The Process of Building Matters More Than What AI Produces

Frequently articles I read are focused around the topic of AI. With the ever increasing trend of vibe coding, RAG, and advanced reasoning. As AI tools increase in value an important part of our lives decreases. The process.

Imagine you head to the gym. You’re lifting weights, focused on your exercise until you turn to see someone drive in a forklift. Carefully they position it under the barbell rack and lift the entire rack off the ground. They hop out, pump their fists in the air and shout, “I lifted the most weights! I’m the strongest one here!” Obviously this situation is ridiculous. It’s easy to tell that the goal of going to the gym is not getting the weight off the ground, it’s the process of lifting the weight that’s important.

Yet it’s becoming increasingly difficult for individuals to see that during the pursuit of an end product they are losing sight of the ever more important process. Simply, the journey before the destination.

Vibe Coding and the Illusion of Building

Vibe coding to me is the clearest example of this in the AI space. Programming is ever more accessible and yet embarrassingly inaccessible as tools begin to streamline the process. It’s insanely easy to create prototypes of apps thinking that AI is amazing only to realize when you productionalize things you have a massive code base with zero knowledge of how anything works. The important part of prototyping isn’t the product, it’s understanding the roadblocks, requirements and intricacies of the application.

I run into this with my own AI agents. I can have Claude scaffold an entire FastAPI service in minutes. But when something breaks at 2am and the logs make no sense, the person who prompted the code into existence and the person who wrote it by hand are in very different positions. One of them understands the system. The other one has a system.

There’s a version of this that’s even more subtle. You can vibe code something that works perfectly and still have learned nothing. The app runs. Users are happy. But you didn’t build the mental model that would let you extend it, debug it, or know when it’s doing something wrong. You got the weight off the ground but you didn’t get any stronger.

The Learning Problem

One area I’m increasingly anxious about is AI’s impact on learning. As students and lifelong learners alike begin to utilize AI to streamline their learning experience through things like reading, note-taking, and brainstorming, these tools are eroding the more important part. The journey itself.

The note you write and the note AI writes for you have very different outcomes. Your notes are likely not very well written, take a long time compared to AI, and may be difficult for another person to consume.

But that’s the point. The struggle of translating an idea from your head into words on a page is where understanding actually happens. When you wrestle with how to explain something you’re forced to confront the gaps in your own thinking. The AI-generated note is cleaner but it’s not yours. You didn’t earn it. And three weeks from now you won’t remember what it says because you never had to fight for it.

I have almost 4,000 notes in Obsidian. The ones I remember, the ones that actually changed how I think, are the messy ones I wrote myself at midnight. Not the polished summaries an AI could have generated in seconds.

This isn’t an argument against using AI for learning. I use it constantly. But I use it the way you’d use a sparring partner, not a ghostwriter. I write my thoughts first, then ask the AI to push back on them. That way the thinking still happens in my head. The AI just makes the thinking harder, which is exactly what good learning should do.

The Destination is a Mirage

Here’s what I keep coming back to. The destination, the finished product, the shipped app, the completed note, the polished report, those things feel like the goal. But they’re actually just evidence that the process happened. The value was created along the way, in the struggle and the iteration and the moments where you almost quit but figured it out instead.

AI is incredibly good at producing destinations. It’s terrible at producing journeys. And if we’re not careful we’ll optimize ourselves into a world where everything looks finished but nobody actually knows anything.

The forklift operator didn’t get stronger. Don’t be the forklift operator.

The Process of Building Matters More Than What AI Produces

Vibe Coding and the Illusion of Building

The Learning Problem

The Destination is a Mirage

See Also