New post: If You're Gonna Vibe, Vibe With Your Boots On
If You're Gonna Vibe, Vibe With Your Boots On
How Intuition Is Made
The Advantage No One Talks About
AI in coding gives you one huge advantage that no one talks about, if you are willing to put in the work. I'm not talking about using AI to learn or having it explain topics, those are all well and good. No, I'm talking about something much more practical.
Remember that old show on Discovery, How It's Made? Ever wonder how we got there? All those machines just to make a belt buckle, a nail, or whatever they were showing. I remember this other show I watched religiously as a kid, Modern Marvels, oh it was my jam.
AI automates a lot of the hand coding, yes. You can never touch a line manually and still build something great. But what separates the successes from the failures is simple. We call it operations, processes, whatever. In less fancy words: why it's made the way it is.
All deep understanding starts by asking why, multiple times.
Why Does AI Want What It Wants?
When you see AI build, do you wonder why it makes those choices? I'm not talking about the attention scores in the representative matrix and whatnot, transformer architecture explanation. I'm talking about the abstract concept of want, the same way a bee wants to find a flower.
Why does AI want to use certain patterns? What context causes what outcome, and why?
Some Asshole Named Deming
W. Edwards Deming was the father of manufacturing as we know it, and he had a saying: if you control the process, you control the results.
Sounds mechanistic, rigid. Hell, if you've worked in the industry, you've probably felt like a cog in the machine, or like you're polishing the fangs of an eldritch beast.
Because they took this deterministic, quasi-fatalist rigidity to heart. Deming had a point, an incomplete one. He viewed people as single functions, as operators with one capability. If you don't wanna read me yapping so informally, go read The Toyota Way. You will reach similar conclusions in a less jarring linguistic experience.
The disciples of this school of thought didn't realize the potential of humans to improve the processes. Then comes in Kaizen, the continuous improvement, and the idea that the people on the front lines have the most actionable insights. They just need to be aggregated. Call it signal, since that's the Twitter word of this month.
The Gap
When you do agentic coding, you are in a novel situation. On the one hand you could treat AI as a single function entity. That's how tab completion in Cursor used to be a thing. Programmers were thinking at the function and method level, letting AI do narrow scope, well defined tasks.
But the potential for more is clear.
AI doesn't behave like a person. Not yet. There is no initiative, no permanent memory, no continuous learning. And though many novel architectures are gaining traction, transformers are what most coding agents are based on.
Software Is Not a Smoke Detector
In manufacturing, these processes are always built to be efficient, cost effective. And in production code, it was always a different beast. If you make smoke detectors, the design comes in planned cycles. If the battery component is upgraded then there is retooling and the show goes on.
But software, it's not like that.
At its worst it becomes a hodge podge of horrors that ensnare the mind and keep the therapists flush with cash. At its best, you get continuous iteration.
And the difference between panic-attack-inducing codebases and tolerable ones is process. Not just a process, but the creation and improvement of one.
The idea is simple, same as any other problem: you analyze, propose a solution, execute, evaluate, and then iterate until it is done.
What I Mean By Process
Let me clarify. What I mean by process is just how you work. I can give you concrete examples.
My Claude Code setup has a hook that gives Claude the current time. Thus if it's an ungodly hour of the night, it might check in and ask if I'm awake. This is fair because hyperfocus is a real danger. Tunnel vision is a form of subtle incapacitation. Subscribe and hit the bell so you don't miss that one lol.
Another thing I do is hooks on session start that give Claude global context about my working preferences. No matter the project, there is a million weird things I do, and they work, because I have tested them, adapted, rebuilt, and tested again.
Operational know-how and intuition comes from that.
Here Is How
If you are expecting me to give you hooks, prompts, MCP servers, you are missing the point.
Vibecoding gives us a unique opportunity. We are the factory worker, the floor supervisor, the engineer, and the stakeholder all melded into one. We experience the process and its ups and downs at every step. This is powerful.
We have the why. Now you need to build your how.
The first step isn't reading books or watching six sigma tutorials. The first step is grabbing a notebook, and no joke, you are gonna talk to yourself as a caveperson would.
I am 100% serious.
Ooga Booga Debugging
"Ooga booga me want to add Discord auth but AI break login, why? Why AI changing login like that?"
Sounds silly. Works amazing.
By the time you are laughing your ass off, you have a better understanding of where the pitfalls are. By reducing every difficulty to cave speak, you are removing everything that is superfluous.
So grab your notebook. Go do the funny voice and write down: - What broke last time you vibecoded? - What did AI do that confused you?
People will yap about context engineering, and that's important. AI needs context. But AI also needs structure, procedures, process.
You don't have to be a six sigma black belt to know that if the AI hallucinates what framework to use, you give an explicit instruction to use the correct one.
Go through your usual flow of work. See what is bogging you down. And then when you have an idea of what to change, do one by one. Otherwise you can't really separate cause and effect. If you test 5 improvements at once, which one "moved the needle"? Oh, Twitter loves that phrase too.
Your Process, Not Mine
You are going to build your own process, and it won't be the same as mine. It cannot be. Each team, each individual works differently.
That's why enterprise work can be exhausting. You play a character that someone defined in a manual 3 years ago and no one bothered to check if it was human.
But you make your own processes now. It's not just vibes anymore. It's vibes with purpose.
We humans have intuition, shaped by experience. Every time you work on process improvement, you hone that. You become better.
Do not fall into premature optimization as a trap. Take the pain as it comes, step by step.
The Compound Effect
When you make your own process, you are constantly testing it. For efficiency, ergonomics, ease of use, quality. And you have to test it for a significant amount of time.
There is no shortcut here. You just have to jump in, talk like a caveperson, and learn.
Maybe 3 months down the line you look back and wonder why you didn't have a process at all. One year down the line and you realize just how much more focus can go into achieving your goals when the operational side is handled.
Twitter loves the phrase "compounding" but it fits here. Commit to the work and it will free your mind from the minutiae while keeping the skills that are relevant to the work nice and sharp.
Commit to the Work
Akio Toyoda once said something about "purposeful passion and action." Go read about him if you want.
But above all else, go and get started.
Purpose builds process. Process empowers purpose. Intuition arises.
Commit to the work.