I'm Petrified of Having Chauffeur Knowledge
Published: November 18, 2025 • 6 min read
I was listening to a podcast by Scott D Clary on Audible yesterday, and he told a story that hasn't left my mind since.
The Max Planck Story
Max Planck, after receiving a Nobel Prize in Physics for his groundbreaking work in quantum mechanics, toured all around Germany giving lectures. He was always driven around by a specific chauffeur who had to sit through every single lecture.
After sitting through dozens and dozens of the exact same lecture, the chauffeur had basically memorized it. He could recite it word for word, even without having a deep understanding of quantum mechanics itself.
One day, for whatever reason, the chauffeur gave the lecture in place of Max Planck. He delivered it flawlessly. But when it came time for questions, someone in the audience asked something complex about the underlying physics.
The chauffeur laughed and said, "That question is so simple that even my chauffeur could answer it," pointing at Max Planck sitting in the audience.
He had the words. He had the performance. But he didn't have the knowledge.
Why This Terrifies Me
I don't want to be the chauffeur of my own code.
I don't want to be able to recite the implementation without understanding the mechanics beneath it. I don't want to build applications that I can demo but can't debug when something breaks in a way I've never seen before.
And right now, I'm worried that's exactly where AI-assisted development is taking me.
The Old Way: Trial, Error, and Deep Understanding
When I first learned how to write code and started building applications for school and professional work, I had to go through a period of trial and error. My solutions failed. I'd iterate through multiple methods of solving a problem until I arrived at the result I wanted.
This process of trying, failing, and changing strategies forced me to think. I was deeply engaged with my work. I understood everything I built down to the very core of the project.
That frustration of debugging for hours? That was the price of real understanding.
The New Reality: Speed Without Struggle
That process of trial, error, and constant iteration has been largely replaced by AI.
If you've been following my blog posts, you know that I've worked a lot with AI and I'm deeply excited by the opportunities it gives me, especially when it comes to creating and building. I want to use AI to contribute good to the world, whether through the projects I create or the prompts I design.
But every once in a while, I am fascinated by the power of AI in a way that makes me wonder: should I be excited about this, or should I be scared?
I took the French Writing Playground from version 1 to version 2 in only 7 days.
That's insane.
I take full credit for every single idea implemented in that application. The architecture decisions, the feature priorities, the user experience choices. Those were all mine. But I didn't struggle during the building process the way I used to. I described what I wanted and watched it come to life after multiple iterations of prompting, not building.
So now I fear I have chauffeur knowledge about the app I built.
But Wait: Is Struggle Even Necessary?
Here's where my thinking gets complicated.
Why do I feel like I need to struggle to feel good or confident about the work I've built? Is this just school and society's conditioning? The idea that results achieved without suffering have no or little value?
Maybe the struggle should be different now.
As the barriers between ideas and creation lessen and lessen, maybe what matters is the speed at which I can execute using the tools available, without sacrificing quality, while ensuring that the products I build solve real problems.
Maybe the value isn't in the struggle itself. Maybe the value is in the thinking, the decision-making, the problem identification.
But even as I write that, something feels off. I don't fully believe it yet.
My First Attempt to Combat This
One of the reasons I created the Q&A series (beginner, intermediate, and senior developer questions about the French Writing Playground) was to force myself to think from the perspective of someone who wants to know more about the decisions made during development.
Writing those posts required me to articulate why I chose PostgreSQL triggers over WebSockets, why I used JWT sessions instead of database sessions, why I implemented four layers of validation before the OpenAI API call. I had to understand these decisions deeply enough to explain them clearly.
That helped. But I'm not sure it was enough.
Where This Leaves Me
As I continue to make sense of how I feel and reflect on this question, I'm committing to solving real problems using the depth of knowledge I have about AI blended with other areas I'm knowledgeable about.
My first step is a case study I'm now working on. I'll use LLM Instance Cloning, a concept I introduced in a previous blog post, to design a prompt that leverages the best features of LLM tools like Claude and ChatGPT to help students with Dyscalculia better understand mathematics.
This will probably take a while to finish. There will be a lot of trial and error, breaking and rebuilding, and guiding the model through countless iterations. I'll also need to do deep research to see the world through the eyes of a person with Dyscalculia as I train an LLM instance to help them understand concepts they might normally struggle with.
When this case study is ready, you'll be able to access it here.
The Uncomfortable Truth
Here's what I've realized writing this post: I don't have a clean answer.
I'm genuinely torn between "AI is letting me build at a speed I never thought possible" and "but am I actually becoming a better developer, or just a faster one?"
Maybe both can be true. Maybe I can be faster AND deeper, if I'm intentional about it.
But that intentionality is key. Without it, I'm just the chauffeur. Performing the lecture without understanding the physics.
As always, thanks for reading!