in a long time.
Its core message is that today’s most powerful AI systems—the ones that look like they’re “thinking”—aren’t actually thinking. They simulate reasoning and generate long, impressive outputs. But when complexity increases, they fall apart. Essentially, they think less when they should be thinking more.
The illusion is in the performance. And if we mistake it for real understanding, we risk building a future on sand.
So, this week’s letter is less about the paper's technical details and more about what I believe to be its broad message and what this moment means for founders and builders. While some focus on chasing hype cycles, the future still needs real builders. And you don’t need to be an AI expert to answer this call. You just need to know what’s real—and be willing to build it.
In simple terms, Apple’s The Illusion of Thinking found that when AI systems are given easy problems, they can sometimes “overthink”—adding too many unnecessary steps. When given complex problems, they tend to break down. They stop thinking altogether or go off track. And even when you give them the right algorithm to follow, they often fail to execute it. They don’t really reason. They generate.
The takeaway from Apple’s paper is a powerful reminder that we may be overestimating how close we are to real artificial general intelligence (AGI). Or worse—designing our businesses, investments, and policies based on the illusion that it’s already here.
This is actually an opportunity to drive clarity. It means there’s still room to build what’s true.
Reading this study brought me back to my own journey as a founder. In 2013, I left Israel and moved to London to start my first venture, Kano. We believed every child should be able to build and shape technology, not just consume it.
We turned that belief into a computer kit that let kids make their own computer, code with it, and invent. Eventually, we built a platform used in over 100 countries and partnered with institutions like Microsoft, Warner Bros., and Disney. But what drove us wasn’t the market—it was the clarity of a problem that felt real, urgent, and unsolved.
And that same clarity is missing in so much of the AI discourse today.
The best founders I know—whether first-timers or repeat builders—aren’t chasing trends. They’re trying to solve real problems that matter to humans. Illusions don’t fool them. They’re building systems that align with reality.
If you’re building something or exploring building something, like me and many others, Apple’s paper should change how we approach our work.
It reminds us that not all that glitters is gold. The longest AI output is not the smartest, and the flashiest startup may not be the most enduring.
It reminds us that depth matters. If your product or company relies on the idea that LLMs can “think,” you may want to be careful.
It reminds us that we need new architectures—human and technical. AI alone won’t solve our biggest challenges. We need new ways to think, build, and reason with technology, not just through it.
So, what should founders do with this takeaway? Here’s how I’m thinking about it, and how you might too:
1. Build systems, not just wrappers
It is tempting to create an AI agent or tool to complement GPT. But now we know the limits. The future might belong to founders who combine AI with real-world context, tools for logic, memory, or symbolic reasoning, and interfaces that reveal, not obscure, the model’s confidence and limitations
Think less about what your AI can say, and more about what it can solve.
2. Design for augmentation, not automation
The best AI tools won’t replace people. They’ll amplify them.
Build tools that help doctors diagnose faster, not replace their judgment. Build tools that help teachers tailor learning, not eliminate the classroom. Build tools that help humans do more of what makes them human.
3. Embrace the unknown
Apple’s paper shows that we’re still far from understanding machine or human intelligence. That’s humbling. But it’s also freeing. It means we don’t need to have all the answers. We need a problem worth solving, a truth worth testing, and a willingness to learn.
Paul Graham has written:
“Live in the future, then build what’s missing.”
But to do that, we have to see the present clearly. Right now, it is filled with shiny illusions. Our job as founders is to look deeper.
This is about what kind of builders we want to be. This isn’t just about AI. It’s about building a more human future. Otherwise, what’s the damn point if technology if not to advance the human condition?
When Melanie Perkins founded Canva, she wasn’t a design expert. She was a university student frustrated by how hard it was to make a clean presentation. She didn’t wait for credentials—she built the simplest, most accessible design tool in the world.
When Demis Hassabis started DeepMind, he wasn’t just interested in AI. He was obsessed with understanding intelligence itself. He focused on long-term breakthroughs. His team cracked protein folding with AlphaFold—solving a 50-year scientific mystery and opening the door to radical medical advances.
These founders were responding to a world they knew needed to change, and that’s the posture we need now—not blind optimism or cynicism, but clarity, courage, and massive action.
We’re at a fork in the road. One path is filled with demos, sizzle reels, and promises of thinking machines that will do it all for us. That path leads to fragility, to trust misplaced in systems that can’t deliver, and to failure in moments of real complexity.
The other path is quieter, slower, but far more profound. It’s the path of founders who know that real intelligence—human or artificial—is hard won, that truth takes time, and who are willing to build anyway.
If you’re reading this and thinking, “Maybe I should wait until the models get better,” don’t. What the world needs now are not perfect systems, but courageous builders who know the difference between illusion and insight.
Thanks for reading,
Yon
👋 Hello! My mission with Beyond with Yon is to ignite awareness, inspire dialogue, and drive innovation to tackle humanity's greatest existential challenges. Join me on the journey to unf**ck the future and transform our world.